Taking Advantage of Collective Operation Semantics for Loosely Coupled Simulations ∗

Although a loosely coupled component-based framework offers flexibility and versatility for building and deploying large-scale multi-physics simulation systems, the perfor-mance of such a system can suffer from excessive buffer-ing of data objects which may or may not be transferred between components. By taking advantages of the collec-tive properties of parallel simulation components, which is common for data-parallel scientific applications, an opti-mization method, which we call buddy-help, can greatly enhance overall performance. Buddy-Help can reduce the time taken for buffering operations in an exporting compo-nent, when there are timing differences across processes in the exporting component. The optimization enables skip-ping unnecessary buffering operations, once another pro-cess, which has already performed the collective export op-eration, has determined that the buffered data will never be needed. Because an analytical study would be very difficult due to the complexity of the overall coupled simulation sys-tem, the performance improvement enabled by buddy-help is investigated via a micro-benchmark specifically designed to illustrate the behavior of coupled simulation scenarios under which buddy-help can provide performance gains.