A Design Environment for Structured Mapping of Signal Processing Applications on Parallel Processors

  • Moe Razaz

Abstract

We present an integrated software environment called Taurus, which is capable of structured mapping of signal processing applications on parallel computers. An application is first converted into a directed graph representation which is then turned into a multiprocessor code with the help of a code generator. Given the hardware interconnection topology and specification, a scheduler determines in what order the multiprocessor code should be mapped onto the individual processors in the hardware platform. The parallel implementation is shown in the form of a Gannt chart so that the user can see graphically the speed-up and processor utilisation. The major advantages of this environment are: (i) an intensive signal processing application can be easily implemented on a parallel platform (ii) processor specification and interconnection topology are user definable so that the same software can be used for implementation on different hardware platforms and (iii) a user can interact with the environment to enhance the performance of the parallel implementation. The design philosophy and the organisation of the integrated environment are presented and discussed.

Keywords

Convolution Editing 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    W.B. Ackerman, Data flow languages, IEEE Comput.vol. 15 (1982).Google Scholar
  2. 2.
    T. Agerwala, Putting Petri Nets to work, IEEE Compt.12:85 (1979).CrossRefGoogle Scholar
  3. 3.
    R.E. Crochiere, and A.V. Oppenheim, Analysis of linear digital networks, Proc. IEEE 63:581–595 (1975).CrossRefGoogle Scholar
  4. 4.
    A.L. Davis, and R.M. Keller, Data flow program graphs, IEEE Comput. vol.15 (1982).Google Scholar
  5. 5.
    R.M. Karp et al., The organisation of computations for uniform recurrence equations, J. ACM 14:563–590(1967).CrossRefGoogle Scholar
  6. 6.
    G. Korn, High speed block diagram languages for microprocessors, miniomputers in instrumentation, control and simulation, Comput. Elec.Eng.4:143–159 (1977).CrossRefGoogle Scholar
  7. 7.
    D.G. Messerschmit, A tool for structured functional simulation, lEEE J.Comm.vol. SAC-2 (1984).Google Scholar
  8. 8.
    J.L. Peterson, Petri Nets, Comput. Survey vol. 9, no. 3 (1967).Google Scholar
  9. 9.
    S. French, Sequencing and Scheduling, Ellis Horwood (1982).Google Scholar
  10. 10.
    T.C. Hu, Parallel sequencing and assembly line problems, Oper. Res.pp841–848 (1961).Google Scholar
  11. 11.
    N.F. Chen and C.L. Liu, Proc. Sagamore Comp. Con. on Parallel Processing, ppl-16, Springer Verlag, N.Y. (1974).Google Scholar
  12. 12.
    T.L. Adam et al., Comm. ACM 17:685–690 (1974).CrossRefGoogle Scholar
  13. 13.
    M. Razaz, Structured DSP Design and Development in a Multiprocessing System, BT Research Report, ppl-74, (March 1993).Google Scholar
  14. 14.
    M. Razaz and K.A. Marlow, Toward an automatic mapping of DSP algorithms onto parallel processors, Applications of Supercomputers in Eng.3:463–477 (1993).Google Scholar
  15. 15.
    M. Razaz and K.A. Marlow, An integrated multiprocessor system for DSP Design and Development, in: Int. Symp. on DSP for Communication Systems (Sept. 1992).Google Scholar
  16. 16.
    M. Razaz and K.A. Marlow, Scheduling DSP algorithms for parallel multiprocessor environment, in: 3rd IMA Conf. Maths. in Signal Processing.pp10–27 (Dec. 1992).Google Scholar
  17. 17.
    K.A. Marlow and M. Razaz, A new precompiler for mapping DSP applications to multiprocessing systems, Transputer Applications and Systems, 1:296–311 (1993).Google Scholar
  18. 18.
    K.A. Marlow and M. Razaz, Visualisation and analysis of multiprocessor DSP design implementations, in: Advances in Parallel & Vector Processing, B. Topping and M. Papadrakakis eds.,pp245–249 (1994).Google Scholar
  19. 19.
    M. Razaz and K.A. Marlow, A multiprocessor algorithm scheduler for signal and image processing, Signal Processing VII: Theories and Applications (Holt, Cowan, Grant and Sandham eds.) 3:1605–1608 (1994).Google Scholar
  20. 20.
    M. Razaz and K.A. Marlow, Performance comparison of two scheduling algorithms for parallel architectures, Advances in Parallel and Vector Processing, B. Topping and M. Papadrakakis eds., pp59–68 (1994).Google Scholar
  21. 21.
    M. Razaz and K.A. Marlow, A transputer-based parallel DSP environment, in: Proc. IEE Colloq. on High Performance Applications of Parallel Architectures, pp8.1-8.6, London, (Feb. 1994).Google Scholar
  22. 22.
    M. Razaz and K.A. Marlow, Designing DSP-based systems using multiple processor hardware platforms, Proc. IEEEISCAS, pp109–112 (1994).Google Scholar
  23. 23.
    M. Razaz, T. Spendiff and K.A. Marlow, Code generator for parallel implementation of intensive algorithms on multiple DSP chips, in: Conf. on DSP Chips in Real Time Measurement and Control, UK,pp8.1-8.10 (Sept. 1997).Google Scholar
  24. 24.
    M.. Razaz and K.A. Marlow, Taurus: A multiprocessor DSP prototyping environment, IEEE SiPS, pp263–272 (1997).Google Scholar
  25. 25.
    Texas Instruments, TMS320C40 Parallel-processing DSPs, Product Bulletin (1993).Google Scholar
  26. 26.
    Meiko Ltd., Computing Surface, C for CS TOOLS, Technical User Manual (1991).Google Scholar
  27. 27.
    Signal Processing Worksystem (SPW), Comdisco Systems, Inc., Foster City, CA, (1991).Google Scholar
  28. 28.
    J.C. Bier et al., Gabriel, IEEE Micro.pp28–45 (Oct. 1990).Google Scholar
  29. 29.
    The Almagest: Manual for Ptolemy Version 0.3.1, Department of EECS, University of California, Berkeley, USA, (Jan. 1992).Google Scholar
  30. 30.
    3L Ltd., C4x Parallel C Version 2.0.2, User manual (1995).Google Scholar
  31. 31.
    J.L. Hennessey and D.A. Patterson, Computer Architecture A Quantitative Approach, Morgan Kaufmann, San Francisco, pp378 (1995).Google Scholar

Copyright information

© Springer Science+Business Media New York 1999

Authors and Affiliations

  • Moe Razaz
    • 1
  1. 1.School of Information SystemsUniversity of East AngliaNorwichUK

Personalised recommendations