| |
| | CS267: Notes for Lecture 9 (part 2), Feb 13, 1996 |
 | | This algorithm is well suited to an s-by-s mesh of processors, for which we will now measure the performance. |
 | | Rather than describe this algorithm, due to Ho, Johnsson and Edelman, in detail, we refer the reader to paper Parallel Numerical Linear Algebra, by Demmel, Heath, and van der Vorst, in volume 7 of the Class Reference Material. |
 | | As mentioned above, matrix multiplication with a 1D layout on a ring of processors can be done using a communication pattern very similar to the one needed for computing gravity in the "Sharks and Fish 2 problem. |
| www.cs.berkeley.edu /~demmel/cs267/lecture11/lecture11.html (3038 words) |
|