I'm really not clear on your question. Perhaps you could refine it a bit. There are two basic architectures that I am familiar with. They are called "parallel" and "pipeline".
In a "parallel" architecture you arrange some number, usually a power of 2, processors in a square array or grid. Each processor has local memory that only it can access and is has memory which is common to some number of its neighbors, maybe four or eight. The problem is divided into parts and each processor works on a part.
In a "pipeline" architecture parts of a problem are handled by a string of processors. As each processor finishes its task it passes results on to the next processor in the string.
Remember -- Google and Wikipedia are your friends. Use them.