Question about multi-GPU support mentioned in the NPP Plus introduction

Accelerated Computing GPU-Accelerated Libraries
1

On NVIDIA Performance Primitives (NPP) | NVIDIA Developer. The “Download NPP +” section mentions that the “NPP + library, … , plus the added advantage of multi-GPU support.” Does the term “multi-GPU support” in this context refer to a specific operator in NPP+ that enables multiple GPU devices to work together to compute the current operator by setting the nCudaDeviceId parameter (or other parameters) of NppStreamContext? If so, can you provide a specific example? If not, what is the meaning of “multi-GPU support”? Thanks.