本文選題：GPU + 異構(gòu)平臺(tái)�。� 參考：《振動(dòng)與沖擊》2014年13期

【摘要】：為了解決傳統(tǒng)的串行有限元分析方法計(jì)算耗時(shí)多精度低的問(wèn)題,基于GPU并行計(jì)算能力在CUDA架構(gòu)下建立了一套兼顧精度和效率的高層結(jié)構(gòu)有限元分析的CPU-CPU的異構(gòu)平臺(tái)�；贑PU-GPU異構(gòu)平臺(tái)研究了高層結(jié)構(gòu)地震響應(yīng)算法,將整個(gè)時(shí)間步積分在GPU中計(jì)算完成,每一時(shí)間步下利用基于GPU的預(yù)處理共軛梯度迭代法求解線性方程組獲得該時(shí)刻的位移,最終實(shí)現(xiàn)了基于GPU的Newmark-β法。通過(guò)算例驗(yàn)證了本文所提方法的高精度、高效率。
[Abstract]:In order to solve the problem that the traditional serial finite element analysis method has low computation time and low precision, a heterogeneous platform of CPU-CPU based on GPU parallel computing ability is established under the CUDA architecture, which takes into account the accuracy and efficiency of finite element analysis of high-rise structures. Based on CPU-GPU heterogeneous platform, the seismic response algorithm of tall structures is studied. The integral of the whole time step is calculated in GPU, and the displacement of linear equations is obtained by using the preprocessing conjugate gradient iteration method based on GPU in each time step. Finally, the Newmark- 尾 method based on GPU is realized. An example is given to verify the high accuracy and high efficiency of the proposed method.
【作者單位】： 哈爾濱工業(yè)大學(xué)深圳研究生院;
【基金】：國(guó)家自然科學(xué)基金重大國(guó)際(中美)合作研究項(xiàng)目(51261120374);國(guó)家自然科學(xué)基金面上項(xiàng)目(51278155);國(guó)家自然科學(xué)基金青年基金項(xiàng)目(51008048)
【分類號(hào)】：TU973.31

【相似文獻(xiàn)】

相關(guān)重要報(bào)紙文章 前1條

1 華榮;不要妄談“協(xié)同設(shè)計(jì)”[N];建筑時(shí)報(bào);2008年

，

本文編號(hào)：1818399