Performance Modelling of Supercomputer Architectures and Algorithms
The need for more processing power has spurred the development of new processor architectures. While these machines bear a high speed potential, this speedup over a conventional machine is only partly achieved, depending on the type of calculations. In this paper an overview is given of the blend of techniques which allow to estimate the processing rate for various existing computer architectures in function of their serial versus vector calculations.
KeywordsExecution Time Service Time Clock Period Array Processor Memory Bank
Unable to display preview. Download preview PDF.
- 1.M.C. Vanwormhoudt, Efficient data handling using multiprocessor structures, in: Flanders Technology International Seminar: “Parallel Information Processing and Artificial Intelligence in Simulation”, 21–39, 28 Feb. 1985.Google Scholar
- 3.R.W. Hockney, and C.R. Jesshope, ”Parallel Computers“, 416 pages, Adam Hilger Ltd. (1981).Google Scholar
- 5.R.B. Lee, ”Performance Characterization of Parallel Computer Organizations“, Ph.D. Thesis, Stanford University (1980).Google Scholar
- 6.E.H. D’Hollander, Speed-up bounds for continuous system simulation on a homogeneous multiprocessor,in: International Conference on Parallel Processing, 318–324 (1981).Google Scholar
- 7.S.S. Lavenberg,”Computer Performance Modelling Handbook“, Notes and Reports in Computer Science and Applied Mathematics, Academic Press (1982).Google Scholar
- 8.C.H. Sauer, and K.M. Chandy, Computer systems performance modelling, IEEE Computer15: 157 –158 (1982).Google Scholar
- 9.R.M. Hord, The Illiac IV: the first Supercomputer, IEEE Spectrum21: 15 (1984).Google Scholar