re PR target/39445 (failure of SSE2 tests)

gcc/

2009-03-12  H.J. Lu  <hongjiu.lu@intel.com>

PR target/39445
* config/i386/i386.c (ix86_expand_push): Don't set memory
alignment.

gcc/testsuite/

2009-03-12  H.J. Lu  <hongjiu.lu@intel.com>

PR target/39445
* gcc.target/i386/pr39445.c: New.

* gcc.target/i386/push-1.c: XFAIL.

From-SVN: r144823

re PR target/39327 (Incorrect addsub/unpck patterns in sse.md)

2009-03-12  H.J. Lu  <hongjiu.lu@intel.com>

PR target/39327
* config/i386/sse.md (avx_addsubv8sf3): Correct item bits.
(avx_addsubv4df3): Likewise.
(*avx_addsubv4sf3): Likewise.
(sse3_addsubv4sf3): Likewise.

From-SVN: r144819

[multiple changes]

2009-03-11  Andrew Haley  <aph@redhat.com>

* java/lang/natClassLoader.cc (_Jv_RegisterClassHookDefault): Clear
INTERPRETED access modifier.

2009-03-11  Tom Tromey  <tromey@redhat.com>

* link.cc (_Jv_Linker::find_field): Pass the field_type to
_Jv_CheckOrCreateLoadingConstraint, not the class that is
requesting the field.

From-SVN: r144818

re PR target/38824 (performance regression of sse code from 4.2/4.3)

2009-03-12  H.J. Lu  <hongjiu.lu@intel.com>

PR target/38824
* config/i386/i386.md: Compare REGNO on the new peephole2
patterns.

From-SVN: r144817

re PR debug/39432 (gdb.base/store.exp failures)

2009-03-12  Vladimir Makarov  <vmakarov@redhat.com>

PR debug/39432
* ira-int.h (struct allocno): Fix comment for calls_crossed_num.
* ira-conflicts.c (ira_build_conflicts): Prohibit call used
registers for allocnos created from user-defined variables.

From-SVN: r144812

re PR target/39181 (complex int arguments cause ICE)

PR target/39181
* config/spu/spu.c (spu_expand_mov): Handle invalid subregs
of non-integer mode as well.

From-SVN: r144811

* MAINTAINERS:  Move myself into the write after approval list.

From-SVN: r144810

gimplify.c (gimplify_call_expr): Don't set CALL_CANNOT_INLINE_P for functions for which the parameter types are...

* gimplify.c (gimplify_call_expr): Don't set CALL_CANNOT_INLINE_P
for functions for which the parameter types are unknown.

testsuite/
* gcc.dg/inline-33.c: New test.

From-SVN: r144801

vect-82.c: Combine dg-do and dg-require-effective-target into dg-skip-if.

* gcc.dg/vect/vect-82.c: Combine dg-do and
dg-require-effective-target into dg-skip-if.
* gcc.dg/vect/vect-83.c: Likewise.

From-SVN: r144800

Daily bump.

From-SVN: r144799

re PR target/39137 (-mpreferred-stack-boundary=2 causes lots of dynamic realign)

PR target/39137
* cfgexpand.c (get_decl_align_unit): Use LOCAL_DECL_ALIGNMENT
macro.
* defaults.h (LOCAL_DECL_ALIGNMENT): Define if not yet defined.
* config/i386/i386.h (LOCAL_DECL_ALIGNMENT): Define.
* config/i386/i386.c (ix86_local_alignment): For
-m32 -mpreferred-stack-boundary=2 use 32-bit alignment for
long long variables on the stack to avoid dynamic realignment.
Allow the first argument to be a decl rather than type.
* doc/tm.texi (LOCAL_DECL_ALIGNMENT): Document.

* gcc.target/i386/stackalign/longlong-1.c: New test.
* gcc.target/i386/stackalign/longlong-2.c: New test.

From-SVN: r144792

* g++.old-deja/g++.jason/thunk2.C: Skip on SPU.

From-SVN: r144785

re PR target/5362 (Undocumented target options)

        PR target/5362
        * config/mcore/mcore.opt: Remove deprecated m4align and m8align
        options.
        Add description to mno-lsim option.
        * config/mcore/mcore.h: Remove comment about deprecated m4align
        option.
        (TARGET_DEFAULT): Remove deprecated MASK_M8ALIGN.
        * doc/invoke.texi: Add description of mno-lsim and
        mstack-increment options.

From-SVN: r144783

re PR target/5362 (Undocumented target options)

        PR target/5362
        * config/fr30/fr30.opt: Document the -mno-lsim option.
        * doc/invoke.texi: Add descriptions of the FR30's -msmall-model
        and -mno-lsim options.

From-SVN: r144780

fold-const.c (fold_comparison): Only call fold_inf_compare if the mode supports infinities.

ChangeLog:

* fold-const.c (fold_comparison): Only call fold_inf_compare
if the mode supports infinities.

testsuite/ChangeLog:

* gcc.c-torture/execute/ieee/inf-3.c: New test.
* gcc.c-torture/execute/ieee/inf-2.c: Fix typo.

From-SVN: r144779

shared_ptr.h: Add include guards.

2009-03-11  Jonathan Wakely  <jwakely.gcc@gmail.com>

* include/bits/shared_ptr.h: Add include guards.
* include/tr1/shared_ptr.h: Likewise.

From-SVN: r144778

iostream: Fix doxygen link.

2009-03-11  Jonathan Wakely  <jwakely.gcc@gmail.com>

* include/std/iostream: Fix doxygen link.
* include/bits/forward_list.h: Fix doxygen markup.
* include/ext/vstring.h: Escape backslash in doxygen comment.
* include/bits/basic_string.h: Likewise.

From-SVN: r144777

Updated my email address

From-SVN: r144776

Move Ada ChangeLog entry to proper place.

From-SVN: r144775

trans.c (gnat_to_gnu): In range check processing...

ada/
* gcc-interface/trans.c (gnat_to_gnu) <case N_Slice>: In range
check processing, remove unintended TREE_TYPE walk on index type
to get to base type.

testsuite/
* gnat.dg/slice_enum.adb: New test.

From-SVN: r144774

MAINTAINERS: Remove myself as auto-vectorizer maintainer.

2009-03-11  Dorit Nuzman  <dorit@il.ibm.com>

        * MAINTAINERS: Remove myself as auto-vectorizer maintainer.

From-SVN: r144773

re PR debug/39086 (ICE in decl_ultimate_origin, at dwarf2out.c:5770 when compiling with -fno-tree-sra)

        PR debug/39086
        * tree-nrv.c (tree_nrv): Don't do this optimization if the front
        end already did.  Notice GIMPLE_CALL modifications of the result.
        Don't copy debug information from an ignored decl or a decl from
        another function.

From-SVN: r144772

faq.xml: Fix links to implementation status pages.

2009-03-10  Paolo Carlini  <paolo.carlini@oracle.com>

* doc/xml/faq.xml: Fix links to implementation status pages.

From-SVN: r144771

complex (operator+(const complex<>&, const _Tp&), [...]): Do not assume real() returns a reference (against DR 387).

2009-03-10  Paolo Carlini  <paolo.carlini@oracle.com>

* include/std/complex (operator+(const complex<>&, const _Tp&),
operator+(const _Tp&, const complex<>&),
operator-(const complex<>&, const _Tp&),
operator-(const _Tp&, const complex<>&)): Do not assume real()
returns a reference (against DR 387).
* testsuite/26_numerics/complex/dr387_2.cc: New.

From-SVN: r144770

Daily bump.

From-SVN: r144768

Remove myself as vectorizer maintainer

From-SVN: r144760

re PR testsuite/39422 (Failing SPU vectorizer testcases)

PR tree-optimization/39422
* gcc.dg/vect/costmodel/spu/costmodel-vect-76b.c: Move
constant array with static initialization to global memory.
* gcc.dg/vect/costmodel/spu/costmodel-vect-76c.c: Likewise.

From-SVN: r144754

pr39339.c: Bitfield sizes changed to avoid padding.

  * gcc.c-torture/execute/pr39339.c: Bitfield sizes changed to avoid padding.

From-SVN: r144752

re PR middle-end/37850 (infinite recursive call to __mulsc3 when multiplying not-constant complexs)

PR middle-end/37850
* libgcc2.c (__mulMODE3): Use explicit assignments to form the
result.
(__divMODE3): Likewise.

Co-Authored-By: Nathan Froyd <froydnj@codesourcery.com>
From-SVN: r144751

MAINTAINERS: Add myself as auto-vectorizer maintainer.

2009-03-10  Ira Rosen <irar@il.ibm.com>

* MAINTAINERS: Add myself as auto-vectorizer maintainer.

From-SVN: r144747

* libgcc/config/i386/t-cygming: Remove executable property.

From-SVN: r144742

Daily bump.

From-SVN: r144738

re PR c++/39371 (Incorrectly rejects switch((unsigned int)boolvar))

PR c++/39371
* semantics.c (finish_switch_cond): Don't call get_unwidened.
* decl.c (finish_case_label): Pass SWITCH_STMT_TYPE as 3rd argument
instead of TREE_TYPE (cond).

* g++.dg/opt/switch2.C: Add -w to dg-options.
* g++.dg/warn/Wswitch-1.C: Adjust expected warnings.
* g++.dg/warn/switch1.C: New test.
* g++.dg/other/switch3.C: New test.

From-SVN: r144732

re PR tree-optimization/39394 (ICE in copy_tree_body_r)

PR tree-optimization/39394
* gimplify.c (gimplify_type_sizes): Gimplify DECL_SIZE and
DECL_SIZE_UNIT of variable length FIELD_DECLs.

* gcc.c-torture/compile/pr39394.c: New test.

From-SVN: r144727

recog.c (verfiy_changes): Disallow renaming of hard regs in inline asms for register asm ("") declarations.

2009-03-09  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* recog.c (verfiy_changes): Disallow renaming of hard regs in
inline asms for register asm ("") declarations.

2009-03-09  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* gcc.target/s390/20090223-1.c: New testcase.

From-SVN: r144726

* fold-const.c (fold_unary): Fix comment.

From-SVN: r144723

Fix PR number reference in ChangeLog from 39409 to 39402.

From-SVN: r144720

re PR libfortran/39402 (gfortran 20090306: internal write of  0.0 with F0.3 gives  **)

2009-03-08  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libfortran/39402
* gfortran.dg/fmt_f0_1.f90: New test.

From-SVN: r144719

re PR libfortran/39402 (gfortran 20090306: internal write of  0.0 with F0.3 gives  **)

2009-03-08  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR libfortran/39409
* io/write_float.def (output_float): Handle F0.d formatting correctly
for any d when value is 0.0.

From-SVN: r144718

Daily bump.

From-SVN: r144716

re PR target/39361 (Many new neon test failures)

PR target/39361
* tree-inline.c (setup_one_parameter): Do replacement of const argument
by constant in SSA form.

From-SVN: r144713

re PR c++/39060 (ICE with lots of invalid member functions)

gcc/cp/

2009-03-08  H.J. Lu  <hongjiu.lu@intel.com>

PR c++/39060
* parser.c (cp_parser_late_parsing_default_args): Continue
the loop when cp_parser_assignment_expression returns
error_mark_node.

gcc/testsuite/

2009-03-08  H.J. Lu  <hongjiu.lu@intel.com>

PR c++/39060
* g++.dg/other/new1.C: Adjusted.
* g++.dg/parse/crash40.C: Likewise.
* g++.dg/parse/defarg12.C: Likewise.
* g++.dg/template/error15.C: Likewise.

* g++.dg/other/pr39060.C: New.

From-SVN: r144710

(Add missing PR reference)

2009-02-25  Laurent GUERBY  <laurent@guerby.net>

PR ada/39221
* a-teioed.adb (Expand): Fix Result overflow.

From-SVN: r144708

i386.c (ix86_maybe_switch_abi): Use named constants instead of magic numbers.

* config/i386/i386.c (ix86_maybe_switch_abi): Use named
constants instead of magic numbers.

From-SVN: r144706

Daily bump.

From-SVN: r144705

Correct filename.

From-SVN: r144702

re PR c++/39367 (ICE at tree-inline.c:1042 with -O)

        PR c++/39367
        * init.c (build_new_1): Don't use a VLA type.
        (build_vec_init): Handle getting a pointer for BASE.

From-SVN: r144697

re PR middle-end/38028 (eh failures on spu-elf)

PR middle-end/38028
* function.c (assign_parm_setup_stack): Use STACK_SLOT_ALIGNMENT to
determine alignment passed to assign_stack_local.
(assign_parms_unsplit_complex): Likewise.
* except.c (sjlj_build_landing_pads): Likewise.

From-SVN: r144696

re PR target/37520 (NO_DOLLAR_IN_LABEL should be defined on x86 targets)

2009-03-06  H.J. Lu  <hongjiu.lu@intel.com>

PR c++/37520
* cp-tree.h: Check NO_DOT_IN_LABEL before NO_DOLLAR_IN_LABEL
when mangling symbols.

From-SVN: r144692

Daily bump.

From-SVN: r144690

re PR c++/33492 (Broken diagnostic: 'must_not_throw_expr' not supported by dump_expr)

/cp
2009-03-06  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/33492
* error.c (dump_expr): Don't try to print THROW_EXPRs in full.

/testsuite
2009-03-06  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/33492
* g++.dg/other/error32.C: New.

From-SVN: r144686

re PR middle-end/39360 (ICE in referenced_var_lookup, at tree-dfa.c:563)

PR middle-end/39360
* tree-flow.h (add_referenced_var): Return bool instead of void.
* tree-dfa.c (add_referenced_var): Return result of
referenced_var_check_and_insert call.
* tree-inline.c (expand_call_inline): Call add_referenced_var instead
of referenced_var_check_and_insert.

* gcc.c-torture/compile/pr39360.c: New test.

From-SVN: r144683

re PR debug/39372 (Missing DW_AT_location for constructor static variable)

PR debug/39372
* dwarf2out.c (add_abstract_origin_attribute): Return
origin_die.
(gen_variable_die): Emit DW_AT_location on abstract static variable's
DIE, don't emit it if abstract origin already has it.
* tree-cfg.c (remove_useless_stmts_bind): GIMPLE_BINDs with any
BLOCK_NONLOCALIZED_VARS in its gimple_bind_block aren't useless.

* g++.dg/debug/dwarf2/static-local-var-in-ctor.C: New test.

From-SVN: r144682

genpreds.c: (needs_variable): Fix parentheses at variable name detection.

* genpreds.c: (needs_variable): Fix parentheses at variable name
detection.
(write_tm_constrs_h): Indent generated code.

From-SVN: r144679

Fix isr documentation for ARM

From-SVN: r144668

re PR debug/39387 (Wrong DW_AT_decl_line on DW_TAG_imported* for using directives in C+ functions)

PR debug/39387
* dwarf2out.c (dwarf2out_imported_module_or_decl_1): For IMPORTED_DECL
take locus from its DECL_SOURCE_LOCATION instead of input_location.

From-SVN: r144666

decl.c (record_builtin_java_type): Use canonicalized integer types.

* decl.c (record_builtin_java_type): Use canonicalized integer
types.

From-SVN: r144664

simplify.c (gfc_simplify_transfer): Zero-initialize the buffer.

* simplify.c (gfc_simplify_transfer): Zero-initialize the
buffer.

From-SVN: r144663

Daily bump.

From-SVN: r144656

bfin.c (bfin_discover_loop): When retrying fails, mark the loop as bad.

* config/bfin/bfin.c (bfin_discover_loop): When retrying fails, mark
the loop as bad.

From-SVN: r144645

re PR c++/38908 (Unexplained "'<anonymous>' is used uninitialized in this function" warning in cc1plus -m64)

        PR c++/38908
        * class.c (is_really_empty_class): New fn.
        * cp-tree.h: Declare it.
        * cp-objcp-common.c (cp_expr_size): Use it.

From-SVN: r144643

re PR debug/39379 (DW_TAG_imported* no longer emitted)

PR debug/39379
* tree-cfg.c (remove_useless_stmts_bind): Don't remove GIMPLE_BINDs
with blocks containing IMPORTED_DECLs in BLOCK_VARS.

* g++.dg/debug/dwarf2/imported-module-3.C: New test.
* g++.dg/debug/dwarf2/imported-module-4.C: New test.

From-SVN: r144640

i386.md (R8_REG, R9_REG): New constants.

* config/i386/i386.md (R8_REG, R9_REG): New constants.
* config/i386/i386.h (CONDITIONAL_REGISTER_USAGE): Use named
constants instead of magic numbers.
(HARD_REGNO_CALLER_SAVE_MODE): Ditto.
(QI_REG_P): Ditto.
* config/i386/i386.c (x86_64_int_parameter_registers): Ditto.
(x86_64_ms_abi_int_parameter_registers): Ditto.
(x86_64_int_return_registers): Ditto.
(ix86_expand_call): Ditto for clobbered_registers array.
(ix86_hard_regno_mode_ok): Ditto.
(x86_extended_QIreg_mentioned_p): Ditto.

From-SVN: r144638

re PR c++/13549 (Problem compiling Boost.Python test)

        PR c++/13549
        * semantics.c (perform_koenig_lookup): Handle TEMPLATE_ID_EXPR.
        * parser.c (cp_parser_postfix_expression): Call it for
        TEMPLATE_ID_EXPR.
        * tree.c (is_overloaded_fn): Look through TEMPLATE_ID_EXPR.
        (get_first_fn): Likewise.

From-SVN: r144636

Fix bugzilla references in ChangeLog for last recent checkins.

From-SVN: r144635

* cse.c (cse_insn): Fix loop to stop at VOIDmode.

From-SVN: r144634

Daily bump.

From-SVN: r144633

combine.c (gen_lowpart_for_combine): Use omode when generating clobber.

* combine.c (gen_lowpart_for_combine): Use omode when generating
clobber.

From-SVN: r144630

predcom-1.C: New test.

2009-03-04  Nathan Sidwell  <nathan@codesourcery.com>

* g++.dg/torture/predcom-1.C: New test.

From-SVN: r144629

re PR rtl-optimization/39235 (get_simple_loop_desc returns uninitialized memory)

PR39235:
* loop-iv.c (get_simple_loop_desc): Use XCNEW.

From-SVN: r144628

graphite.c (nb_reductions_in_loop): Update simple_iv arguments.

* graphite.c (nb_reductions_in_loop): Update simple_iv arguments.

From-SVN: r144627

re PR tree-optimization/39362 (ICE failing to coalesce ab SSA names)

2009-03-04  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/39362
* tree-ssa-sccvn.c (visit_use): Stores and copies from SSA_NAMEs
that occur in abnormal PHIs should be varying.

* g++.dg/torture/pr39362.C: New testcase.

From-SVN: r144626

add DR note

From-SVN: r144619

re PR c++/9634 ([DR224] Injected class name as qualifier should not make the name dependent)

        PR c++/9634
        PR c++/29469
        PR c++/29607
        * decl.c (make_typename_type): Do look inside currently open classes.
        * parser.c (cp_parser_lookup_name): Likewise.
        (cp_parser_template_name): Likewise.
        * pt.c (dependent_scope_p): New function.
        * cp-tree.h: Declare it.
        * class.c (currently_open_class): Return fast if T isn't a class.

Co-Authored-By: Giovanni Bajo <giovannibajo@gcc.gnu.org>
From-SVN: r144618

tree-scalar-evolution.c (analyze_scalar_evolution_in_loop): Extend comments.

* tree-scalar-evolution.c (analyze_scalar_evolution_in_loop):
Extend comments.
(simple_iv):  Take loop as an argument instead of statement.
* tree-scalar-evolution.h (simple_iv): Declaration changed.
* tree-ssa-loop-niter.c (number_of_iterations_exit): Update calls
to simple_iv.
* tree-ssa-loop-ivopts.c (determine_biv_step, find_givs_in_stmt_scev):
Ditto.
* tree-parloops.c (loop_parallel_p, canonicalize_loop_ivs): Ditto.
* matrix-reorg.c (analyze_transpose): Ditto.
* tree-data-ref.c (dr_analyze_innermost): Ditto.
* tree-vect-analyze.c (vect_analyze_data_refs): Ditto.
* tree-predcom.c (ref_at_iteration): Ditto.
* tree-ssa-loop-prefetch.c (idx_analyze_ref): Ditto.

From-SVN: r144615

re PR testsuite/39357 (gcc.dg/vect/vect-iv-6.c)

PR testsuite/39357
* gcc.dg/vect-iv-6.c (dg-require-effective-target): Change from
vect_int to vect_int_mult.

From-SVN: r144614

re PR tree-optimization/39358 (Wrong aliasing warning with lists)

2009-03-04  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/39358
* tree-ssa-structalias.c (do_sd_constraint): Fix check for
escaped_id and callused_id.
(solve_graph): Likewise.

* g++.dg/warn/Wstrict-aliasing-bogus-escape-2.C: New testcase.
* g++.dg/warn/Wstrict-aliasing-bogus-escape-3.C: Likewise.

From-SVN: r144602

* es.po: Update.

From-SVN: r144600

re PR tree-optimization/39339 (SRA miscompilation of vte)

2009-03-04  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/39339
* tree-sra.c (try_instantiate_multiple_fields): Make it
no longer ICE on the above.

* gcc.c-torture/execute/pr39339.c: New testcase.

From-SVN: r144598

emit-rtl.c (adjust_address_1): Reduce offset to a signed value that fits within Pmode.

* emit-rtl.c (adjust_address_1): Reduce offset to a signed value
that fits within Pmode.

testsuite:
* gcc.c-torture/compile/20090303-1.c,
gcc.c-torture/compile/20090303-2.c: New tests.

From-SVN: r144595

Daily bump.

From-SVN: r144594

* es.po: Update.

From-SVN: r144591

re PR middle-end/10109 (documentation of target macro LIBCALL_VALUE)

PR middle-end/10109
* tm.texi (LIBCALL_VALUE): Update description.

From-SVN: r144586

re PR middle-end/34443 (The GCC manual about section variable attribute is incorrect)

PR middle-end/34443
* doc/extend.texi (section): Update description.

From-SVN: r144582

re PR tree-optimization/39345 (ICE in copy_tree_body_r, at tree-inline.c:1020)

2009-03-03  H.J. Lu  <hongjiu.lu@intel.com>

PR middle-end/39345
* tree-inline.c (remapped_type): New.
(can_be_nonlocal): Call remapped_type instead of remap_type.

From-SVN: r144581

Cleanup new JDK directory support in GCJ.

2009-03-03  Andrew John Hughes  <ahughes@redhat.com>

* Makefile.am:
Remove dangling src.zip and javac symlinks.
Fix java.security symlink to use toolexeclibdir
as classpath/resource/Makefile.am does.
* Makefile.in,
* configure: Regenerated.
* configure.ac: Drop hard-coded prefix,
thus removing sdk_dir and making jre_dir="jre".
Map x86_64 to amd64 as used by OpenJDK.
* gcj/Makefile.in,
* include/Makefile.in,
* testsuite/Makefile.in: Regenerated.

From-SVN: r144576

re PR fortran/39354 (bad codegen for openmp atomics (Intel64 Fortran logical ops))

PR fortran/39354
* gimplify.c (goa_stabilize_expr): Handle tcc_comparison,
TRUTH_ANDIF_EXPR and TRUTH_ORIF_EXPR.

* gfortran.dg/gomp/pr39354.f90: New test.

From-SVN: r144575

re PR tree-optimization/39272 (-D_FORTIFY_SOURCE=2 no longer warns with new glibc headers)

2009-03-03  Richard Guenther  <rguenther@suse.de>

PR middle-end/39272
* tree.c (tree_nonartificial_location): New function.
* tree.h (tree_nonartificial_location): Declare.
* builtins.c (expand_builtin_memory_chk): Provide location
of the call location for artificial function pieces.
(maybe_emit_chk_warning): Likewise.
(maybe_emit_sprintf_chk_warning): Likewise.
(maybe_emit_free_warning): Likewise.
* expr.c (expand_expr_real_1): Likewise.

From-SVN: r144573

re PR tree-optimization/39343 (Wrong result for __builtin_object_size (x, 1))

PR tree-optimization/39343
* tree-ssa-ccp.c (maybe_fold_offset_to_address): Don't check if
COMPONENT_REF t has ARRAY_TYPE.

* gcc.dg/pr39343.c: New test.

From-SVN: r144571

re PR tree-optimization/39248 (FAIL: gcc.dg/vect/vect-complex-1.c)

PR tree-optimization/39248
* gcc.dg/vect/vect-complex-1.c: Add attribute aligned
to the arrays.
* gcc.dg/vect/vect-iv-6.c: Don't expect to fail to vectorize
on targets without vector misalignment support.
* lib/target-supports.exp
(check_effective_target_vect_short_mult): Add
check_effective_target_arm32.

From-SVN: r144569

re PR middle-end/39335 (ICE in GCC 4.4 with -O[123] -floop-interchange)

2009-03-02  Sebastian Pop  <sebastian.pop@amd.com>

PR middle-end/39335
* tree-parloops.c (canonicalize_loop_ivs): Call fold_convert
when the type precision of the induction variable should be
larger than the type precision of nit.
(gen_parallel_loop): Update use of canonicalize_loop_ivs.
* graphite.c (graphite_loop_normal_form): Same.
* tree-flow.h (canonicalize_loop_ivs): Update declaration.

* testsuite/gcc.dg/graphite/pr39335_1.c: New.
* testsuite/gcc.dg/graphite/pr39335.c: New.

From-SVN: r144564

fix ChangeLog date

From-SVN: r144561

Daily bump.

From-SVN: r144560

i386.md (ST?_REG, MM?_REG): New constants.

* config/i386/i386.md (ST?_REG, MM?_REG): New constants.
(*call_1_rex64_ms_sysv): Use named constants instead of magic
numbers to describe clobbbered registers.
(*call_value_0_rex64_ms_sysv): Ditto.
* config/i386/mmx.md (emms): Ditto.
(femms): Ditto.

From-SVN: r144554

mips.c (mips_mdebug_abi_name): Fix the handling of ABI_64.

gcc/
* config/mips/mips.c (mips_mdebug_abi_name): Fix the handling
of ABI_64.

From-SVN: r144552

spu.c (TARGET_SECTION_TYPE_FLAGS): Define.

* config/spu/spu.c (TARGET_SECTION_TYPE_FLAGS): Define.
(spu_section_type_flags): New function.

From-SVN: r144550

pthread4.cc: Move...

2009-03-02  Benjamin Kosnik  <bkoz@redhat.com>

* testsuite/thread/pthread4.cc: Move...
* testsuite/21_strings/basic_string/pthread4.cc: ...here.
* testsuite/thread/pthread5.cc: Move...
* testsuite/23_containers/list/pthread5.cc: ...here.
* testsuite/thread/pthread6.cc: Move...
* testsuite/23_containers/map/pthread6.cc: ...here.
* testsuite/thread/pthread7-rope.cc: Move...
* testsuite/ext/rope/pthread7-rope.cc: ...here.
* testsuite/thread/guard.cc: Move...
* testsuite/18_support/pthread_guard.cc: ...here.
* testsuite/thread/18185.cc: Move...
* testsuite/21_strings/basic_string/pthread18185.cc: ...here.
* testsuite/thread/pthread1.cc: Move...
* testsuite/23_containers/list/pthread1.cc: ...here.
* testsuite/thread/pthread2.cc: Move...
* testsuite/27_io/basic_ofstream/pthread2.cc: ...here.
* testsuite/thread/pthread3.cc: Move...
* testsuite/27_io/basic_ostringstream/pthread3.cc: ...here.

* testsuite/libstdc++-dg/conformance.exp: Remove thread directory.
* testsuite/thread: Remove.

From-SVN: r144547

pthread4.cc: Move...

2009-03-02  Benjamin Kosnik  <bkoz@redhat.com>

* testsuite/thread/pthread4.cc: Move...
* testsuite/21_strings/basic_string/pthread4.cc: ...here.
* testsuite/thread/pthread5.cc: Move...
* testsuite/23_containers/list/pthread5.cc: ...here.
* testsuite/thread/pthread6.cc: Move...
* testsuite/23_containers/map/pthread6.cc: ...here.
* testsuite/thread/pthread7-rope.cc: Move...
* testsuite/ext/rope/pthread7-rope.cc: ...here.
* testsuite/thread/guard.cc: Move...
* testsuite/18_support/pthread_guard.cc: ...here.
* testsuite/thread/18185.cc: Move...
* testsuite/21_strings/basic_string/pthread18185.cc: ...here.
* testsuite/thread/pthread1.cc: Move...
* testsuite/23_containers/list/pthread1.cc: ...here.
* testsuite/thread/pthread2.cc: Move...
* testsuite/27_io/basic_ofstream/pthread2.cc: ...here.
* testsuite/thread/pthread3.cc: Move...
* testsuite/27_io/basic_ostringstream/pthread3.cc: ...here.

From-SVN: r144546

return-4.c: Scan andl/andq instead of andl.

2009-03-02  H.J. Lu  <hongjiu.lu@intel.com>

* gcc.target/i386/stackalign/return-4.c: Scan andl/andq instead
of andl.
* gcc.target/i386/stackalign/return-5.c: Likewise.
* gcc.target/i386/stackalign/return-6.c: Likewise.

From-SVN: r144544

i386.h (CONDITIONAL_REGISTER_USAGE): Do not copy reg_class_contents of FLOAT_REGS into a temporary.

* config/i386/i386.h (CONDITIONAL_REGISTER_USAGE): Do not copy
reg_class_contents of FLOAT_REGS into a temporary.

From-SVN: r144542