论文部分内容阅读
采用基于列交换的Ｇａｕｓｓ－Ｊｏｒｄａｎ并行算法来解决空气动力学中超音速高阶面元法的稠密矩阵求逆问题，该方法采取了块循环数据分配方式，尤其对超立方体结构的并行机系统来说具有通讯优势。在４台ＳＧＩ工作站构成的２×２网格上进行的实验表明，对秩为１０００左右的矩阵可得到５７％～６４％的效率。
The Gauss-Jordan parallel algorithm based on column exchange is used to solve the dense matrix inversion problem of hypersonic high-order surface element method in aerodynamics. This method adopts block cyclic data allocation method, especially for hypercube structure parallel machine system Has the advantage of communication. Experiments on 2 × 2 grids made up of 4 SGI workstations show that 57% to 64% efficiency is achieved for a matrix with a rank of about 1000.