Advanced loop parallelization and integrated vectorization
Project/Area Number |
18K18032
|
Research Category |
Grant-in-Aid for Early-Career Scientists
|
Allocation Type | Multi-year Fund |
Review Section |
Basic Section 60050:Software-related
|
Research Institution | The University of Tokyo |
Principal Investigator |
Sato Shigeyuki 東京大学, 大学院情報理工学系研究科, 助教 (90779464)
|
Project Period (FY) |
2018-04-01 – 2022-03-31
|
Project Status |
Completed (Fiscal Year 2021)
|
Budget Amount *help |
¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000)
Fiscal Year 2020: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2019: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
Fiscal Year 2018: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
|
Keywords | プログラム合成 / 自動並列化 / コンパイラ / ループ並列化 / ベクトル化 |
Outline of Final Research Achievements |
In this project, we studied automatic parallelization of complicated reduction loops. Our developed techniques enable us to systematically transform serial intuitive specifications into divide-and-conquer implementations amenable to various parallel computers. Specifically, we developed 1) a branch elimination method by operator extraction, 2) a parallelization method based on dynamic behaviors, and 3) a vectorization method based on data shuffling as part of the fundamental compiler technology. To evaluate these techniques, we also developed 4) a benchmark suite of reduction loops. Furthermore, we developed implementation techniques for 5) parallel lexing and 6) parallel regular expression matching as case studies on applications of complicated reductions.
|
Academic Significance and Societal Importance of the Research Achievements |
今日において,既にさまざまな並列計算機が至る所にあると同時に,新しい並列計算機が次々と開発されている.その並列計算機のハードウェア性能を引き出すには,並列プログラムが必要になる.しかし,並列プログラムを上手く作ることは人間にはとても難しく,しばしば,間違った計算をしたり,性能が悪くなったりする.本研究で扱った自動並列化技術は,人間にとって直観的で間違えにくい計算仕様を,並列計算機に都合が良い計算形式に変換するものである.これにより,多様な並列計算機のハードウェア性能を,簡単かつ安全に引き出すことができるようになる.
|
Report
(5 results)
Research Products
(5 results)