Scientific computing workloads are well suited to parallel accelerators such as GPGPUs and the Intel Xeon Phi. While these accelerators can provide greater performance than traditional CPUs due to their parallel architectures ...