Parallel Computing And Efficiency Analysis On Daubechies Wavelets Density Functional

Daubechies wavelat density functional method has broad application in the field of materials science and nanosimulations, etc. Daubechies wavelet has its own characteristics of localization, which make it an ideal choice for electronic structure calculations basic set. BigDFT is a DFT massively parallel electronic structure code using a wavelet basis set with the capability to use a linear scaling method. BigDFT can satisfy many applications both fit accuracy and locality requirements, and can also take advantage of MPI, OpenMP, CUDA and OpenCL to accelerate simulation system. With the development of computer technology, the computing power of supercomputers growing, hardware architecture is complex. Under this situation,how to combine the characteristics of code and computing hardware features is a big challenge.This dissertation mainly discuss the parallel calculation of BigDFT wavelet program and its efficiency running on high performance computing platform. Further applications of BigDFT code on different sizes of silicon atoms have been performed.After analysis on the parallel efficiency based on various simulations, we find the most optimal solutions for our model simulations.The main work can be divided into the following four aspects:(1) We firstly report the recent development on the high performance parallel computing technology about electronic structure simulations.(2) We built suitable parallel complier environment to supply the applications of BigDFT parallel computing. A variety of the efficiency of compliers and math library have been compared to test the suitability for the parallel application of BigDFT.(3) Before the numerial experiments or simulations, we systematic explore the user-friendly interface between high-performance computing platform and the software, including the basis commands and often used tools such as LSF. We complete the installation of BigDFT code in this constructed high-performance computing platform.(4) We test the different version of BigDFT.By testing, we find the best solution of parallel computing on BigDFT. The optimal solution is reached. Under the condition that computing resources are limited and under the usage rules of super computing platform, when the processes number of each node is close to the full load state of the node number and the processes number is consistent, the parallel efficiency is the best and computing resources utilization efficiency is high.