Dear Cedric,
I found out that batch QR, LU, Chol decomposition need both a batched TRSM and GEMM. But current TRSM (inspired by magma) only support a single matrix, is it easy to implement a batched version with an extra page dimension?https://docs.nvidia.com/cuda/cusolverdx/get_started/trsm.html
Thanks.
Best wishes,
Jinchuan
Dear Cedric,
I found out that batch QR, LU, Chol decomposition need both a batched TRSM and GEMM. But current TRSM (inspired by magma) only support a single matrix, is it easy to implement a batched version with an extra page dimension?https://docs.nvidia.com/cuda/cusolverdx/get_started/trsm.html
Thanks.
Best wishes,
Jinchuan