Computer Engineering Commons^™

Towards Comparative Profiling Of Parallel Applications With Pperfdb, Christian Leland Hansen

Dissertations and Theses

Due to the complex nature of parallel programming, it is difficult to diagnose and solve performance related problems. Knowledge of program behavior is obtained experimentally, with repeated runs of a slightly modified version of the application or the same code in different environments. In these circumstances, comparative performance analysis can provide meaningful insights into the subtle effects of system and code changes on parallel program behavior by highlighting the difference in performance results across executions.

I have designed and implemented modules which extend the PPerfDB performance tool to allow access to existing performance data generated by several commonly used tracing …