Hardware Computing
Vision Related Acceleration in FPGA


Time Series Data Mining
There are some interesting characters in the time series data mining. The streaming data can be generated at a high speed with unknown length. The information is both the data and the order of data. There are also some high throughput and real time analysis requirements. Based on these features, we are inspired to pipeline and parallel processing.

Programming Model and Platforms
Machine learning and data mining are gaining increasing attentions of the computing society. FPGA provides a highly parallel, low power, and flexible hardware platform for this domain, while the difficulty of programming FPGA greatly limits its prevalence. MapReduce is a parallel programming framework that could easily utilize inherent parallelism in algorithms. We describe FPMR, a MapReduce framework on FPGA, which provides programming abstraction, hardware architecture, and basic building blocks to developers.
Another platform of us about integrating FPGA into Clouds is our 'wukong' platform. In this work, we divide the FPGA into slots, and each slot can be configured as accelerator for at runtimey using Dynamic Partial Reconfiguration (DPR) technology. Using DPR, we solve the abstraction, sharing, compatibility and security problem when integrating FPGA into Clouds. To maximize the performance of our heterogeneous FPGA-based cloud platform, we propose the 'benefit-based' metric and the scheduling algorithm accoding to the metric.
