MultiGPU

Can we use atomic exchange sorting algorothm in MultiGPU? Why?