MPI and HPF Performance in a DEC Alpha Cluster
Abstract
There are several different types of parallel computer architectures in use today. Some of these are large machines that house hundreds of processors with low- to mid-range computing capabilities. A different type of parallel computer architecture becoming increasingly popular is that of the cluster. Clusters are basically networked workstations: each containing 1 to approx. 30 processors with mid- to high-range computing capabilities. While arguments can be made for both paradigms, clusters seem to be gaining in popularity. They provide fast computation through multiplicity and fast processor throughput. Furthermore, code reuse on different cluster environments is now possible with the adoption of standard interprocessor communication tools like the Message-Passing Interface (MPI) and High Performance FORTRAN (HPF). This report compares and contrasts the performance of MPI and HPF on a currently available computer cluster.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1998
- Accession Number
- ADA343800
Entities
People
- Dale Shires
Organizations
- United States Army Research Laboratory