p2pBandwidthLatencyTest cudaDeviceCanAccessPeer cudaDeviceEnablePeerAccess cudaDeviceDisablePeerAccess cudaEventCreateWithFlags cudaEventElapsedTime cudaMemcpy whole ./ ../ ../../common/inc Performance Strategies Asynchronous Data Transfers Unified Virtual Address Space Peer to Peer Data Transfers Multi-GPU CUDA Performance multi-GPU support peer to peer true p2pBandwidthLatencyTest.cu 1:CUDA Basic Topics 1:Performance Strategies sm35 sm37 sm50 sm52 sm60 sm61 sm70 sm72 sm75 sm80 x86_64 linux windows7 x86_64 macosx arm ppc64le linux all Peer-to-Peer Bandwidth Latency Test with Multi-GPUs exe