基于CUDA的SMAC算法并行化

发布时间：2018-06-15 17:36

本文选题：计算流体力学 + 统一计算设备架构　；参考：《西安邮电大学学报》2016年05期

【摘要】：改进SMAC(Simplified Marker and Cell)算法,增强其对流体模拟的实时处理能力。采用点差分格式对求解压力场和速度更新的偏微分方程进行离散化;引入消除数据相关性的存储算法以减少数据传输,并借助分层存储机制提高访存比,采用并行归约增加线程并行度;在统一计算设备架构平台下,对离散化的SMAC算法进行改进、优化及并行化实现。纯粹计算及多次迭代模拟实验结果显示,改进算法提速明显,可实现对一般场景的实时模拟。
[Abstract]:The algorithm of simplified Marker and cell is improved to enhance its real time processing ability for fluid simulation. The point difference scheme is used to discretize the partial differential equations to solve the pressure field and the velocity update, the storage algorithm to eliminate the data correlation is introduced to reduce the data transmission, and the hierarchical storage mechanism is used to improve the memory access ratio. Parallel reduction is used to increase the degree of thread parallelism, and the discrete SMAC algorithm is improved, optimized and parallelized on the platform of unified computing equipment architecture. The experimental results of pure computation and iterative simulation show that the improved algorithm increases the speed obviously and can realize the real-time simulation of general scenes.
【作者单位】：西安邮电大学电子工程学院;西安邮电大学计算机学院;
【基金】：国家自然科学基金资助项目(61136002) 西安市科技发展计划资助项目(CXY1440(10))
【分类号】：O35;TP338.6

【相似文献】