VexCL seems to be a very attractive library for gpu programming.
Unfortunately, its a very young library and there are few information over there. I've been searching how to execute a matrix-vector multiplication, but the only matrix representation I've found is vex::SpMat, which holds a sparse matrix.
If matrix is dense, then the sparse representations are less efficient for computations, usually.
All my matrix are dense and I want to know how to execute this efficiently in VexCL.