Fast and Accurate Evaluation of Memory Peformance Upper-Bound
Grigori Fursin, Mike O'Boyle, University of Edinburgh, and Olivier Temam, Gregory Watt, Paris South University
- Full Paper(.ps version)
In many cases, compiler-based optimizations for memory performance tuning are too limited and programmers have to resort to manual optimization techniques. However, this process is tedious and time-consuming, especially since there is no indication on when the process can stop, i.e., when optimal memory performance has been achieved or sufficiently approached. Architecture simulators can provide such information but designing an accurate model of an existing architecture is very difficult, and simulation times are excessively long. In this article, we propose a technique that is both fast and reasonably accurate for estimating the memory performance upper-bound of many scientific applications. This technique has been tested on several programs and can be used to guide a manual optimization process, or even to drive iterative compilation techniques.
Please contact our
webadmin with any comments or changes.
Unless explicitly stated otherwise, all material is
copyright © The University of Edinburgh.