"Performance Monitoring and Analysis Components in Adaptive PDE-Based Simulations"
B. Norris and I. Veljkovic
Preprint Version: [pdf]
Newton-Krylov nonlinear and linear solver components for performance monitoring using the TAU performance tools. To reduce the performance monitoring and component adaptation overhead, we employ two databases. The first is created and destroyed during runtime and stores performance data for code segments of interest, as well as various application-specific performance events in the currently running application instance. The second database is persistent and contains performance data from various applications and different instances of the same application. It can also contain performance information derived through offline analysis of raw data. We describe a prototype implementation of this infrastructure and show how adaptive linear solver algorithms are employed in a driven cavity flow simulation code.