Previous | [ 1] | [ 2] | [ 3] | [ 4] | [ 5] | [ 6] | [ 7] | [ 8] | [ 9] | [ 10] | [ 11] | [ 12] | [ 13] | [ 14] | [ 15] | [ 16] | [ 17] | [ 18] | [ 19] | [ 20] |
¡@
Chun-Yuan Lin and Yeh-Ching Chung*
Institute of Molecular and Cellular Biology
*Department of Computer Science
National Tsing Hua University
Hsinchu, 300 Taiwan
E-mail: ychung@cs.nthu.edu.tw
Array operations are useful in a large number of important scientific codes, such as
molecular dynamics, finite-element methods, climate modeling, etc. It is a challenging
problem to provide an efficient data distribution for irregular problems. Multi-dimensional
(MD) sparse array operations can be used in atmosphere and ocean sciences, image
processing, etc., and have been an extensively investigated problem. In our previous
work, a data distribution scheme, Encoding-Decoding (ED), was proposed for twodimensional
(2D) sparse arrays. In this paper, ED is extended to be useful for MD sparse
arrays first. Then, the performance of ED is compared with that of Send Followed Compress
(SFC) and Compress Followed Send (CFS). Both theoretical analysis and experimental
tests were conducted and then shown that ED is superior to SFC and CFS for all
of evaluated criteria.
Received January 6, 2005; revised May 25 & September 28, 2005; accepted November 2, 2005.
Communicated by Gen-Huey Chen.