BSR implimnatation

Accelerated Computing GPU-Accelerated Libraries
1

Hi All,

Is there an open source implementation of the Blocked CSR (BSR) SpMV kernel? I didn’t find any in CUSP library.

I would really appreciate any help, thanks.

2

cuSPARSE has support for BSR matrices. You can use the

cusparse<t>bsrmv()

function for BSR matrix-vector products.

Here is the manual, it has some examples as well. Just remember to include the header and link the library via -lcusparse:
http://docs.nvidia.com/cuda/cusparse/#cusparse-level-2-function-reference