2012 Fiscal Year Final Research Report
Research on Data Resident Computing
| Project/Area Number |
22650008
|
|---|
| Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Computer system/Network
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
NAKAMURA Hiroshi 東京大学, 大学院・情報理工学系研究科, 教授 (20212102)
|
|---|
| Project Period (FY) |
2010 – 2012
|
| Keywords | コンピューティング / 実行モデル / アーキテクチャ / 高性能化 / 低消費電力化 |
|---|
| Research Abstract |
Currently, inefficiency of data transfers between computing components prevents computer systems from further performance improvement or power reduction, rather than computing components themselves. To overcome this problem, this research proposes a new computing paradigm called Data Resident Computing, which can directly optimize data transfers. This paradigm is applied to many-core processors, and two new computing methods are proposed. One is data compression by making use of locative information of paths of data transfers, and the other is novel process scheduling by considering the interferences between data transfers of different cores. Experimental results reveal the effectiveness of the proposed techniques.
|
