fun hot tip, inside your $HFS/ocl/compute folder, the team at sesi has included both a simple sparse matrix vector multiplication function as well as a bunch of opencl vector functions, including parallel reduction for large dot product operations 😮
definitely worth taking a look at