Study Of Folding Structure Prediction Of Protein Optimization Algorithm In Heterogeneous Distributed & Parallel Environment

Since the structure of protein molecules is very complex, its structure prediction should build a simplified model, and use optimization algorithms to search the minimum energy value under the steady state, but the amount of calculation is very great. Therefore, making use of distributed & parallel technology to enhance execution efficiency of the algorithm is the key point of protein folding structure prediction. The heterogeneous distributed & parallel environment which formed by the CPU and GPU, can improve the execution speed and processing power of computing system, which provides a theoretical basis for increasing the execution efficiency of the optimization algorithm.This thesis analyzes the organizing structure, function theory and performance-optimizing of the heterogeneous distributed & parallel environment, and researches the programming mathod of the CUDA platform. Moreover, the folding structure prediction of protein algorithm is optimized on heterogeneous distributed & parallel environment, and several optimization approaches for the genetic annealing algorithm are designed. Firstly, the paper constructs the program slicing-scheduling mechanism. It adjusts and slices the programme by evaluating the intensity of calculation. Secondly, it dynamically allocates the resource of the system according to the size of the data and the status of the calculation. Thirdly, it realizes the algorithm on the heterogeneous environment, based on share memory optimization, unrolling optimization, instruction optimization and memory coalescing techmology. Thus, the algorithm is transformed into a parallel processing mode which can enhance the execution efficiency.On the basis of 2D AB off-lattice model, this paper runs an experiment on CUDA platform, using the optimized GAA. These results show the validity of the optimization of protein folding structure prediction algorithm. Comparing to the experiment under the CPU serial environment, the high execution speed and better rate can be got by using this method, which has great meaning and value on the research of the folding structure of protein.