计算机科学与探索
計算機科學與探索
계산궤과학여탐색
JOURNAL OF FRONTIERS OF COMPUTER SCIENCE & TECHNOLOGY
2014年
8期
897-905
,共9页
赵慰%赵永华%刘晓辉%何力新
趙慰%趙永華%劉曉輝%何力新
조위%조영화%류효휘%하력신
统一计算架构(CUDA)%格点积分%第一性原理%局域轨道法
統一計算架構(CUDA)%格點積分%第一性原理%跼域軌道法
통일계산가구(CUDA)%격점적분%제일성원리%국역궤도법
compute unified device architecture (CUDA)%grid integrals%first principle%local orbitals
第一性原理计算软件在密度泛函理论的发展中起着重要作用。相比平面波方法,局域轨道法更适合处理大规模多体问题。随着问题规模的不断增大和计算机计算能力的提升,软件的并行加速成为一个重要课题,MPI(message passing interface)结合GPU(graphic processing unit)实现的异构并行是一个新的解决方案。基于局域轨道法的第一性原理计算软件MESIA(massive electronic simulation based on systematically improvable atomic bases)经过MPI+OpenMP+CUDA三级并行,单GPU取得了约15倍的加速比,同时表现出了良好的可扩展性。测试结果同时验证了使用GPU计算可以保证计算精度。
第一性原理計算軟件在密度汎函理論的髮展中起著重要作用。相比平麵波方法,跼域軌道法更適閤處理大規模多體問題。隨著問題規模的不斷增大和計算機計算能力的提升,軟件的併行加速成為一箇重要課題,MPI(message passing interface)結閤GPU(graphic processing unit)實現的異構併行是一箇新的解決方案。基于跼域軌道法的第一性原理計算軟件MESIA(massive electronic simulation based on systematically improvable atomic bases)經過MPI+OpenMP+CUDA三級併行,單GPU取得瞭約15倍的加速比,同時錶現齣瞭良好的可擴展性。測試結果同時驗證瞭使用GPU計算可以保證計算精度。
제일성원리계산연건재밀도범함이론적발전중기착중요작용。상비평면파방법,국역궤도법경괄합처리대규모다체문제。수착문제규모적불단증대화계산궤계산능력적제승,연건적병행가속성위일개중요과제,MPI(message passing interface)결합GPU(graphic processing unit)실현적이구병행시일개신적해결방안。기우국역궤도법적제일성원리계산연건MESIA(massive electronic simulation based on systematically improvable atomic bases)경과MPI+OpenMP+CUDA삼급병행,단GPU취득료약15배적가속비,동시표현출료량호적가확전성。측시결과동시험증료사용GPU계산가이보증계산정도。
First principle calculation package plays an important role in the development of density functional theory. Local orbital method is more suitable than plane wave method when calculating big system. Parallelism of the pack-age becomes more and more important as the increasing of system scale and the development of HPC (high perfor-mance computing). The combination of MPI (message passing interface) and CUDA (compute unified device archi-tecture) is a new way to accelerate the package. Accelerating the MESIA (massive electronic simulation based on systematically improvable atomic bases) package based on LCAO (linear combination of atomic orbitals) method with MPI+OpenMP+CUDA provides 15-fold speedup per GPU and good scalability. The test results also prove that computing with GPU can keep the accuracy of results.