I have folded my long vector into 2D using column-major ordering in Brook. I want to apply sparse matrix multiply using x[index]*A in a kernel on my 2D streams.
1. How can I get and operate on each column of my 2D stream? (If I do this, I can still use my 1D sparse matrix multiplication kernel)
2. Is there an easier way to implement this matrix operation using a construction similar to (x[index]*A) but perhaps utilizing 2D arrays? This would avoid me calling my 1D kernel.