a predecessor if it has a non-zero first index and the predecessor
ends with a struct index.

llvm-svn: 3982

llvm-svn: 3981

llvm-svn: 3980

llvm-svn: 3979

rather than in random order.  This causes LICM to be DRAMATICALLY more
effective. For example, on bzip2.c, it is able to hoist 302 loads and
2380 total instructions, as opposed to 44/338 before.  This  obviously
makes other transformations much more powerful as well!

llvm-svn: 3978

llvm-svn: 3977

llvm-svn: 3976

llvm-svn: 3975

llvm-svn: 3974

llvm-svn: 3973

llvm-svn: 3972

llvm-svn: 3971

llvm-svn: 3970

created here, simply by handling all implicit operands (which should
have been done anyway).

llvm-svn: 3969

callee does not return a value.

llvm-svn: 3968

handled like normal operands, not like other call arguments.

llvm-svn: 3967

since it is defined by the instruction.

llvm-svn: 3966

created here.  Instead they are created in LiveRangeInfo.cpp.  This
simplifies the code here quite a bit.

llvm-svn: 3965

needs the RegClass list to be passed in.

llvm-svn: 3964

llvm-svn: 3963

llvm-svn: 3962

Don't allow direct FP-to-uint conversion (must be eliminated by preselection).
Address arithmetic for arrays is now entirely 64-bit so no sign-ext needed.

llvm-svn: 3961

llvm-svn: 3960

llvm-svn: 3959

llvm-svn: 3958

llvm-svn: 3957

llvm-svn: 3956

llvm-svn: 3955

revision as I start to use it though.

llvm-svn: 3954

* Linearize list of directories so I don't have to keep dealing with
  wrapping problems.

llvm-svn: 3953

llvm-svn: 3952

 - Do not clone instructions then insert clone outside of loop.  Just move them.

llvm-svn: 3951

llvm-svn: 3950

llvm-svn: 3949

llvm-svn: 3948

llvm-svn: 3947

llvm-svn: 3946

has not yet been removed.

llvm-svn: 3945

- break-crit-edges pass does not invalidate loop-preheader pass.

llvm-svn: 3944

Checkin new loop-preheader insertion pass.

llvm-svn: 3943