1996 Fiscal Year Final Research Report Summary
Implementation of Parallel Functional Programming Systems
| Project/Area Number |
06558039
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (A)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 試験 |
|---|
| Research Field |
計算機科学
|
|---|
| Research Institution | University of Tokyo |
|---|
Principal Investigator |
TAKEICHI Masato University of Tokyo, Graduate School of Engineering, Professor, 大学院・工学系研究科, 教授 (10011165)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
TANAKA Tetsuro University of Tokyo, Educational Computer Center, Assistant Professor, 教育用計算機センター, 助教授 (60251360)
MATSUOKA Satoshi Tokyo Institute of Technology, Assistant Professor, 大学院・情報理工学研究科, 助教授 (20221583)
IWASAKI Hideya Tokyo University of Agriculture and Technology, Faculty of Engineering Assistant, 工学部, 助教授 (90203372)
YONEZAWA Akinori University of Tokyo, Graduate School of Science, Professor, 大学院・理学系研究科, 教授 (00133116)
|
|---|
| Project Period (FY) |
1994 – 1996
|
| Keywords | Functional Programming / Parallel Evaluation / Program Optimization |
|---|
| Research Abstract |
This project aims at development of functional programming systems for parallel computers. Implementation of a parallel functional system Parallel Gofer on the AP1000 computer has been finished and under evaluation. This implementation is based on the Gofer system developed by Mark Jones and accepts any Gofer programs for sequential evaluation. Programs are allowed to include references to extended library functions for parallelization.
Several new ideas for this implementation have been published already. One of such ideas is so-called "unboxing" techniques for data construction. This implementation showed that the idea is promising while some optimization should be considered for practical application.
Along with this implementation, a novel idea for optimization has been explored and implemented. Although most optimization so far relies on heuristics, our new system is completely mechanical. It is based on hylomorphisms which comes from research on constructive algorithmics. This technique is applicable to sequential and parallel functional programs.
These results have been made public at international conferences and published in the proceedings.
|
