Matrix compression by common subexpression elimination
(English)Manuscript (preprint) (Other academic)
In this report a method for common subexpression elimination in matrices isexplored. The method is applied to several types of matrices occurring in numericalsimulations. In all cases, the cost of a matrix-vector multiplication is reduced by asignificant amount. The amount of storage required for the eliminated matrices isalso less than that required for the original matrices. When the proposed method isapplied to the Fourier transform matrix, the output is equivalent to the fast Fouriertransform. For some matrices used in the fast multipole method for dislocationdynamics, the cost of a matrix-vector multiplication is reduced from O(p^6) to O(p^4.5),where p is the expansion order. Using an expansion order of 5, one can expect a factorof four speedup of the fast multipole part of a dislocation dynamics code.
IdentifiersURN: urn:nbn:se:kth:diva-11566OAI: oai:DiVA.org:kth-11566DiVA: diva2:277754
QC 201008052009-11-202009-11-202010-08-05Bibliographically approved