基于整数规划的最大目标阵列重构算法

Integer programming-based reconfiguration algorithm for maximum target arrays

为了求解得到最大规模的目标处理器阵列,提出了一种基于整数规划的阵列重构算法。该算法将阵列中的无故障可用单元转化为整数规划问题中的变量,把处理器网状结构的约束指定为变量的等式或不等式,从而将求解最大规模目标阵列的问题转化为最大化变量的和的整数规划问题,并使用相应的整数规划求解器对其进行求解。实验表明,与现有的算法相比,该算法能增大目标阵列的规模,从而提高处理器系统的可靠性。

In order to obtain a maximal size target processor array, an algorithm based on integer programming is proposed to reconfigure processor arrays. In this algorithm, fault-free process elements in the host array are converted into variables of the integer programming problem, and equations about these variables are defined according to the constraints of mesh structure of processor arrays. Thus, the problem of constructing maximum size target arrays can be converted into the inte- ger programming problem that maximizing the sum of these variables. The problem can be solved by using relevant integer programming solvers. Experimental results show that compared with existing algorithms, the proposed algorithm can in- crease the scale of target arrays effectively and improve the reliability of processor system.