FPGA Based Big Data Accelerator Design in Teaching Computer Architecture and Organization

摘要

In the past few years big data applications are becoming diverse and ubiquitous. There is a renewed interest in teaching senior level students to be professional in accelerator based computer architecture design and engineering. However, it poses a significant challenge to tutor the students with sufficient knowledge and practical skills in this area. In this paper, we propose a big data accelerator design project implemented on field-programmable gate array (FPGA) in teaching a computer architecture and organization course. The experimental system is carried out on a heterogeneous architecture using Xilinx Virtex 5 development boards. To achieve a modular accelerator implementation, several milestones are set to facilitate the on-time complete of the project. With the assistance of the FPGA-based experiment, most students have obtained a much more comprehensive understanding of the processor architecture and the accelerator design paradigm. Student feedback and survey illustrates the effectiveness and popularity of the FPGA-based project with milestones over simulation based experiments.