哈尔滨工程大学学报
哈爾濱工程大學學報
합이빈공정대학학보
JOURNAL OF HARBIN ENGINEERING UNIVERSITY
2014年
6期
661-666
,共6页
杨志国%黄兴%郑兴%段文洋
楊誌國%黃興%鄭興%段文洋
양지국%황흥%정흥%단문양
GPU%并行计算%CUDA%SPH方法%溃坝%水动力学%数值计算
GPU%併行計算%CUDA%SPH方法%潰壩%水動力學%數值計算
GPU%병행계산%CUDA%SPH방법%궤패%수동역학%수치계산
graphics processing unit ( GPU)%parallel computing%compute unified device architecture ( CUDA)%SPH method%dam breaking%hydrodynamic%numerical simulation
SPH方法是一种无网格的粒子方法,对于求解强非线性水动力学问题具有重要意义。随着粒子数增加,该方法的计算效率成为限制其大规模工程应用的重大瓶颈。可将大规模并行计算引入SPH方法中,以得到良好的计算加速效果。采用将GPU运用于SPH方法并行计算的技术,借助CUDA硬件计算架构,研究SPH方法的并行计算通用性问题。以二维溃坝问题作为数值算例,对GPU计算结果的稳定性和收敛性进行验证,比较CPU与GPU的计算效率。通过计算,验证了GPU在SPH方法并行计算应用中的可靠性、可行性以及高效性,为提高SPH方法的计算效率提供一种重要的参考途径。
SPH方法是一種無網格的粒子方法,對于求解彊非線性水動力學問題具有重要意義。隨著粒子數增加,該方法的計算效率成為限製其大規模工程應用的重大瓶頸。可將大規模併行計算引入SPH方法中,以得到良好的計算加速效果。採用將GPU運用于SPH方法併行計算的技術,藉助CUDA硬件計算架構,研究SPH方法的併行計算通用性問題。以二維潰壩問題作為數值算例,對GPU計算結果的穩定性和收斂性進行驗證,比較CPU與GPU的計算效率。通過計算,驗證瞭GPU在SPH方法併行計算應用中的可靠性、可行性以及高效性,為提高SPH方法的計算效率提供一種重要的參攷途徑。
SPH방법시일충무망격적입자방법,대우구해강비선성수동역학문제구유중요의의。수착입자수증가,해방법적계산효솔성위한제기대규모공정응용적중대병경。가장대규모병행계산인입SPH방법중,이득도량호적계산가속효과。채용장GPU운용우SPH방법병행계산적기술,차조CUDA경건계산가구,연구SPH방법적병행계산통용성문제。이이유궤패문제작위수치산례,대GPU계산결과적은정성화수렴성진행험증,비교CPU여GPU적계산효솔。통과계산,험증료GPU재SPH방법병행계산응용중적가고성、가행성이급고효성,위제고SPH방법적계산효솔제공일충중요적삼고도경。
As a mesh-free particle method, the SPH method is very important for dealing with the hydrodynamic problems with strong nonlinear problems. However, following the increase of the particle number, the calculation efficiency becomes a bottleneck for applying the method to the engineering practice. Good acceleration performance can be attained by applying massively parallel computing into the SPH method. In order to study the general parallel purpose of the SPH method, the parallel technology through the use of GPU on the compute unified device architec-ture platform has been used. The convergence and the stability of the computing results generated by GPU have been verified and the computational efficiency of GPU and CPU has been compared according to the computing re-sults by using a numerical example of the two-dimensional dam-breaking problem. The capability, feasibility and high efficiency of GPU computing have been proved by the results which provide important reference channels to in-crease the calculation efficiency of the SPH method.