The paper presents present a mechanism for automatic management of the memory hierarchy, including secondary storage, in the context of a global address space parallel programming framework. The programmer specifies the parallelism and locality in the computation. The scheduling of the computation into stages, together with the movement of the associated data between secondary storage and global memory, and between global memory and local memory, is automatically managed by the framework. A novel formulation of hypergraph partitioning is used to model the optimization problem of minimizing disk I/O by improving locality of access. Experimental evaluation of the proposed approach using a sub-computation from the quantum chemistry domain shows a reduction in the disk I/O cost by upto a factor of 11, and a reduction in turnaround time by upto 97%, as compared to alternatives used in state-of-the-art quantum chemistry codes.
Revised: July 31, 2007 |
Published: November 15, 2007
Citation
Krishnamoorthy S., U. Catalyurek, J. Nieplocha, A. Rountev, and P. Sadayappan. 2007.Hypergraph Partitioning for Automatic Memory Hierarchy Management. In Conference on High Performance Networking and Computing. Proceedings of the 2006 ACM/IEEE Conference on Supercomputing SC '06, Tampa, FL, 11-17 Nov. 2006, 12 pages. New York, New York:ACM Press.PNNL-SA-50740.doi:10.1109/SC.2006.36