Efficient organization of control structures in distributed implementations

Abstract

A new technique for the management of control structures in distributed implementations of dynamic process systems is presented. Instead of storing the runtime stacks of parallel processes as linked lists of activation blocks in a heap structure, the local stacks of several parallel processes, which are executed on the same processor element, are stored in an interleaved manner on a single physical stack (within each processor element), called the meshed stack. The technique ensures that there is almost no overhead for the evaluation of single processes due to the parallel environment. In principle, the meshed stack technique is independent of the implemented language. We explain it for the parallel implementation of functional languages.