search

NVIDIA / CUDALibrarySamples

CUDA Library Samples
Other
1.56k stars 333 forks source link

[cuSPARSELT] Performance differences between cusparseLtSpMMACompress and cusparseLtSpMMACompress2? #165

Closed cj831880 closed 10 months ago

cj831880 commented 10 months ago

Are there are any performance differences between cusparseLtSpMMACompress and cusparseLtSpMMACompress2 or any other reason why using one over the other?

fbusato commented 10 months ago

no, both functions call the same internal kernel. The only difference is that cusparseLtSpMMACompress2 is decoupled from the Plan