An investigation of Unified Memory Access performance in CUDA | IEEE Conference Publication | IEEE Xplore