Optimizing atomic functions using compile-time information

Functions that invoke operations on multiple objects atomically are a useful extension of object-based parallel languages, such as Orca. This paper introduces atomic functions and shows how compile-time information can drive run-time optimizations of such functions.