A linear array of processors with partially shared memory for parallel solution of PDE

A multiprocessor architecture with partially shared memory blocks is proposed which is suited for the successive approximation of scientific computing problems, such as matrix operations and partial differential equations. The topology is a ring, but the system configuration is a subset of an orthogonal multiprocessor that I.D. Scherson (1989) and K. Hwang (1989) have already proposed. An experimental system was built which was used to solve a heat-conduction problem. The computing time is about 600 s in the case of 256*256 grid points.<
>