Benchmarking and Programming Practices for the Intel Xeon Phi Coprocessor

2016The Intel Xeon Phi is a relative newcomer to the scientific computing scene. In the recent years, GPUs have been used extensively for mathematical simulations. The Xeon Phi is Intel’s response to the use of these cards. Like the GPU, it is highly parallelizable but can be programmed like a CPU. Since the card is relatively new, little literature is written about it. The focus of this paper is to benchmark the card and find programming practices that are conducive to higher performance and to compare those techniques and their performance to those of more mainstream processors