DMCA
DRAFT: UPC vs. MPI and OpenMP: Analysis of a Hybrid Approach to Parallel Programming (2002)
BibTeX
@MISC{Berlin02draft:upc,
author = {Konstantin Berlin},
title = {DRAFT: UPC vs. MPI and OpenMP: Analysis of a Hybrid Approach to Parallel Programming},
year = {2002}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract. The main idea behind UPC is that users should be able to view the underlying machine model as a collection of threads operating in a common global address [5]. In such an environment fine-grain algorithms are preferred to coarse-grain algorithms since they are easier to code. While such algorithms, written in UPC, should have similar performance in shared memory environments compared to OpenMP, they are currently slower. In addition they have unacceptably long execution time in a cluster environment compared to coarse-grain algorithms written in MPI and UPC itself. Also UPC lacks the flexibility and functionality of OpenMP and MPI in their respective environments. 1.
