Development of General Computation Models for Many-core Environments

2014 Fiscal Year Final Research Report

• Project/Area Number: 23240002
• Research Category: Grant-in-Aid for Scientific Research (A)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Fundamental theory of informatics
• Research Institution: The University of Tokyo (2014); National Institute of Informatics (2011-2013)
• Principal Investigator: SADAKANE Kunihiko 東京大学, 情報理工学系研究科, 教授 (20323090)
• Co-Investigator(Kenkyū-buntansha): SHIBUYA Tetsuo 東京大学, 医科学研究所, 准教授 (60396893) ; FUKUNAGA Alex 東京大学, 大学院総合文化研究科, 准教授 (90452002) ; HU Zhenjiang 国立情報学研究所, アーキテクチャ科学研究系, 教授 (50292769)
• Project Period (FY): 2011-04-01 – 2015-03-31
• Keywords: メニーコア / GPGPU / 並列アルゴリズム / 省スペースアルゴリズム

Outline of Final Research Achievements

We have proposed a parallel computation model AGPU to analyze asymptotic behavior of GPU algorithms. Though precise complexities of algorithms depend on architectures of GPU devices, these are bounded from above by a constant factor. By using our AGPU model, we can estimate performance of algorithms which do not depend on specification of GPU devices nor input data. We have also proposed a comparison-based sort algorithm which has optimal I/O complexity. Experimental results on real GPUs show that our algorithm runs 1.9 times faster than an existing algorithm.

Free Research Field

アルゴリズムとデータ構造