void L1DistanceMatrix(direction_t dirA, direction_t dirB, T alpha,
const El::ElementalMatrix<T> &APre, const El::ElementalMatrix<T> &BPre,
T beta, El::ElementalMatrix<T> &CPre) {
if (dirA == base::COLUMNS && dirB == base::COLUMNS) {
// Use a SUMMA-like routine, with C as stationary
// Basically an adaptation of Elementals TN case for stationary C.
const El::Int m = CPre.Height();
const El::Int n = CPre.Width();
const El::Int sumDim = BPre.Height();
const El::Int bsize = El::Blocksize();
const El::Grid& g = APre.Grid();
El::DistMatrixReadProxy<T, T, El::MC, El::MR> AProx(APre);
El::DistMatrixReadProxy<T, T, El::MC, El::MR> BProx(BPre);
El::DistMatrixReadWriteProxy<T, T, El::MC, El::MR> CProx(CPre);
auto& A = AProx.GetLocked();
auto& B = BProx.GetLocked();
auto& C = CProx.Get();
// Temporary distributions
El::DistMatrix<T, El::STAR, El::MC> A1_STAR_MC(g);
El::DistMatrix<T, El::STAR, El::MR> B1_STAR_MR(g);
A1_STAR_MC.AlignWith(C);
B1_STAR_MR.AlignWith(C);
El::Scale(beta, C);
for(El::Int k = 0; k < sumDim; k += bsize) {
const El::Int nb = std::min(bsize,sumDim-k);
auto A1 = A(El::IR(k,k+nb), El::IR(0,m));
auto B1 = B(El::IR(k,k+nb), El::IR(0,n));
A1_STAR_MC = A1;
B1_STAR_MR = B1;
L1DistanceMatrix(base::COLUMNS, base::COLUMNS, alpha,
A1_STAR_MC.LockedMatrix(), B1_STAR_MR.LockedMatrix(),
T(1.0), C.Matrix());
}
}
// TODO the rest of the cases.
}
inline void
internal::GemmTNC
( Orientation orientationOfA,
T alpha, const DistMatrix<T,MC,MR>& A,
const DistMatrix<T,MC,MR>& B,
T beta, DistMatrix<T,MC,MR>& C )
{
#ifndef RELEASE
PushCallStack("internal::GemmTNC");
if( A.Grid() != B.Grid() || B.Grid() != C.Grid() )
throw std::logic_error
("{A,B,C} must be distributed over the same grid");
if( orientationOfA == NORMAL )
throw std::logic_error("GemmTNC assumes A is (Conjugate)Transposed");
if( A.Width() != C.Height() ||
B.Width() != C.Width() ||
A.Height() != B.Height() )
{
std::ostringstream msg;
msg << "Nonconformal GemmTNC: \n"
<< " A ~ " << A.Height() << " x " << A.Width() << "\n"
<< " B ~ " << B.Height() << " x " << B.Width() << "\n"
<< " C ~ " << C.Height() << " x " << C.Width() << "\n";
throw std::logic_error( msg.str().c_str() );
}
#endif
const Grid& g = A.Grid();
// Matrix views
DistMatrix<T,MC,MR> AT(g), A0(g),
AB(g), A1(g),
A2(g);
DistMatrix<T,MC,MR> BT(g), B0(g),
BB(g), B1(g),
B2(g);
// Temporary distributions
DistMatrix<T,STAR,MC> A1_STAR_MC(g);
DistMatrix<T,STAR,MR> B1_STAR_MR(g);
// Start the algorithm
Scal( beta, C );
LockedPartitionDown
( A, AT,
AB, 0 );
LockedPartitionDown
( B, BT,
BB, 0 );
while( AB.Height() > 0 )
{
LockedRepartitionDown
( AT, A0,
/**/ /**/
A1,
AB, A2 );
LockedRepartitionDown
( BT, B0,
/**/ /**/
B1,
BB, B2 );
A1_STAR_MC.AlignWith( C );
B1_STAR_MR.AlignWith( C );
//--------------------------------------------------------------------//
A1_STAR_MC = A1; // A1[*,MC] <- A1[MC,MR]
B1_STAR_MR = B1; // B1[*,MR] <- B1[MC,MR]
// C[MC,MR] += alpha (A1[*,MC])^T B1[*,MR]
// = alpha (A1^T)[MC,*] B1[*,MR]
internal::LocalGemm
( orientationOfA, NORMAL, alpha, A1_STAR_MC, B1_STAR_MR, (T)1, C );
//--------------------------------------------------------------------//
A1_STAR_MC.FreeAlignments();
B1_STAR_MR.FreeAlignments();
SlideLockedPartitionDown
( AT, A0,
A1,
/**/ /**/
AB, A2 );
SlideLockedPartitionDown
( BT, B0,
B1,
/**/ /**/
BB, B2 );
}
#ifndef RELEASE
PopCallStack();
#endif
}
inline void
RowEchelon( DistMatrix<F>& A, DistMatrix<F>& B )
{
#ifndef RELEASE
CallStackEntry entry("RowEchelon");
if( A.Grid() != B.Grid() )
LogicError("{A,B} must be distributed over the same grid");
if( A.Height() != B.Height() )
LogicError("A and B must be the same height");
#endif
const Grid& g = A.Grid();
// Matrix views
DistMatrix<F>
ATL(g), ATR(g), A00(g), A01(g), A02(g), APan(g),
ABL(g), ABR(g), A10(g), A11(g), A12(g),
A20(g), A21(g), A22(g);
DistMatrix<F>
BT(g), B0(g),
BB(g), B1(g),
B2(g);
// Temporary distributions
DistMatrix<F,STAR,STAR> A11_STAR_STAR(g);
DistMatrix<F,STAR,VR > A12_STAR_VR(g);
DistMatrix<F,STAR,MR > A12_STAR_MR(g);
DistMatrix<F,MC, STAR> A21_MC_STAR(g);
DistMatrix<F,STAR,VR > B1_STAR_VR(g);
DistMatrix<F,STAR,MR > B1_STAR_MR(g);
DistMatrix<Int,STAR,STAR> p1_STAR_STAR(g);
// In case B's columns are not aligned with A's
const bool BAligned = ( B.ColShift() == A.ColShift() );
DistMatrix<F,MC,STAR> A21_MC_STAR_B(g);
// Pivot composition
std::vector<Int> image, preimage;
// Start the algorithm
PartitionDownDiagonal
( A, ATL, ATR,
ABL, ABR, 0 );
PartitionDown
( B, BT,
BB, 0 );
while( ATL.Height() < A.Height() && ATL.Width() < A.Width() )
{
RepartitionDownDiagonal
( ATL, /**/ ATR, A00, /**/ A01, A02,
/*************/ /******************/
/**/ A10, /**/ A11, A12,
ABL, /**/ ABR, A20, /**/ A21, A22 );
RepartitionDown
( BT, B0,
/**/ /**/
B1,
BB, B2 );
View2x1
( APan, A12,
A22 );
A12_STAR_VR.AlignWith( A22 );
A12_STAR_MR.AlignWith( A22 );
A21_MC_STAR.AlignWith( A22 );
B1_STAR_VR.AlignWith( B1 );
B1_STAR_MR.AlignWith( B1 );
if( ! BAligned )
A21_MC_STAR_B.AlignWith( B2 );
p1_STAR_STAR.ResizeTo( A11.Height(), 1 );
//--------------------------------------------------------------------//
A11_STAR_STAR = A11;
A21_MC_STAR = A21;
lu::Panel( A11_STAR_STAR, A21_MC_STAR, p1_STAR_STAR, A00.Height() );
ComposePivots( p1_STAR_STAR, A00.Height(), image, preimage );
ApplyRowPivots( APan, image, preimage );
ApplyRowPivots( BB, image, preimage );
A12_STAR_VR = A12;
B1_STAR_VR = B1;
LocalTrsm
( LEFT, LOWER, NORMAL, UNIT, F(1), A11_STAR_STAR, A12_STAR_VR );
LocalTrsm( LEFT, LOWER, NORMAL, UNIT, F(1), A11_STAR_STAR, B1_STAR_VR );
A12_STAR_MR = A12_STAR_VR;
B1_STAR_MR = B1_STAR_VR;
LocalGemm( NORMAL, NORMAL, F(-1), A21_MC_STAR, A12_STAR_MR, F(1), A22 );
if( BAligned )
{
LocalGemm
( NORMAL, NORMAL, F(-1), A21_MC_STAR, B1_STAR_MR, F(1), B2 );
}
else
{
A21_MC_STAR_B = A21_MC_STAR;
LocalGemm
( NORMAL, NORMAL, F(-1), A21_MC_STAR_B, B1_STAR_MR, F(1), B2 );
}
A11 = A11_STAR_STAR;
A12 = A12_STAR_MR;
B1 = B1_STAR_MR;
//--------------------------------------------------------------------//
SlidePartitionDownDiagonal
( ATL, /**/ ATR, A00, A01, /**/ A02,
/**/ A10, A11, /**/ A12,
/*************/ /******************/
ABL, /**/ ABR, A20, A21, /**/ A22 );
SlidePartitionDown
( BT, B0,
B1,
/**/ /**/
BB, B2 );
}
}
