2003 Fiscal Year Final Research Report Summary
Research on the execution environment for algorithms on a powerful model BSR^+ of parallel computation
| Project/Area Number |
14580398
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
計算機科学
|
|---|
| Research Institution | Kyushu Sangyo University |
|---|
Principal Investigator |
XIANG Limin Kyushu Sangyo University, Faculty of Information Science, Associate Professor, 情報科学部, 助教授 (60341294)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
USHIJIMA Kazuo Kyushu Sangyo University, Faculty of Information Science, Professor, 情報科学部, 教授 (40037750)
|
|---|
| Project Period (FY) |
2002 – 2003
|
| Keywords | Parallel Computation / Model of Parallel Computation / BSR / Parallel Algorithms / PRAM / BROADCAST Instruction |
|---|
| Research Abstract |
The PRAM (Parallel Random Access Machine) is a popular model of parallel computation. In the variants of PRAM, BSR^+ is the most powerful in computation. Based on the model BSR (Broadcasting with Selective Reduction), BSR^+ was proposed by the head investigator, etc., in IEEE Transactions on Parallel and Distributed Systems (vol. 10, no. 8) in 1999. Along with the LAN being made universal and the PC being of low cost, it is possible to build an execution environment for algorithms on BSR^+ model of parallel computation by use of several PC's in laboratory. One PC used as the Interconnection Unit and the Control Unit, and P (4 【less than or equal】 P 【less than or equal】 6) PC's as the P processors, the function of BSR^+ was implemented in this research. The number P of processors is a constant in any physical parallel computer, but the size N is a variable of the problem in many applications. By the principle that the design of algorithm is independent on the physical computer, the function transforming N into P should be supplied in the environment of parallel computation. Built on the LAN, the execution environment with the transforming function was tested and verified for algorithms on BSR^+ model of parallel computation in this research. The environment that BSR^+ algorithms can be executed on is helpful to the research on parallel algorithms and to the application of parallel algorithms.
|
