inline void
LocalTrrkKernel
( UpperOrLower uplo,
T alpha, const DistMatrix<T,MC, STAR>& A,
const DistMatrix<T,STAR,MR >& B,
T beta, DistMatrix<T,MC, MR >& C )
{
#ifndef RELEASE
PushCallStack("LocalTrrkKernel");
CheckInput( A, B, C );
#endif
const Grid& g = C.Grid();
DistMatrix<T,MC,STAR> AT(g),
AB(g);
DistMatrix<T,STAR,MR> BL(g), BR(g);
DistMatrix<T,MC,MR> CTL(g), CTR(g),
CBL(g), CBR(g);
DistMatrix<T,MC,MR> DTL(g), DBR(g);
const int half = C.Height()/2;
ScaleTrapezoid( beta, LEFT, uplo, 0, C );
LockedPartitionDown
( A, AT,
AB, half );
LockedPartitionRight( B, BL, BR, half );
PartitionDownDiagonal
( C, CTL, CTR,
CBL, CBR, half );
DTL.AlignWith( CTL );
DBR.AlignWith( CBR );
DTL.ResizeTo( CTL.Height(), CTL.Width() );
DBR.ResizeTo( CBR.Height(), CBR.Width() );
//------------------------------------------------------------------------//
if( uplo == LOWER )
internal::LocalGemm( NORMAL, NORMAL, alpha, AB, BL, T(1), CBL );
else
internal::LocalGemm( NORMAL, NORMAL, alpha, AT, BR, T(1), CTR );
internal::LocalGemm( NORMAL, NORMAL, alpha, AT, BL, T(0), DTL );
AxpyTriangle( uplo, T(1), DTL, CTL );
internal::LocalGemm( NORMAL, NORMAL, alpha, AB, BR, T(0), DBR );
AxpyTriangle( uplo, T(1), DBR, CBR );
//------------------------------------------------------------------------//
#ifndef RELEASE
PopCallStack();
#endif
}
Symbol() {
refcount = 0;
DBR(delegated = 0);
}
