Blocked CSR SPMV Implementation

Has anybody implemented Blocked CSR/ELL SPMV in CUDA?? Where can I get the source code?. I'm fin开发者_运维技巧ding it hard to block the matrix in CSR format

I don't think anyone will give you the implementation. You can check this paper on Efficient SPMV on GPUs using CUDA. They even talk about CSR kernel.