Abstract: Research on SIMD architecture has been at a low tide. One of the reasons is that there isn’t a good algorithm to perform conflict-free aceess to parallel memory. Judging from the coverage of access patterns, we believe that prime memory is the most ideal. At the end of the 1970s, BSP parallel computer adopted prime memory,However because of some problems unsolved, the effect is not ideal. Such problems are crossbar interconnection network, waste of memory space and inequality between the number of memory modules and processors. This paper presents a new approach to its implementation.