## Computing eigenvalues and eigenvectors of a dense real symmetric matrix on the Ncube 6400 |

### What people are saying - Write a review

We haven't found any reviews in the usual places.

### Common terms and phrases

6400 Shirish Chinchalkar allnodes applied backward accumulation based on row communication overhead computational effort compute Pjx2 compute_vec(i computes the Householder conquer algorithm contains the upperdiagonal Cornell Theory Center DENSE REAL SYMMETRIC diagonal matrix distributed by rows divide and conquer Dunigan 91 Efficiency of parallelization eigenvalues and eigenvectors Eispack subroutine end end endif end floating point operations flops formed by backward forward accumulation Householder transformations Householder vector based hypercube interprocessor communication large problems matrix Qi Mflops n x n orthonormal matrix n)v/vTv NCUBE 6400 Shirish node processors number of floating number of processors numerical experiments orthonormal matrix Qi parallel algorithms Pjx2 in order problem size pTv)v/vTv Q is orthonormal Q1Q2 where Q2 QDQT Qi are distributed QjAQi QR algorithm QR method real symmetric eigenvalue real symmetric matrix rows of Qi sequential shows the variation single sweep looks speedup total number TQL2 Algorithm TRED2 Algorithm TRED2 and TQL2 Variation in efficiency wrap fashion