|
Math @ Duke
|
Publications [#322671] of Amanda Randles
Papers Published
- Kale, V; Randles, A; Gropp, WD, Locality-optimized mixed static/dynamic scheduling for improving load balancing on SMPs,
ACM International Conference Proceeding Series, vol. 09-12-September-2014
(September, 2014),
pp. 115-116, ACM Press, ISBN 9781450328753 [doi]
(last updated on 2025/12/31)
Abstract: Application performance can be degraded significantly due to node-local load imbalances during application execution. Prior work suggested using a mixed static/dynamic scheduling approach for handling this problem, specifically in the context of fine-grained, transient load imbalances. Here, we consider an alternate strategy for more general load imbalances where fine-grained, transient load imbalance may be coupled with coarse-grained load imbalance. Specifically, we implement a scheduling scheme in which we modify the data layout in mixed static/dynamic scheduling, and add an additional tuned constraint in the dequeue function of our scheduler. Through experimentation of an n-body particle simulation code on modern multi-core architectures, our technique gives a 19.4% performance gain over dynamic scheduling, and an overall 48.6% performance gain over standard static scheduling.
|
|
|
|
dept@math.duke.edu
ph: 919.660.2800
fax: 919.660.2821
| |
Mathematics Department
Duke University, Box 90320
Durham, NC 27708-0320
|
|