Hi,

This is not a problem/bug, just wondering if anyone can help with optimizing the following expression if possible.

Code: Select all

 
return A*(a.asDiagonal())*(A.adjoint());


A is MatrixXcd, a is VectorXcd and the result is MatrixXcd.

Thanks!

You can exploit that the result is self-adjoint this way:

Code: Select all

MatrixXcd B = a.asDiagonal()*A;
MatrixXcd res(a.size(),a.size());
res.triangularView<Lower>() = A*B.adjoint();
res.triangularView<Upper>() = res.adjoint(); // if you need the full matrix only


If your matrices are large enough (>200x200), then you should get an almost x2 speedup.