Out of core computation

Accelerated Computing GPU-Accelerated Libraries
,
1

Hi,

Do cuBLAS and cuSPARSE support multiplying large matrices that do not fit in GPU memory (out of core computation)?

2

cublas supports it using cublasXt:

There are no restriction on the sizes of the matrices as long as they can fit into the host memory.

I’m not aware of any support for it in cusparse, currently

3

Thank you very much.

4

Hi. We don’t have such feature in cuSparse. What functionality you want to do?

5

Hi @qanhpham,

Thank you. I simply want to multiply a large sparse matrix (20Mx400K) by a dense matrix (400Kx500). It is used as a part of image processing project.

Kind regards,