Experiences with the Intel Hypercube

ABSTRACT

This paper presents a comparative analysis of the performance of two different parallel algorithms for solving the prefix problem on the Intel Hypercube. Estimates of the key parameters that determine the cost of communication between nodes are also given. It turns out that the ratio of communication cost to that of unit computation is quite high (is in the range of 240 to 470) for this class of machines.

REFERENCES

• 1: Y. Saad and M. Schultz. "Topological Properties of Hypercube", Department of Computer Science, Technical Report, DCS-RR-389, Yale University, 1985.
• 2: S. Lakshmivarahan and S. K. Dhall. "A new Hierarchy of Hypercube Architectures and their application", Second conference on Hypercube Multiprocessors, Oakridge National Laboratory, September 1986.
• 3: iPSC System Overview - Intel Scientific Computers, 1985.
• 4: S. Lakshmivarahan and S. K. Dhall. "New Parallel Algorithms for Solving First-order and Certain Second Order Linear Recurrence", Proceedings of the International Conference on Parallel Processing, 1981, pp. 841-846.
• 5: R. W. Hochney and C. R. Jesshope. "Parallel Computers", Adam Hilger, 1981.
• 6: D. Heller. "A Survey of Parallel Algorithms in Numerical Analysis", SIAM Review, Vol. 20, 1978, pp. 740-777.
• 7: Y. Saad and M.H. Schultz. "Data Communication in Hypercube", Research Report DCS-RR-428, Yale University, October 1985.

INDEX TERMS

Primary Classification:
  G. Mathematics of Computing
  G.1 NUMERICAL ANALYSIS
      G.1.0 General
          Subjects: Parallel algorithms

Additional Classification:
  C. Computer Systems Organization
  C.1 PROCESSOR ARCHITECTURES
      C.1.2 Multiple Data Stream Architectures (Multiprocessors)
          Subjects: Interconnection architectures (e.g., common bus, multiport memory, crossbar switch)

  I. Computing Methodologies
  I.3 COMPUTER GRAPHICS
      I.3.1 Hardware architecture
          Subjects: Parallel processing

General Terms:
Algorithms, Design, Performance