Development of data-driven custom pipeline processing system for accelerating big data processing

2017 Fiscal Year Final Research Report

• Project/Area Number: 26540019
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Computer system
• Research Institution: Tokyo Institute of Technology (2015-2017); Japan Advanced Institute of Science and Technology (2014)
• Principal Investigator: Sato Yukinori 東京工業大学, 学術国際情報センター, 特任講師 (30452113)
• Research Collaborator: LUK Wayne Imperial College London, Department of Computing, Professor ; MENCER Oskar Maxeler Technologies Ltd, CEO/CTO
• Project Period (FY): 2014-04-01 – 2018-03-31
• Keywords: FPGAアクセラレータ / カスタムコンピューティング / ビッグデータ / ハード・ソフト協調設計 / データ局所性

Outline of Final Research Achievements

In this research, we developed a data-driven custom pipeline processing system for accelerating big data processing, which attempts to make full use of FPGA accelerators that can customize their memory hierarchy and datapath with high-level synthesis tool chains. We focused on the following core technologies that enhance data locality within time and space dimensions; (1) data locality optimization by loop tiling, (2) a new dataflow profiling technique based on semantics of data manipulation, (3) a novel framework for large-scale data stream processing. As a result of implementation and evaluation of these technologies, we showed that our proposal is beneficial to improving productivity of customization.

Free Research Field

計算機アーキテクチャ