Limits on the Performance Benefits of Multithreading and Prefetching

        The performance benefits of multithreading and software prefetching depend heavily on hardware and application-specific parameters, such as the multiprocessor's remote memory access latency and the cache miss rate of parallel applications. This paper considers the relationship between these parameters and the maximum performance improvement achievable through multithreading and software prefetching. Specifically, the paper presents new analytical models and experimental measurements of several parallel applications on the MIT Alewife multiprocessor and uses them to predict the benefits of these latency-tolerating techniques. For the first time, both techniques are evaluated on a real machine as opposed to simulations. The models determine the regions in the parameter space where the techniques may be most effective, while the measurements determine the regions where applications actually lie. We find that these regions do not always overlap significantly. The multithreading model explains why 2-4 contexts is sufficient to achieve most of this technique's potential benefit in the current generation of shared-memory multiprocessors. For these multiprocessors, the maximum improvement in running time from multithreading is less than 10% for most of the applications in this paper. The model also shows that multithreading can significantly improve the performance of the same applications in future multiprocessors with much longer latencies, provided sufficient hardware contexts exist to tolerate the longer latencies. Reducing context-switch overhead will not be crucial in these multiprocessors (ScalParSys).

By: Beng-Hong Lim, Ricardo Bianchini (Federal Univ. of Rio de Janeiro, Brazil)

Published in: RC20238 in 1995


This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.

Questions about this service can be mailed to .