This Article 
 Bibliographic References 
 Add to: 
Analytical Estimation of Vector Access Performance in Parallel Memory Architectures
May 1993 (vol. 42 no. 5)
pp. 616-624

A limiting factor in high-performance vector computers is the rate at which data can be moved to and from memory during vector loads and stores. To increase the bandwidth of vector memory operations, several investigators have proposed the use of noninterleaved storage mappings, also known as storage schemes. A method of analyzing the performance of vector references for the class of storage schemes referred to as XOR schemes is described. The proposed measure of relative performance, the variability of an access, is defined and its computation is outlined. The use of variability as a performance indicator is demonstrated and compared with performance measurements made using simulation. One of the key aspects of the variability measure is its capability to lend insight to the transient behavior of vector accesses.

[1] CONVEX Computer Corp.,CONVEX Architecture Reference, Oct. 1988.
[2] P. Budnik and D. Kuck, "The organization and use of parallel memories,"IEEE Trans. Comput., vol. C-20, no. 12, pp. 1566-1569, Dec. 1971.
[3] D. Lawrie and C. Vora, "The prime memory system for array access,"IEEE Trans. Comput., vol. C-31, no. 5, pp. 435-442, May 1982.
[4] D. T. Harper III and J. R. Jump, "Vector access performance in parallel memories using a skewed storage scheme,"IEEE Trans. Comput., vol. C-36, no. 12, pp. 1440-1449, 1987.
[5] J. Frailong, W. Jalby, and J. Lenfant, "XOR-Schemes: A flexible data organization in parallel memories," inProc. Int. Conf. Parallel Processing, 1985, pp. 276-283.
[6] A. Norton and E. Melton, "A class of boolean linear transformations for conflict-free power-of-two stride access," inProc. Int. Conf. Parallel Processing, 1987, pp. 247-254.
[7] B. Rau, M. Schlansker, and D. Yen, "The Cydra 5 Stride-Insensitive Memory System," inProc. Int. Conf. Parallel Processing, 1989, pp. 1242-1246.
[8] G. Sohi, "High-bandwidth interleaved memories for vector processors--A simulation study," Tech. Rep., Comput. Sci. Dep., Univ. of Wisconsin-Madison, Sept. 1988.
[9] B. R. Rau, "Pseudo-random interleaved memory," inProc. 18th Int. Symp. Comput. Architecture, Toronto, Canada, May 1991, pp. 74-83.
[10] R. Raghavan and J. P. Hayes, "On randomly interleaved memories," inProc. Supercomputing '90 Conf.(New York), Nov. 1990.
[11] D. T. Harper III, "Block, multistride vector, and FFT accesses in parallel memory systems,"IEEE Trans. Parallel Distributed Syst., vol. 2, pp. 43-51, Jan. 1991.
[12] W. Oed and O. Lange, "On the effective bandwidth of interleaved memories in vector processing systems,"IEEE Trans. Comput., vol. C-34, no. 10, pp. 949-957, Oct. 1985.
[13] W. Oed and O. Lange, "Modelling, measurement, and simulation of memory interference in the Cray X-MP,"Parallel Comput., vol. 3, pp. 343-358, 1986.
[14] S. J. Leon,Linear Algebra with Applications. New York: Macmillan, 1986.
[15] R. J. McEliece,Finite Fields for Computer Scientists and Engineers. Boston, MA: KIuwer Academic, 1987.

Index Terms:
vector access performance; parallel memory architectures; high-performance vector computers; vector memory operations; noninterleaved storage mappings; vector references; XOR schemes; performance measurements; variability measure; memory architecture; parallel architectures; performance evaluation.
D.T. Harper, III, Y. Costa, "Analytical Estimation of Vector Access Performance in Parallel Memory Architectures," IEEE Transactions on Computers, vol. 42, no. 5, pp. 616-624, May 1993, doi:10.1109/12.223682
Usage of this product signifies your acceptance of the Terms of Use.