Creation of a synthetic application of benchmarking

Synthetic applications performing benchmarks have been suggested as a means to streamline the benchmarking process. The idea is to effectively and efficiently evaluate computing performance without having to obtain licenses for proprietary codes, and without risking the release of sensitive or classified codes, code fragments, or algorithms. We present one approach to solve this problem. Based on that approach, we then present results measured on habu, an IBM Power3 supercomputer at the Major Shared Resource Center at the Naval Oceanographic Office.