conjugateGradientMultiDeviceCG -ewp -maxrregcount=64 --std=c++11 cudaMemAdvise cudaMemPrefetchAsync cudaLaunchCooperativeKernelMultiDevice cudaStreamSynchronize cudaOccupancyMaxActiveBlocksPerMultiprocessor whole ./ ../ ../../Common Unified Memory Linear Algebra Cooperative Groups MultiDevice Cooperative Groups CUDA Sparse Matrix Unified Memory Multi-GPU CPP11 cudadevrt true conjugateGradientMultiDeviceCG.cu UVM MDCG CPP11 1:CUDA Advanced Topics 3:Linear Algebra sm60 sm61 sm70 sm72 sm75 sm80 sm86 x86_64 linux ppc64le linux windows aarch64 6.0 exe