Advertisement

Vector Processing Techniques on CYBER 205

  • Karl-Heinz Schlosser

Abstract

In this lecture we discuss some aspects of algorithm development on today’s vector computers, especially on the CYBER 205. After a short discussion of the different kinds of parallelism in the architecture of this machine it is shown, that for the same algorithm the optimal instruction is a function of the data structure. Then the weakness of automatic vectorization is shown and the necessity for semantic vectorization — that means the development of new algorithms — is pointed out. Therefore there is a requirement of programming languages with some parallel features allowing the scientists to express parallel numerical algorithms in a natural manner.

We will discuss such language kernels and give an overview of the realization in the new CYBER 205 PASCALV compiler. Finally we will discuss some numerical algorithms
  1. (1)

    iterative linear equation solvers

     
  2. (2)

    adaptive quadrature methods

     
  3. (3)

    numerical optimization with evolution strategies in respect of semantic vectorization. We will give an example of the optimal PASCALV implementation of well suited algorithms and some performance specifications.

     

Keywords

Vector Addition Index Expression Arithmetic Unit Startup Time Semantic Vectorization 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Bernutat-Buchmann, Ulrike and Krieger, Jost, 1982, Analysis of elementary vector algorithms, in: “Conferences on CYBER 200 in Bochum, Proceedings”, Bernutat-Buchmann, U., Ehlich, H., Schlosser K.-H., eds., Rechenzentrum, Ruhr-Universität, Bochum, F.R.G., p. 125–142.Google Scholar
  2. Bernutat-Buchmann, Ulrike and Krieger, Jost, 1984, Evolution strategies in numerical optimization on vector computers, in: “Parallel Computing 83, Proceedings”, Feilmeier, M., Joubert, G., Schendel, U., eds., North-Holland, Amsterdam, p. 99–105.Google Scholar
  3. Ehlich, Hartmut, 1984, “PASCALV — der PASCAL compiler für den vektorrechner CYBER 205”, Bochumer Schriften zur Parallelen Datenverarbeitung 5, Ehlich, H., ed., Rechenzentrum, Ruhr-Universität, Bochum, F.R.G.Google Scholar
  4. Händler, Wolfgang, 1977, The impact of classification schemes on computer architecture, in: “Int. Conf. on Parallel Processing 1977”, Baer, J.L., ed., IEEE Inc., New York, p. 7–15.Google Scholar
  5. Helmbrecht, Detlef and Schlosser, Karl-Heinz, 1984, Erfahrungen mit PASCALV, in:“Conferences on CYBER 200 in Bochum 1984, Proceedings”, to be published.Google Scholar
  6. Lambiotte, Jules Joseph and Voigt, Robert G., 1975, The solution of tridiagonal linear systems on the CDC STAR 100 computer, ACM Transactions on Mathematical Software 1: 4, p. 308–329.Google Scholar
  7. Peuser, Michael,1982, Eine systematik der CYBER 205 vektorbefehle, in: “Conferences on CYBER 200 in Bochum, Proceedings”, Bernutat-Buchmann, U., Ehlich, H., Schlosser, K.-H., ed., Rechenzentrum, Ruhr-Universität Bochum, F.R.G., p. 107–124.Google Scholar
  8. Ramamoorthy, C.V. and Li, H.F., 1977, Pipeline archi- tecture, Computing Surveys 9: 1, p. 61–102.Google Scholar
  9. Robinson, I.G., 1971, Adaptive Gaussian integration, The Australian Computer Journal 3, No. 3, p. 126–129.Google Scholar
  10. Robinson, I.G., 1976, An algorithm for automatic integration using the adaptive Gaussian technique, The Australian Computer Journal 8, No. 3, p. 106–115.Google Scholar
  11. Schlosser, Karl-Heinz, 1982, Vektoriziering auf Sprache- bene, in: “Conferences on CYBER 200 in Bochum, Proceedings”, Bernutat-Buchmann, U., Ehlich, H., Schlosser, K.-H., eds., Rechenzentrum, Ruhr-Universität Bochum, F.R.G., p. 35–86.Google Scholar
  12. Schumann, Martin, 1983, Vektorizierung adaptiver quadraturverfahren, Diplomarbeit, Bochum.Google Scholar
  13. Schwefel, H.-P., 1977, “Numerische Optimierung von Computer Modellen der Evolutionsstrategie”, Birkhäuser, Basel. Traub, Joseph F., 1973, Iterative solution of tridiagonal systems on parallel or vector computers, in: “Complexity of Sequential and Parallel Numerical Algorithms”, Traub, J.F., ed., Academic Press, New York, p. 49–82.Google Scholar
  14. Wieczorek, Martin, 1984, “Zur exakten formalen spezifikation der Syntax und Semantik von Machinensprachen”, Bochumer Schriften zur Parallelen Datenverarbeitung 6, Ehlich, H., ed., Rechenzentrum, Ruhr-Universität, Bochum, F.R.G.Google Scholar

Copyright information

© Plenum Press, New York 1985

Authors and Affiliations

  • Karl-Heinz Schlosser
    • 1
  1. 1.Ruhr-Universität BochumBochumGermany

Personalised recommendations