MATRIX COMPUTATIONS ON THE CM-200
Abstract:
Abstract. The parallel computer CM-200 consists of a very large number of simple processors connected in a mesh. The peak performance is very high, but it is not clear how easy it is to write efficient programs in high-level languages. A straight-forward implementation of the Householder QR algorithm in CM Fortran is shown to be slow. Another implementation is presented, with better performance, but still not comparable to the low-level implementation in CMSSL.
Citations
| 1143 | Matrix Computations – Golub, Loan - 1989 |
| 129 | A Users Guide to PVM Parallel Virtual Machine – Beguelin, Dongarra, et al. - 1991 |
| 9 | Bj orck, Numerical Methods – Dahlquist, A - 1974 |
| 3 | High Performance Fortran Language Speci – F - 1993 |
| 3 | Fortran 90 Explained, Oxford Science – Metcalf, Metcalf, et al. - 1990 |
| 1 | Bit-level Jacobi-like algoritms for eigenvalue and singular value decompositions – Schimmel - 1991 |
| 1 | An assessment of the connection machine, tech. rep – Schreiber - 1990 |
