HPC Benchmark Suite NMx, Phase II

In the phase II effort, Intelligent Automation Inc., (IAI) and University of Central Florida (UCF) propose to develop a comprehensive numerical test suite for benchmarking current and future high performance computing activities that will include: (1) dense and unsymmetrical matrix problems faced in space aviation and problems in thermally driven structural response and radiation exchange, (2) implicit solution algorithms with production models and benchmarks for indefinite matrices and pathological cases, (3) configurations scaling for large systems in shared, distributed and mixed memory conditions, (4) documentation for strengths, weaknesses, and limitations of the toolkits used together with recommendations and (5) precision and round-off studies on serial and parallel machines, comparison of solutions on serial and parallel hardware with study of wall clock performance with respect to the number of processors We successfully demonstrated in phase I that we can accurately and precisely benchmark run time solvers of dense complex matrices in hybrid-distributed memory architecture. We achieved highly scalable super-linear speed-up and scalability of the algorithm for large problem sizes. The tools developed in phase II will greatly improve the performance and efficiency to adapt the benchmarks to HPC systems different hardware architectures at NASA facilities and for non-NASA commercial applications.