Efficient Offloading of Parallel Kernels Using MPI_Comm_Spawn | IEEE Conference Publication | IEEE Xplore