You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
CUTLASS uses CUDA streams to launch GEMM kernels and can be used in a multi-GPU application.
The kernels CUTLASS launches are, however, single-GPU operations that an application would need to compose to compute larger computations distributed across multiple GPUs and multiple nodes.
Dear sir:
Does cutlass support multi-GPU automatically.
The text was updated successfully, but these errors were encountered: