Dynamic limited directory scheme using data locality

The locality of data reference affects system performance significantly in distributed shared memory systems. One method of improving the performance is to use data allocation methods reflecting the locality of data references. This paper proposes a cache coherence scheme by utilizing the near processor locality, and evaluates its performance. The proposed scheme reduces memory overhead by using a presence table and a directory pool. If a processor is within a predefined distance, its identification is set to the presence table. Otherwise, its identification is stored to a directory pool. This scheme can solve the disadvantages of the limited directory scheme and the dynamic point allocation directory scheme. Three application programs, which have various shared data, are simulated to evaluate the performance of proposed scheme. The results show that the optimal data allocation method with the proposed scheme reduces the directory pool access up to 4.9 times.