simpleCudaGraphs cudaGraphAddMemsetNode cudaGraphsUsingStreamCapture cudaMemcpyAsync cudaGraphGetNodes cudaStreamDestroy cudaMallocHost cudaGraphClone cudaEventCreate cudaPitchedPtr cudaGraphCreate cudaMalloc cudaPos cudaGraphAddMemcpyNode cudaStreamEndCapture cudaGraphExecDestroy cudaStreamBeginCapture cudaGraphAddHostNode cudaGraphsManual cudaStreamSynchronize cudaGraphDestroy cudaGraphLaunch cudaStreamWaitEvent cudaFree cudaEventRecord cudaStreamCreate cudaGraphInstantiate cudaLaunchHostFunc cudaExtent cudaMemsetAsync cudaFreeHost cudaGraphAddKernelNode whole ./ ../ ../../../Common CUDA Graphs Stream Capture CUDA GPGPU CUDA Graphs true simpleCudaGraphs.cu 1:CUDA sm35 sm37 sm50 sm52 sm53 sm60 sm61 sm70 sm72 sm75 sm80 sm86 sm87 x86_64 linux windows7 x86_64 macosx arm sbsa ppc64le linux 3.5 exe