2014 Fiscal Year Final Research Report
Research and development of macro parallelization methods targeted to stream-based programs
| Project/Area Number |
24300020
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Partial Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Computer system/Network
|
|---|
| Research Institution | University of Tsukuba |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
WADA Koichi 筑波大学, システム情報系, 教授 (30175145)
NAKANO Koji 広島大学, 工学研究科院, 教授 (30281075)
YUNOKI Seiji 独立行政法人理化学研究所, 柚木計算物性物理化学研究室, 准主任研究員 (70532141)
|
|---|
| Project Period (FY) |
2012-04-01 – 2015-03-31
|
| Keywords | ハイパフォーマンスコンピューティング / GPUコンピューティング / ストリーム指向 |
|---|
| Outline of Final Research Achievements |
The stream-based program has been applied among scientific and industrial computations due to the widely populated manycore accelerators such as GPU. The performance of single GPU is high due to the massively parallel execution of the program. However, considering parallel computing using multiple accelerators, programmers must consider a method for parallelization of the program and its communication timings of the divided programs. Therefore, it is hard to maintain the performance from the parallelized program. This research project aims to develop a programming foundation that can parallelize the stream-based program into multiple GPUs automatically and achieves a scalable performance from the parallel program for GPU.
|
| Free Research Field |
並列分散処理
|
