Slow sparse matrix-vector product with symmetric matrices

This has nothing to do with global scope or by dense being faster etc etc. The explanation is that:

julia> a = sprand(5,5,0.5); b = rand(5); c = similar(b);

julia> @edit A_mul_B!(c, Symmetric(a), b)

falls back to the generic matrix multiplication which will of course be extremely slow for sparse matrices.

3 Likes