Finding the Best HPCMP Architectures Using Benchmark Application Results for TI-09

Time-to-solution and scalability of an application can vary greatly from one computer architecture to another. It is important then to consider the appropriateness of architectures with respect to applications to make the most efficient use of available resources. However, the number of applications currently used on systems within the High Performance Computing Modernization Program (HPCMP) is very large. This paper, therefore, focuses solely on codes within the HPCMP Technology Insertion 2009 (TI-09) Applications Benchmarking Suite. These codes include Adaptive Mesh Refinement (AMR), Air Vehicles Unstructured Solver (AVUS), CTH, General Atomic and Molecular Electronic Structure System (GAMESS), HYbrid Coordinate Ocean Model (HYCOM), Improved Concurrent Electromagnetic Particle In Cell (ICEPIC), and Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS). This paper evaluates the parallel performance-per-processor of each of these seven codes on major systems within the Program. This, in turn, can guide users and developers in selecting appropriate architectures for each code.