Design Space Exploration of Biometrics on Intel Many integrated Core Architecture (NICS)

The Beacon system provides high-performance computing resource – large amount of conventional cores plus innovative accelerator cores along with high memory bandwidth. The system also provides mass storage system and high I/O bandwidth to manage the large datasets efficiently. All of these features provide us an ideal platform to analyze the biometric information in a timely and efficient manner. The biometrics applications contain tremendous amount of data parallelism. The data elements handled in biometrics are sometimes as simple as a vector of values and the main operations are sometimes as simple as vector operations. Thus, biometric information can be distributed to the cores and then processed in a parallel fashion. As for our research, our datasets include the biometric data such as face, finger, voice, iris, etc. On the Beacon platform, we will port and optimize our application onto the Intel Many Integrated Core (MIC) architecture and develop experimental methods and tools to explore the impact of MIC architecture on the biometric applications in the perspectives of performance evaluation and energy-efficiency.