CusparseSpMV and cusparseCsrmvEx perform the same sparse matrix with dense matrix multiplication as per the toolkit documentation. Any ideas of the benefits of using one function over the other would be appreciated?
the only difference i could find was cusparseCsrmvEx was fixed to CSR. there isnt much info on what exactly is going on.
also what exactly does this mean “pointers should be aligned with 128 bytes” (as per cusparseCsrmvEx limitation)