Performance Characteristics of the MultiZone NAS Parallel Benchmarks.pdf

Performance Characteristics of the Multi-Zone NAS Parallel Benchmarks Haoqiang Jin and Rob F. Van der Wijngaart* NASA Advanced Supercomputing Division M/S T27A-1, NASA Ames Research Center, Mogett Field, CA 94035-1000 {hjin,wijngaar}@nas .nasa.gov Abstract We describe a new suite of computational benchmarks that models applications featuring multiple levels of parallelism. Such parallelism is often available in realistic flow computations on systems of grids, but had not previously been captured in bench- marks. The new suite, named NPB Multi-Zone, is extended from the NAS Parallel Benchmarks suite, and involves solving the application benchmarks LU, BT and SP on collections of loosely coupled discretization meshes. The solutions on the meshes are updated independently, but after each time step they exchange boundary value information. This strategy provides relatively easily exploitable coarse-grain paral- lelism between meshes. Three reference implementations are available: one serial, one hybrid using the Message Passing Interface (MPI) and OpenMP, and another hybrid using a shared memory multi-level programming model (SMP+OpenMP). We exam- ine the effectiveness of hybrid parallelization paradigms in these implementations on three different parallel computers. We also use an empirical formula to investigate the performance characteristics of the multi-zone benchmarks. 1 Introduction The NAS Parallel Benchmarks (NPB) [l]are well-known problems for testing the capabilities of parallel computers and parallelization tools. They exhibit mostly fine-grain exploitable parallelism