Avoiding the square-root bottleneck in the Choleski factorization of a matrix on a parallel computer (Q1072331)
From MaRDI portal
 | This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use this page instead for the normal view: Avoiding the square-root bottleneck in the Choleski factorization of a matrix on a parallel computer |
scientific article; zbMATH DE number 3942901
| Language | Label | Description | Also known as |
|---|---|---|---|
| English | Avoiding the square-root bottleneck in the Choleski factorization of a matrix on a parallel computer |
scientific article; zbMATH DE number 3942901 |
Statements
Avoiding the square-root bottleneck in the Choleski factorization of a matrix on a parallel computer (English)
0 references
1986
0 references
The main difficulty in the parallel computation of the Cholesky factorization of a matrix is the fact that the square roots must be computed sequentially. We present a new way to compute the Cholesky factorization which allows the square roots to be (partially) overlapped. The biggest improvement is obtained when square roots are much slower than divides and multiplies (i.e. when they are implemented in software).
0 references
square-root bottleneck
0 references
parallel computation
0 references
Cholesky factorization
0 references
