2018 Fiscal Year Final Research Report
Study on Automatic Code Generation of Java Coarse Grain Parallel Processing for Multi-Platform
Project/Area Number |
16K00174
|
Research Category |
Grant-in-Aid for Scientific Research (C)
|
Allocation Type | Multi-year Fund |
Section | 一般 |
Research Field |
High performance computing
|
Research Institution | Meiji University |
Principal Investigator |
|
Project Period (FY) |
2016-04-01 – 2019-03-31
|
Keywords | 粗粒度タスク並列処理 / ローカルタスク協調実行 / 並列化コンパイラ / Java / Fork/Join Framework / メニーコア / POWER8 / ダイナミックスケジューリング |
Outline of Final Research Achievements |
Many-core processors are used to realize high-performance computing for servers and smartphones. In order to maximize the performance of multicore processors, it is required to utilize the inherent parallelism of target programs. Therefore, this research proposed the task-drive coarse grain parallel processing that realizes the coarse grain parallelization for Java programs. Also, in order to use many-core processor efficiently, this research proposed the local task cooperative execution method to use not only coarse grain parallelism but also local task parallelism. These schemes are implemented on the developed parallelizing compiler. From the results of performance evaluation on multi-platform environment such as x86, POWER and ARM, effectiveness of the Java coarse grain parallelization was confirmed.
|
Free Research Field |
並列処理
|
Academic Significance and Societal Importance of the Research Achievements |
本研究はマルチプラットフォーム環境において,Java処理系に備わっているFork/Join Framework(並列処理ソフトウェア)を利用して,粗粒度タスク並列処理を実現するタスク駆動型粗粒度並列処理手法を提案している.本手法は,並列処理の実行環境に依存せず,Javaプログラムの並列性を最大限に活用できることに優位性がある.また,本手法は高性能なサーバからPCやAndroidスマートフォンに至るまで,容易に利用可能なシステムソフトウェア技術であり,学術的意義に加えて社会的意義も大きい.
|