This project covers the following four activities in the area of restructuring compilers and the application of the technology to applications. (1) The previous work of the PI on data alignment will be extended to include data distribution. This will permit data and computations to be mapped to physical processors of a MIMD machine in such a way that parallelism is exploited while minimizing communication overhead. (2) Loop transformation technology for imperfectly nested loops will be developed. Current technology can handle only `perfectly nested loops` where assignments occur only in the innermost loops. However, many practical applications have imperfectly nested loops. In this project techniques for performing linear loop transformations and tiling imperfectly nested loops will be developed. (3) The project will develop methods for producing parallel code for applications involving sparse matrices. (4) Above techniques will be applied to logic simulation applications.
