Cross memory pseudo-random mapping method for parallel computing

The invention discloses a cross memory pseudo-random mapping method for parallel computing, which uses a pseudo-random sequence generated by a maximum sequence code (m-sequence) to form an address mapping matrix, so that parallel access and parallel computing performance can be improved, and storage delay of a dynamic memory is slowed down. An H matrix adjusting method of address mapping is adopted, the weakness of the previous method is overcome, and it is ensured that the same H matrix can be used in forward mapping and reverse mapping. The generated address mapping matrix is combined with an exclusive-or function to form an address mapping circuit, the design is easy, and the structure is simple. For different address widths, an address segmentation mapping method is used for optimizing the pseudo-randomness of a mapping matrix so as to improve the parallel storage performance. 本发明公开了一种供并行计算用的交叉存储器伪随机映射方法，使用最大序列码(m-sequence)生成的伪随机序列来构成地址映射矩阵、可以提高并行存取和并性计算性能，减缓动态存储器的存储延迟。本发明采用了地址映射的一种H矩阵调整方法，克服了之前方法的弱点，保证了正