DenseMatrix<double> ParSparseMatrix<double>::transMat(
const DenseMatrix<double>& B) const {
SparseMatrix<double>::compress(*this);
GCK(*this, B, this->nRows() != B.nRows(), "ParSparseMatrix transpose * Matrix")
DenseMatrix<double> C(nCols(), B.nCols(), true);
SparseMatrix<double> A_local;
partition(*static_cast<ParSparseMatrix<double>*>(&A_local));
// actually do multiplication - end up with partial result matrix
// on each processor
DenseMatrix<double> C_local = A_local.transMat(B);
static_cast<ParSparseMatrix<double>*>(&A_local)->finalize();
// Add all the result vectors together on each processor.
allsum(_comm, C_local.ptr(), C.ptr(), C_local.size());
return C;
}
