The focus of this research is to design and implement an optimizing C* compiler for hypercube multicomputers and distribute the compiler to the hypercube user community. The compiler will translate C* into C. The use of C as a target language takes advantage of existing C compilers to reduce development time and expense. It also makes the compiler more portable. To move the compiler to other hypercubes or shared memory multiprocessors, only the routing library must be reimplemented. Preliminary results are encouraging. Hand-compiled C* programs have achieved reasonable speedup on a commercial hypercube. The long-term goal of the research is to investigate the power of the synchronous, data parallel approach to parallel algorithm development. Many numerical and nonnumerical algorithms can be expressed in an elegant fashion using this methodology. If C* compilers can be written that allow for the reasonable efficient execution of parallel algorithms on the Connection Machine, multicomputers, and shared memory multiprocessors, then there would be a whole class of algorithms for which no reprogramming would be necessary in order to move an application from one parallel architecture to another.
