Challenges in obtaining peak parallel performance with a Convex C240, parallel vector processor

ABSTRACT

This report examines the behavior of the Linpack 300×300 benchmark [Dongarra] on a parallel vector machine. It is observed that the performance of several parallel vector machines on this application is far below their nominal peak performance. Dissection of the internals of the algorithms shows how peak performance is limited. The insights gained provide guidance to algorithm developers as to ways to make maximum use of architectural strengths. System architects may gain insight about which system characteristics to optimize to increase the performance of future designs for this class of application.

REFERENCES

• 1: Patrick McGehearty,Convex Computer Corporation,3000 Waterview Plwy. PO Box 833851,Richardson,TX 75083-3851 (uiicdes.sun,uunet,hardvard,killer,usenix)!convex! patrick

INDEX TERMS

Primary Classification:
  C. Computer Systems Organization
  C.1 PROCESSOR ARCHITECTURES
      C.1.2 Multiple Data Stream Architectures (Multiprocessors)
          Subjects: Parallel processors**

Additional Classification:
  C. Computer Systems Organization
  C.1 PROCESSOR ARCHITECTURES
      C.1.2 Multiple Data Stream Architectures (Multiprocessors)
          Subjects: Array and vector processors
  C.4 PERFORMANCE OF SYSTEMS
      Subjects: Performance attributes

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

General Terms:
Algorithms, Design, Performance