Benchmarking and Performance Characterization
of Parallel Computers
using Algorithmic Structures

A practical methodology has been developed for evaluating and comparing the performance of parallel computers. The methodology determines machine parameters and program parameters separately, and predicts the performance of a given workload on the machines under consideration. Machine parameters are measured using benchmarks that consist of parallel algorithm structures. It has been shown that performance of parallel computers is sensitive to the changes in the workload and therfore any evaluation and comparison of architectures must consider a given user workload. For various example workloads, predicted times have been compared with the actual runs, and the results show that the prediction error is within 10-20%.

The methodology has been implemented on a number of MIMD and SIMD machines and cluster of PCs. Benchmarks can be found here. More information on the implementation of the methodology and the benchmark results can be found in the publications.


Links to other benchmarks