Multitasking the Householder Diagonalization Algorithm on the CRAY X-MP/48
In a previous lecture (computational implementation of the R—matrix method in atomic and molecular collision problems) it was mentioned that an important part of the computation in solving collision problems by R—matrix techniques was the diagonalization of the internal-region Hamiltonian matrices. These matrices are usually real, symmetric and dense (ii.e. non—sparse). All eigenvalues and eigenvectors are required to define the R—matrix. the matrices can be of various sizes, up to an order of a few thousand square.
KeywordsNest Loop Collision Problem Fast Memory Triangle Form Hamiltonian Matrice
Unable to display preview. Download preview PDF.
- Gourlay, A.R. and Watson, G.A., 1973, Computational methods for matrix eigenproblems. Wiley & Sons.Google Scholar
- Householder, A.S. and Bauer, F.L., 1959, Numerische Math. 1 Band 1. Heft, p29.Google Scholar
- Ralston, A. and Wilf, H.S., 1967, Mathematical methods for digital computers. II Wiley & Sons.Google Scholar
- Sawey, M. and Berrington, K.A., 1989, Comp. Phys. Commun.: submitted.Google Scholar