Creating a Massively Parallel and Distributed Computation Framework by Exploiting Short Idle Periods
| Project/Area Number |
16H02801
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Software
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
Ino Fumihiko 大阪大学, 情報科学研究科, 教授 (90346172)
|
|---|
| Project Period (FY) |
2016-04-01 – 2020-03-31
|
| Project Status |
Completed (Fiscal Year 2019)
|
| Budget Amount *help |
¥17,030,000 (Direct Cost: ¥13,100,000、Indirect Cost: ¥3,930,000)
Fiscal Year 2019: ¥2,470,000 (Direct Cost: ¥1,900,000、Indirect Cost: ¥570,000)
Fiscal Year 2018: ¥5,720,000 (Direct Cost: ¥4,400,000、Indirect Cost: ¥1,320,000)
Fiscal Year 2017: ¥3,640,000 (Direct Cost: ¥2,800,000、Indirect Cost: ¥840,000)
Fiscal Year 2016: ¥5,200,000 (Direct Cost: ¥4,000,000、Indirect Cost: ¥1,200,000)
|
|---|
| Keywords | GPU / マルチタスク / 並列分散処理 / 高速化 |
|---|
| Outline of Final Research Achievements |
In this work, we developed software techniques for exploiting milliseconds of idle periods that occur on remote graphics hardware (i.e., GPU) in home and office. The developed techniques are capable of accelerating compute-intensive scientific computation on shared GPUs while providing fast screen updating on the GPUs. We further developed a program translator that automatically generates parallel programs for multi-GPU environments, and demonstrated that the GPUs in home and office are useful for accelerating applications, such as distributed deep learning, that involve communication between computing nodes.
|
| Academic Significance and Societal Importance of the Research Achievements |
分散深層学習のような，計算ノード間の通信を伴う応用は，これまで高性能計算センターに設置された，専有のGPUクラスタ上で高速化されてきた．本研究の成果は，家庭やオフィスに設置された計算機群にGPUを追加するだけで，日常業務を処理しながら，ノード間通信を伴う分散深層学習を高速化できることを明らかにしていて，導入コストの低い共創型高性能計算システムの礎となることが期待でき，社会的にも学術的にも意義深い研究成果である．
|
Report
(5 results)
Research Products
(58 results)
