This reverts commit 2b52880592a525cfe04d8f9008a35da8c2ea94c3.

Needs review.

llvm-svn: 185806

llvm-svn: 185805

I was originally going to use MVC for memmove too, but that's less of
a clear win.  Remove some accidental left-overs in the previous commit.

llvm-svn: 185804

llvm-svn: 185803

Use MVC for memcpy in cases where a single MVC is enough.  Using MVC is
a win for longer copies too, but I'll leave that for later.

llvm-svn: 185802

llvm-svn: 185801

llvm-svn: 185800

llvm-svn: 185799

This is mostly Doxygen formatting, but also updates some C++0x references
to C++11 and clarifies some wording.

llvm-svn: 185798

FIXME: Could we introduce another feature for it?
llvm-svn: 185797

It would emit @llvm.memcpy with "-triple x86_64-(mingw32|win32)" and had been failing since Nick's r185735.

  ; Function Attrs: nounwind
  declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly, i64, i32, i1) #1

llvm-svn: 185796

AST/SourceLocationTest.cpp: Appease MS hosts to suppress CXXUnresolvedConstructExpr.SourceRange, for now.

FIXME: It could pass if MS-compatible mode were disabled with Args.push_back("-fno-delayed-template-parsing").
llvm-svn: 185795

This fixes a bug (found by llvm-stress) in
DAGTypeLegalizer::PromoteIntRes_BUILD_VECTOR where it assumed that the result
type would always be larger than the original operands. This is not always
true, however, with boolean vectors. For example, promoting a node of type v8i1
(where the operands will be of type i32, the type to which i1 is promoted) will
yield a node with a result vector element type of i16 (and operands of type
i32). As a result, we cannot blindly assume that we can ANY_EXTEND the operands
to the result type.

llvm-svn: 185794

llvm-svn: 185793

llvm-svn: 185792

llvm-svn: 185791

llvm-svn: 185790

Introduce a typedef for the type of NewlyDeducedPacks to avoid repeating the small size of the inner SmallVector.

llvm-svn: 185789

llvm-svn: 185788

llvm-svn: 185787

llvm-svn: 185786

llvm-svn: 185785

llvm-svn: 185784

This fixes an oversight that Intrinsic::nearbyint was not being mapped to
ISD::FNEARBYINT (thus fixing the over-optimistic cost we were assigning to
nearbyint calls for some targets).

llvm-svn: 185783

llvm-svn: 185782

This reverts commit 01f8d579f7672872324208ac5bc4ac311e81b22e.

llvm-svn: 185781

llvm-svn: 185780

Sema::MergeFunctionDecl attempts merging two decls even if the old decl
is invalid.  This can lead to interesting circumstances where we
successfully merge the decls but the result makes no sense.

Take the following for example:

template <typename T>
int main(void);

int main(void);

Sema will not consider these to be overloads of the same name because
main can't be overloaded, which means that this must be a redeclaration.

In this case the templated decl is compatible with the non-templated
decl allowing the Sema::CheckFunctionDeclaration machinery to move on
and do bizarre things like setting the previous decl of a non-templated
decl to a templated decl!

The way I see it, we should just bail from MergeFunctionDecl if the old
decl is invalid.

This fixes PR16531.

llvm-svn: 185779

llvm-svn: 185778

llvm-svn: 185777

pointer arguments.

llvm-svn: 185776

since it has no callers today.

llvm-svn: 185775

This is a complete re-write if the bottom-up vectorization class.
Before this commit we scanned the instruction tree 3 times. First in search of merge points for the trees. Second, for estimating the cost. And finally for vectorization.
There was a lot of code duplication and adding the DCE exposed bugs. The new design is simpler and DCE was a part of the design.
In this implementation we build the tree once. After that we estimate the cost by scanning the different entries in the constructed tree (in any order). The vectorization phase also works on the built tree.

llvm-svn: 185774

Added testcases.

llvm-svn: 185773

case inspired by a stackoverflow question.

llvm-svn: 185772

* Fix up \brief documentation;
* Update C++0x references to C++11;
* Doxygen formatting: bulleted lists start with a single hyphen, not two;
* Fix a typo, "assosiate" -> "associate".

llvm-svn: 185771

Upon further reflection, the alias analysis part of r185764 is not a safe
change.

llvm-svn: 185770

[objc-arc] Teach the ARC optimizer that objc_sync_enter/objc_sync_exit do not modify the ref count of an objc object and additionally are inert for modref purposes.

llvm-svn: 185769

llvm-svn: 185768

llvm-svn: 185767