Zynq Cluster for CFD Parametric Survey

Fast Aerodynamics Routines FaSTAR is a state of the art Computational Fluid Dynamics CFD software package to enable high precise analysis. Due to its complicated data structure from unstructured grid, the acceleration with GPU or massively parallel machines is not efficient. Although a hardware accelerator on an FPGA is a hopeful candidate, the complicated FaSTAR program is difficult to pick up time consuming cores and implement them on an FPGA. In practical aircraft design, a parametric survey, which executes FaSTAR jobs in parallel with different conditions is commonly used. Here, we propose a Zynq cluster as a cost and power efficient solution of FaSTAR parametric survey. By introducing high-level synthesis and partial reconfiguration, the FaSTAR job with a specific condition runs on a simple node with a Zynq-7000 AP SoC. Now a part of FaSTAR job can be executed on FPGA of Zynq board about 1.3 times faster than Intel's Xeon E5-2667 2.9i¾?GHz software.

[1]  Satoru Yamamoto,et al.  Systolic Architecture for Computational Fluid Dynamics on FPGAs , 2007 .

[2]  Nachiket Kapre,et al.  Zedwulf: Power-Performance Tradeoffs of a 32-Node Zynq SoC Cluster , 2015, 2015 IEEE 23rd Annual International Symposium on Field-Programmable Custom Computing Machines.

[3]  Bernd Einfeld On Godunov-type methods for gas dynamics , 1988 .

[4]  Hideharu Amano,et al.  Partially reconfigurable flux calculation scheme in advection term computation , 2013, 2013 International Conference on Field-Programmable Technology (FPT).

[5]  G. Botella,et al.  Aerodynamics Analysis Acceleration through Reconfigurable Hardware , 2008, 2008 4th Southern Conference on Programmable Logic.

[6]  Neil W. Bergmann,et al.  Automatic Self-Reconfiguration of System-on-Chip Peripherals , 2007 .

[7]  Hideharu Amano,et al.  Reconfigurable out-of-order mechanism generator for unstructured grid computation in computational fluid dynamics , 2012, 22nd International Conference on Field Programmable Logic and Applications (FPL).