Say I have two matrices mat1 of size 3xN and mat2 of size m x n. And N = m*n , i.e number of cols in mat1 is same as the size of mat2.

I want to multiply column 0 of a matrix mat1 with mat2(0,0), column 1 of mat1 with mat2(0,1), and so on.

If mat2 was a vector of length N, thaen I could have done

Code: Select all

mat1 = mat1 * v.asDiagonal()

How do I do this operation efficiently without creating temporaries , since my N is quite large.

Update:
One solution I found is to map mat2 to a vector by doing:

Code: Select all

Map<VectorXd> v(mat2.data(), mat2.size());
mat1 = mat1 * v.asDiagonal();

But still I would like to know if there is a better way.

P.S I have by default row_major enabled, so the above code does what i expect.

Nothing better than the map trick. However, be careful of the storage order of mat2.

Thanks. I like the .asDiagonal() functionality.

However I feel, it will be nice to have a similar .asVector() functionality for Matrix types. I mean something similar to Matlab's A ( : ).