By Topic

Performing BMMC permutations in two passes through the expanded delta network and MasPar MP-2

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

3 Author(s)
Wisniewski, L.F. ; Dept. of Comput. Sci., Dartmouth Coll., Hanover, NH, USA ; Cormen, T.H. ; Sundquist, T.

This paper examines routing of BMMC (bit-matrix-multiply/complement) permutations on two types of multistage interconnection networks: the expanded delta network and the global router of the MasPar MP-2. BMMC permutations are an important class of permutations that has been well-studied on various multistage networks. The class of BMMC permutations includes as subclasses Gray-code and inverse Gray-code permutations and the entire subclass of bit-permute/complement (BPC) permutations, which in turn includes matrix transpose (with power-of-2 dimensions), bit reversal, vector reversal, hypercube, and matrix reblocking permutations. There are four results in this paper. First, we use linear-algebraic techniques to derive an algorithm to perform any BMMC permutation in at most two passes on the expanded delta network. Second, we use linear-algebraic techniques to derive an algorithm to perform any BMMC permutation in at most two passes on the global router of the MasPar MP-2. Third, we use linear-algebraic and combinatorial analysis to determine the distribution of all BMMC permutations when routed naively through the MP-2 global router and show that most, but not all, BMMC permutations require only one or two passes anyway. We can apply our two-pass algorithms in those cases when naive routing requires more than two passes. Fourth, we present experimental evidence to support our analysis.

Published in:

Frontiers of Massively Parallel Computing, 1996. Proceedings Frontiers '96., Sixth Symposium on the

Date of Conference:

27-31 Oct. 1996