The last problem we have to tackle is the question how to map a vector on the CPU into a texture on the GPU. We choose the easiest mapping that comes to mind: A vector of length N is mapped into a texture of size sqrt(N) by sqrt(N) for LUMINANCE formats (this means we assume N is a power of two), and into a texture of size sqrt(N/4) by sqrt(N/4) for RGBA formats, again assuming N is chosen so that