为了解决计算时间过长这一制约有限元计算在实际工程中应用的瓶颈问题,提出了一种利用并行处理方式加速有限元计算的方法。从有限元计算理论着手,详细分析了引入图形处理器并行处理方式的利弊,协调了图形处理器、中央处理器之间的任务分配和数据交换,并围绕图形处理器的计算资源设计出了全新的有限元计算处理架构。研究结果表明,效率提高了近22倍。
In order to eliminate the lasting calculation bottleneck on applications of finite element,a new method was proposed to accelerate the computing of FEM by using the existing modern GPU and programmable shading tool.Under the FEM theory,this paper analyized the pros and cons of using GPU parallel process,coordinated the task assignment and data exchange between GPU and CPU and designs a new FEM processing architecture by using the computational resources of GPU.The GPU parallel simulation calculation rate is 22 times faster than that by using CPU's,as well as it has useful and practical value in solving the project problems.