GPGPU Programming Framework based on a Shared Memory Model and Scheduling Optimization

2014 Fiscal Year Final Research Report

• Project/Area Number: 24500060
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Computer system/Network
• Research Institution: Mie University
• Principal Investigator: OHNO Kazuhiko 三重大学, 工学(系)研究科(研究院), 講師 (20303703)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 高性能計算 / GPGPU / CUDA / 並列プログラミング言語 / 言語処理系 / プログラミングモデル / 自動最適化

Outline of Final Research Achievements

Although Graphics Processing Units (GPU) is regarded as a promising platform for high performance computing, the productivity and reusability of the current programming framework CUDA are not sufficient. Therefore, we are developing an improved framework named MESI-CUDA. MESI-CUDA provides easier framework hiding low-level architecture, while high performance is achieved by the automatic optimization. As the research results, we introduced logical thread mapping and supported dynamic data structures. To improve the execution performance, we also developed memory access optimization schemes such as utilizing the shared memories and logical-to-physical thread mapping.

Free Research Field

並列プログラミング言語