Key Techniques Research Of The Parallelization L-system Based On CUDA

Plants as a common natural landscape in real life,the construction of 3d model has been the research focus in the field of computer application,and it has been in special effects,agricultural education,computer animation and other fields has been widely used.Now how to use the computer to efficiently and accurately simulate its model is a hot and difficult point in the research of computer graphics.In plant morphological simulation,L-system is the most classical rule-based modeling method.However,this process mainly uses string recursive iteration to extend the grammar of L-system,so its grammar generation efficiency is a key point that cannot be ignored.For the problem,the paper main research contents and conclusions are as follows:(1)Proposes a parallel decomposition algorithm.The algorithm can be classified by the characteristics of L-system grammar,according to different types of L-system using different decomposition algorithm to decompose the related rules.For the non-branching L-system,the algorithm can dynamically handle parallelization according to matrix multiplicative associativity,to solve the problem of character grammar deviation caused by parallelization.For the branch L-system,the algorithm can extraction into the stack string rules to solve the parallel computing brought by branch structure,the control of its character into and out of the stack time,and its string grammar classified by the level of depth,parallel to accelerate at the same level of L-grammar and iteration to a high level,thus the realization decompose multithreaded processing characters,and improve the efficiency of the generation of L-grammar.For the randomized L-system,the algorithm can use the dynamic adjustment of the position index of the string of the module according to the randomness of the character grammar.(2)Parallelization acceleration of grammar iteration in L-system using CUDA computing platform.This method takes the initial grammar and iterative rule of L-system as the prerequisite,6 different types of classical L-system grammars are modeled separately for test samples.Finally,for the modeling results,this paper presents the efficiency comparison between CUDA parallel decomposition algorithm,multithreaded parallel algorithm and single-core high density CPU algorithm.(3)According to the designed parallel algorithm and the established parallel L-system model,this paper designs and develops the L-system based on CUDA computing platform.This system use L-system parallel decomposition algorithm,multithreading algorithm and serial algo-rithm,comparing the grammar of L-system with different orders of magnitude in the experiment.The experimental results show that different string grammars magnitude,the efficiency of grammar improvement is different.When the order of magnitude of string grammar in Lsystem reaches millions,the efficiency of parallel algorithm is 17%-27% higher than that of the high density single-core CPU algorithm.