Abstract
This paper studies a model for pipeline programs with unidirectional data flow. The model takes communication cost into account and imposes a number of restrictions on pipeline structure, e.g., that the time required in a pipe stage to process a message is the same as the processing time in every other stage. Explicit expressions are derived for the execution time and speedup of model pipeline programs, and a necessary and sufficient condition for optimality is derived. The logistics of model pipeline programs are described analytically in terms of the utilization of computation and communication capacity.
Similar content being viewed by others
References
J. Baer,Computer Systems Architecture, Computer Science Press, Rockville Maryland (1980).
E. Gehringer,et al., The Cm* Testbed,Computer, pp. 40–53. (October 1982).
E. Gehringer, et al.,Parallel Processing, the Cm + Experience, Digital Press, Maynard, Massachusetts (1987).
M. Quinn,Designing Efficient Algorithms for Parallel Computers, McGraw Hill, New York (1987).
P. Kogge, Algorithm Development for Pipelined Processor,Proc. Int. Conf. on Par. Proc., pp. 217 (August 1977).
S. Dasgupta,Computer Architecture, Vol. 2 Wiley & Sons, New York (1989).
K. Hwang, F. Briggs,Computer Architecture and Parallel Processing, McGraw Hill, New York (1984).
H. Jagadish, et al., A Study of Pipelining in Computing Arrays,IEEE Trans on Comp,35:431–40 (1986).
G. Alaghband, A Parallel Pivoting Algorithm on a Shared Memory Multiprocessor,Proc. Int. Conf. on Par. Prog., pp. 177–80 (August 1988).
H. Stone,High Performance Computer Architecture, Addison-Wesley, Reading Massachusetts (1987).
S. Kunkel, J. Smith, Optimal Pipelining in Supercomputers,Proc. 13th Symposium on Computer Arch., pp. 404–11 (June 1986).
J. Gustafson, et al., Development of Parallel Methods for a 1024-processor Hypercube,SIAM J. Sci. Stat. Comput.,9:609–38 (1988).
M. Hagan,et al., Parallel Signal Processing Research on the HEP, Proc. Int. Conf. on Par. Proc., pp. 599–606 (August 1985).
T. Hoshino,PAX Computer, Addison-Wesley, Reading, Massachusetts (1985).
P. Burke, The Output of a Queueing System,Operations Res,4:699–704 (1956).
Invert a Matrix by Monte Carlo, Collected Algorithms from ACM, Vol. 1, ed., Webb Miller, Algorithm 166.
Additional information
Digital Equipment Corporation.
Rights and permissions
About this article
Cite this article
Gait, J. Speedup and optimality in pipeline programs. Int J Parallel Prog 18, 277–290 (1989). https://doi.org/10.1007/BF01407860
Received:
Revised:
Issue Date:
DOI: https://doi.org/10.1007/BF01407860