Lin Gan's paper accepted by FPL2014

Atmospheric modeling is an essential issue in the study of climate change. However, due to the complicated algorithmic and communication models, scientists and researchers are facing tough challenges in finding efficient solutions to solve the atmospheric equations. The atmospheric modeling team of HPGC focuses on developing and accelerating the complicated atmospheric equations through the most sophisticated platforms. In the latest work, we accelerate a solver for the three-dimensional Euler atmospheric equations through reconfigurable data flow engines. The left figure underneath shows the 25 point stencil kernel derived from the Euler equations. We first propose a hybrid design (shown as the right figure underneath) that achieves efficient resource allocation and data reuse between the host CPU and the DFE accelerator.

StencilInFPL2014,LinGan.jpeg

Furthermore, through algorithmic offsetting (the right figure underneath), fast memory table, and customizable-precision arithmetic, we map a complex Euler kernel into a single FPGA chip (shown as the left figure underneath), which can perform 956 floating point operations per cycle.

WorkflowInFPL2014,LinGan.jpeg

In a 1U-chassis, our CPU-DFE unit with 8 FPGA chips is 18.5 times faster and 8.3 times more power efficient than a multicore system based on two 12-core Intel E5-2697 (Ivy Bridge) CPUs, and is 6.2 times faster and 5.2 times more power efficient than a hybrid unit equipped with two 12-core Intel E5-2697 (Ivy Bridge) CPUs and three Intel Xeon Phi 5120d (MIC) cards.

The related paper titled A Highly-Efficient and Green Data Flow Engine for Solving Euler Atmospheric Equations was accepted by the 24thInternational Conference on Field Programmable Logic and Applications (FPL2014), which was held in Munich, Germany in 2-4 Sept, 2014. The PhD candidate, Lin Gan, who is also the first author of the paper, attended the conference in Munich and gave a presentation. Besides members from the atmospheric modeling team, co-authors also include faculty and students from Tsinghua University, Chinese Academy of Sciences, and Imperial College London. The paper will soon appear on-line, and will be indexed by the IEEE Xplore.