论文部分内容阅读
通过数据并行的方式对一个成熟的叶轮机多块网格气动计算程序(MAP)进行了并行化处理,利用计算统一设备架构(CUDA)技术实现了在图形处理单元(GPU)上的并行计算.保留了原程序中的2阶空间迎风格式和隐式时间离散格式,并采用了隐式迭代对线性系统进行求解.经过2个叶轮机械算例的测试,与在传统的中央处理器(CPU)上运行的原程序相比,在计算结果完全一致的前提下,单GPU的计算速度最高可达单CPU计算速度的8.89倍,与四核并行的CPU计算相比可以得到2.39倍的加速.
By parallelizing the data of a multi-grid MAP of a mature impeller, parallel computing is implemented on a graphics processing unit (GPU) by using the CUDA technique. The second order space windward and implicit time discretization of the original program is retained, and the implicit iteration is adopted to solve the linear system.After two impeller mechanical examples of the test, compared with the traditional CPU (Central Processing Unit) Compared with the original program running on the premise of the same calculation results, single GPU computing speed up to 8.89 times the computational speed of a single CPU, compared with the quad-core parallel CPU calculation can be 2.39 times the acceleration.