Online Performance Projection for Clusters with Heterogeneous GPUs
Dept. of Computer Science, Virginia Tech.
IEEE International Conference on Parallel and Distributed Systems (ICPADS), 2013
@article{panwar2013online,
title={Online Performance Projection for Clusters with Heterogeneous GPUs},
author={Panwar, Lokendra S and Aji, Ashwin M and Meng, Jiayuan and Balaji, Pavan and Feng, Wu-chun},
year={2013}
}
We present a fully automated approach to project the relative performance of an OpenCL program over different GPUs. Performance projections can be made within a small amount of time, and the projection overhead stays relatively constant with the input data size. As a result, the technique can help runtime tools make dynamic decisions about which GPU would run faster for a given kernel. Usage cases of this technique include scheduling or migrating GPU workloads over a heterogeneous cluster with different types of GPUs.
October 25, 2013 by hgpu
