Performance Monitoring and Analysis Components in Adaptive PDE-Based Simulations 1

摘要: As scientists incorporate more sophisticated models into their simulations, software complexity,aswellasthe underlyingcomputational costof thesemodels,are growing rapidly. Performance evaluation and tuning of applications that are largescale both in terms source code runtime requirements can be challenging andtime-consumingforscientists.Wehavedevelopedasoftwareinfrastructurefor performance monitoring, data management, adaptive algorithm developmentforcomponentPDE-basedsimulations.Newton-Krylovnonlinearand linear solver components instrumented for monitoring using the TAU tools. To reduce component adaptation overhead, we employ two databases serve signiflcantly difierent purposes. The flrst one is created destroyed during runtime, stores segments interest, as well various application-speciflc events currently running application instance. second ispersistentandcontainsperformancedatafromvariousapplicationsanddifierent instances same application. It also contain information derivedthrougho†ineanalysisofrawdata.Wedescribeaprototypeimplementation ofthisinfrastructureandshowhowadaptivelinearsolveralgorithmsareemployed a driven cavity ∞ow simulation code.