PARALLEL JACOBI ALGORITHM FOR MATRIX DIAGONALIZATION ON TRANSPUTER NETWORKS

被引:3
|
作者
TERVOLA, P
YEUNG, W
机构
[1] Department of Physics, Queen Mary and Westfield College, London, E1 4NS, Mile End Road
关键词
SYMMETRICAL MATRICES; EIGENVALUES; EIGENVECTORS; JACOBI METHOD; MIMD MACHINES; TRANSPUTER NETWORKS; DISTRIBUTED MEMORY PROCESSORS;
D O I
10.1016/S0167-8191(05)80102-4
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
We present a parallel algorithm for the determination of the eigenvalues and eigenvectors of a real symmetric matrix. The algorithm allocates a certain number of columns to each of the transputers. The Jacobi cycle of annihilating the off diagonal elements consists of letting all the transputers perform Jacobi rotations concurrently, correcting for overlapping transformations and shuffling the sets of columns among the transputers. We develop formulae for the speedup and efficiency. Using Occam 2 we implement the algorithm on several transputer networks and compare the actual timings with our calculated results. We discuss the merits of this particular implementation.
引用
收藏
页码:155 / 163
页数:9
相关论文
共 50 条