论文部分内容阅读
本文针对实际应用中处理器阵列故障率极小的普遍现象,提出针对故障时阵列的实时高效快速重构算法.不同于传统的基于可用处理器的重组理念,本文利用反向思维方式,从极少数目的故障处理器及其所处的位置出发,通过对阵列中的少量故障处理器进行移位操作,使得故障处理器尽可能地被移动至同一逻辑列,最大限度地保留使用原有的阵列布局,进而实现逻辑阵列的高效快速重构.实验结果表明,对于故障率小于2%的32×32物理阵列,本文提出的算法实现了极限加速,并且不损失任何逻辑列.对故障率小于1%的64×64物理阵列,提出的算法的重构速度至少加速了65%,而逻辑阵列的损失率可控制在1.64%以内.因此,当阵列的故障率极小时,本文算法可以高效快速地重构阵列.
In this paper, aiming at the common phenomenon that the failure rate of the processor array is extremely small in practice, this paper proposes a real-time and efficient fast reconfiguration algorithm for the fault array.Compared with the traditional reconfigurable concept based on available processors, this paper uses the reverse thinking mode, A small number of fault processor and its location, through the array of a small number of fault processor shift operation, making fault processor as possible to be moved to the same logic column, to maximize the use of the original array Layout, and then realize the efficient and rapid reconstruction of the logic array.The experimental results show that the proposed algorithm achieves the limit acceleration for any 32 × 32 physical array with a failure rate of less than 2%, and does not lose any logical columns.For a failure rate of less than 1 % Of the 64 × 64 physical array, the proposed algorithm at least 65% speed up the reconstruction speed, while the logic array loss rate can be controlled within 1.64%. Therefore, when the failure rate of the array is extremely small, the algorithm can be efficient and fast Reconstruction of the array.