Application representations for multiparadigm performance modeling of large-scale parallel scientific codes

Effective performance prediction for large parallel applications on very la rge-scale systems requires a comprehensive modeling approach that combines analytical models, simulation models, and measurement for different applica tion and system components. This paper presents a common parallel program r epresentation, designed to support such a comprehensive approach, with four design goals: (1) the representation must support a wide range of modeling techniques; (2) it must be automatically computable using parallelizing co mpiler technology, in order to minimize the need for user intervention; (3) it must be efficient and scalable enough to model teraflop-scale applicati ons; and (4) it should be flexible enough to capture the performance impact of changes to the application, including changes to the parallelization st rategy, communication, and scheduling. The representation we present is bas ed on a combination of static and dynamic task graphs. It exploits recent c ompiler advances that make it possible to use concise, symbolic static grap hs and to instantiate dynamic graphs. This representation has led to the de velopment of a compiler-supported simulation approach that can simulate reg ular, message-passing programs on systems or problems 10 to 100 times large r than was possible with previous state-of-the-art simulation techniques.