Premium
Portable, parallel transformation: Distributed‐Memory approach
Author(s) -
Covick Lawrence A.,
Sando Kenneth M.
Publication year - 1996
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/(sici)1096-987x(199606)17:8<992::aid-jcc8>3.0.co;2-m
Subject(s) - bottleneck , parallel computing , distributed memory , computer science , computation , transformation (genetics) , shared memory , distributed shared memory , parallel algorithm , algorithm , uniform memory access , memory management , semiconductor memory , computer hardware , embedded system , biochemistry , chemistry , gene
The four‐index transformation has a high ratio of data transfer to computation making it a potential “bottleneck” for parallel correlation energy determination. We present formulas for the communication times on different parallel architectures for an algorithm that is primarily designed for distributed‐memory machines. We also implemented the algorithm on two shared‐memory parallel computers, the Encore Multimax and the Alliant FX‐8, and measured wall clock times for several problem sizes and processor configurations. © 1996 by John Wiley & Sons, Inc.