Data Space Fusion Based Approach for Effective Alignment of Computation and Data

The alignment of computation and data has an important affection on the performance of parallel programs running on distributed memory multicomputers. This paper presents a new approach of the effective alignment of computation and data, namely the data space fusion based approach for partitioning computation and data, which can be used to solve the computation and data decomposition problems on distributed memory multicomputers. This approach can maximize parallelism and minimize communication over processors by exploiting the parallelism of computation space as high as possible and using the technique of data space fusion to optimize data distribution. The approach can be integrated with data replication and offset alignment naturally and therefore can make the communication overhead as low as possible. The experimental results on eight programs show that the approach presented in this paper is effective for aligning computation and data.