CUDA Multi-Process Service (MPS)

CUDA MPS [1] is an alternative to the CUDA application programming interface (CUDA API). The MPS runtime architecture is designed to transparently enable the use of Hyper-Q to CUDA applications that require the use of multiple cooperative processes, usually MPI jobs, on NVIDIA graphics processing units (those based on the Kepler architecture). Hyper-Q allows CUDA kernels to be processed concurrently on the same GPU; thus benefiting performance when it is sub -used by a single process

Versions

• CUDA-8.0

[1]

nvidia, “MULTI-PROCESS SERVICE”, Mayo del 2015. https://docs.nvidia.com/deploy/pdf/CUDA_Multi_Process_Service_Overview.pdf. Accedido por última vez el 11 de julio del 2017