TY - GEN
T1 - FPGA implementation of viscous function in a package for computational fluid dynamics
AU - Mishra, Dipikarani
AU - Hatto, Mao
AU - Kuhara, Takuya
AU - Fujita, Naoyuki
AU - Osana, Yasunori
AU - Amano, Hideharu
PY - 2015/2/27
Y1 - 2015/2/27
N2 - Fast Aerodynamics Routines (FaSTAR) is a recent fluid dynamics software package by JAXA. It is a convenient tool for users since various solvers and automatic generation functions of grid data are provided. Although FaSTAR is one of the fastest CFD software in the world, it is still inefficient in parallel execution because of it's irregular and unpredictable data structure. The acceleration of FaSTAR using FPGAs have been exerted, and some important cores were implemented. Here, a main loop of the viscous function which occupies about 16% of total was selected as a target of accelaration, and implemented on Xilinx's FPGA Virtex-6 achived 2.3 times speed of software execution. The operational freqency was 100 MHz with 144653 slices and the resource usage was 58% for register utilization and 96% for LUT utilization. Considering the complexity of the computation, at least 2.3 times performance improvement is expected compared with the software on Core 2 Duo 2.4GHz.
AB - Fast Aerodynamics Routines (FaSTAR) is a recent fluid dynamics software package by JAXA. It is a convenient tool for users since various solvers and automatic generation functions of grid data are provided. Although FaSTAR is one of the fastest CFD software in the world, it is still inefficient in parallel execution because of it's irregular and unpredictable data structure. The acceleration of FaSTAR using FPGAs have been exerted, and some important cores were implemented. Here, a main loop of the viscous function which occupies about 16% of total was selected as a target of accelaration, and implemented on Xilinx's FPGA Virtex-6 achived 2.3 times speed of software execution. The operational freqency was 100 MHz with 144653 slices and the resource usage was 58% for register utilization and 96% for LUT utilization. Considering the complexity of the computation, at least 2.3 times performance improvement is expected compared with the software on Core 2 Duo 2.4GHz.
KW - Computational fluid dynamics (CFD) field programmable gate arrays (FPGA) scientific computation
UR - http://www.scopus.com/inward/record.url?scp=84925435348&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84925435348&partnerID=8YFLogxK
U2 - 10.1109/CANDAR.2014.33
DO - 10.1109/CANDAR.2014.33
M3 - Conference contribution
AN - SCOPUS:84925435348
T3 - Proceedings - 2014 2nd International Symposium on Computing and Networking, CANDAR 2014
SP - 608
EP - 610
BT - Proceedings - 2014 2nd International Symposium on Computing and Networking, CANDAR 2014
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2nd International Symposium on Computing and Networking, CANDAR 2014
Y2 - 10 December 2014 through 12 December 2014
ER -