Abstract
The authors present the issues in concurrent program execution, and survey some of the current research in this area. Programs for multiprocessor systems require a completely different approach from those for uniprocessor systems. The effective use of multiprocessor systems requires a program representation which clearly indicates sections (tasks) that can be executed concurrently, mechanisms for synchronizing these different concurrent modules, methods for ensuring data integrity, and algorithms that schedule and allocate these modules to the available processors efficiently. Most current work in these areas make very limiting assumptions about the nature of the program, or the nature of the underlying hardware architecture, so that they are not truly flexible.<>