Optimizing the time cost of multiple groups of parallel processes within a parallel structure

Abstract

In a shared-memory parallel processing environment, shared variables facilitate communication between processes. To protect the shared variables from nondetermination, which can result from concurrent access by more than one process at a time, they are placed within a critical section. In previous work the authors introduced scheduling methods to optimize the time cost of a parallel structure when processes within the parallel structure must access the same critical section. Here, they extend the model and assume that there can be more than one group of parallel processes and that each group has one critical section. Assuming that each critical section is stored in a different memory module, they present methods of optimizing the time cost of the parallel structure under two different scenarios. The first scenario assumes that memory modules exist to accommodate all the groups; the second scenario assumes that these are not enough memory modules to accommodate all the groups. >