Is it poosible to use HYB or ELL sparse matrix multiplication(SPMV) in cuda 11?

Accelerated Computing GPU-Accelerated Libraries
1

I found the HYB or ELL format sparse matrix in cuda 11 is deprecated. But for my practical problem, HYB format SPMV run better than CSR format. Is there any method to use HYB format in cuda 11? Like including some other library? Or I must write these kernel code myself?

2

No, as you said, these two formats have been deprecated. What do you mean that HYB runs better? is it about performance? Did you try cusparseSpMV in CUDA 11.2.1? it should provide better performance than HYB

1 Like