Dense LU Factorization with Double Precision

Accelerated Computing GPU-Accelerated Libraries
1

Dear NVIDIA cuSolver developers,

Is there any plan for cuSolver to support double precision Dense LU Factorization using tensor core? Currently there is a function cusolverDnDgetrs to do this job, but it is unclear if this API uses ordinary GPU cores or tensor core.

Thanks!

2

For GETRS FP64, cuSOLVER calls cuBLAS, which runs on DMMA Tensor cores.
You should be able to see TC usage with the latest version of Nsight Systems