Abstract
In this paper we consider mesh connected computers with multiple buses, providing broadcast facilities along rows and columns. A tight bound of 0(n?) is established for the number of rounds required for semigroup computations on n values distributed on a 2-dimensional rectangular mesh of size n with a bus on every row and column. The upper bound is obtained for a skewed rectangular mesh of dimensions n3/8 × n5/8 This result is to be contrasted with the tight bound of θ(n1/6) for the same problem on the square (n1/2 × n1/2) mesh [PR]. This implies that in the presence of multiple buses, a skewed configuration may perform better than a square configuration for certain computational tasks. Our result can be extended to the J-dimensional mesh, giving a lower bound of ω(1/nd2d) and an upper bound of O(d2d+1 n 1/d2d).
| Original language | English |
|---|---|
| Title of host publication | Proceedings of the 1st Annual ACM Symposium on Parallel Algorithms and Architectures, SPAA 1989 |
| Editors | F.T. Leighton |
| Publisher | Association for Computing Machinery, Inc |
| Pages | 138-147 |
| Number of pages | 10 |
| ISBN (Electronic) | 089791323X, 9780897913232 |
| DOIs | |
| State | Published - 1 Mar 1989 |
| Externally published | Yes |
| Event | 1st Annual ACM Symposium on Parallel Algorithms and Architectures, SPAA 1989 - Santa Fe, United States Duration: 18 Jun 1989 → 21 Jun 1989 |
Publication series
| Name | Proceedings of the 1st Annual ACM Symposium on Parallel Algorithms and Architectures, SPAA 1989 |
|---|
Conference
| Conference | 1st Annual ACM Symposium on Parallel Algorithms and Architectures, SPAA 1989 |
|---|---|
| Country/Territory | United States |
| City | Santa Fe |
| Period | 18/06/89 → 21/06/89 |
Bibliographical note
Publisher Copyright:© 1989 ACM.
Funding
*Computer Science Department, Stanford University, Stanford, CA 94305. Supported in part by a Weizmann fellowship and by contract ONR N00014-88-K-0166 l Department of Applied Mathematics, The Weizmann Institute, Rehovot 76100, Israel. Part of this work was carried out while this author was visiting Stanford university. Supported in part by a weizmarm fellowship, by contract ONR N00014-88-K-0166 and by a grant of Stanford Center for Integrated Systems.
| Funders | Funder number |
|---|---|
| ONR N00014-88-K-0166 l Department of Applied Mathematics | ONR N00014-88-K-0166 |
| Stanford Center for Integrated Systems |