| Project/Area Number |
62065002
|
|---|
| Research Category |
Grant-in-Aid for Specially Promoted Research
|
| Allocation Type | Single-year Grants |
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
TANAKA Hidehiko The Univ. of Tokyo, Fac. of Engineering, Professor, 工学部, 教授 (60011102)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KITSUREGAWA Masaru Univ. of Tokyo, Inst. of Indust. Science, Assistant Professor, 生産技術研究所, 助教授 (40161509)
相田 仁 東京大学, 工学部, 講師 (00175712)
|
|---|
| Project Period (FY) |
1987 – 1990
|
| Project Status |
Completed (Fiscal Year 1990)
|
| Budget Amount *help |
¥205,000,000 (Direct Cost: ¥205,000,000)
Fiscal Year 1990: ¥25,000,000 (Direct Cost: ¥25,000,000)
Fiscal Year 1989: ¥42,000,000 (Direct Cost: ¥42,000,000)
Fiscal Year 1988: ¥78,000,000 (Direct Cost: ¥78,000,000)
Fiscal Year 1987: ¥60,000,000 (Direct Cost: ¥60,000,000)
|
|---|
| Keywords | Knowledge Information Processing / Parallel Processing / Symbolic Computing / Inference Machine / Logic Programming / Knowledge Base / Database Machine / Relational Database / 相互結合網 / 多次元クラスタリング / 同時実行制御 / 論理型プログラミング / 並列デバッガ / 並列ガーベジコレクション / 同時実行制御アルゴリズム / 関係データベース処理 / 関係データベース / データべースマシン / 知識ベース / 理論型プロケラミング / データベースマシン / 知識ベースマシン |
|---|
| Research Abstract |
Large scale knowledge information processing system prototype was developed and studied. The system consists of parallel inference subsystem, which executes knowledge information processing programs in high speed by parallel processing, and knowledge base management Subsystem, which manages large ammount of knowledge information efficiently.
As for the research on parallel inference subsystem, Parallel Inference Engine PIE64 was studied and developed. PIE64 consists of 64 processing elements called Inference Unit (IU) and two high performance interconnection network systems which support automatic load balancing. Although some part of LSI design remains, the development, functional test, and evaluation of IU boards and interconnection network were successfully completed. We could show their realizability and effectiveness from these results. "Load Partitioning Method based on Profiling" and "Generation Scavenging GC on Distributed-Memory Parallel Computers" were proposed as new manageme
… More
nt methods of parallel symbolic processing on PIE64.
As for the research on parallel logic programming language, a new parallel logic programming language : FLENG was designed. A prototype FLENG processing system was developed and used extensively as a basic research tool. Highly functional debugger for parallel programs was developed on the system. Knowledge information processing application programs was developed and its characteristics were examined and analyzed.
As for the research on knowledgeーbase subsystem, we evaluated the performance of Functional Disk System in detail, which we have so far investigated to clarify secondary storage system. Based on the experimental results, we built the knowledge base managements subsystem prototype which consists of four clusters, each of which adopts tightly coupled shared memory multiprocessor architecture. We examined it could attain high performance. We also proposed novel multiーdimensionally clustered joint processing algorithm. We implemented and evaluated its effectiveness. In order to support multiーuser access, we proposed new concurrency control protocol for shared nothing machine. Extensive simulation study was performed. We could obtain optimal algorithm. Less
|
