HIGH PERFORMANCE COMPUTING FOR RECONNAISSANCE APPLICATIONS

Parallel programming is vital to fully utilize the multicore architectures that dominate the processor market. The market, however, is constantly evolving, with new processors and new architectures getting released annually. Using an open parallel processing language, such as OpenCL (Open Computing Language), enables the use of a single program across multiple architectures. It also enables a method of evaluation between multiple devices so the best choice can be made for a given application. In this research, OpenCL is used to evaluate the performance of two signal processing algorithms across two graphics processing units and one central processing unit. Experimental results show that for each algorithm, a specific device can clearly be shown to outperform the others.Ensign, United States NavyApproved for public release; distribution is unlimited