https://hgpu.org/?p=11089
All-pairs Shortest Path Algorithm based on MPI+CUDA Distributed Parallel Programming Model