Enabling the CUDA Unified Memory model in Edge, Cloud and HPC offloaded GPU kernels | IEEE Conference Publication | IEEE Xplore