COMMUTATIVITY ANALYSIS - A NEW ANALYSIS FRAMEWORK FOR PARALLELIZING COMPILERS

This paper presents a new analysis technique, commutativity analysis, for automatically parallelizing computations that manipulate dynamic, pointer-based data structures. Commutativity analysis views the comput ation as composed of operations on objects. It then analyzes the progr am at this granularity to discover when operations commute (i.e. gener ate the same final result regardless of the order in which they execut e). If all of the operations required to perform a given computation c ommute, the compiler can automatically generate parallel code. We have implemented a prototype compilation system that uses commutativity an alysis as its primary analysis framework. We have used this system to automatically parallelize two complete scientific computations: the Ba rnes-Hut N-body solver and the Water code. This paper presents perform ance results for the generated parallel code running on the Stanford D ASH machine. These results provide encouraging evidence that commutati vity analysis can serve as the basis for a successful parallelizing co mpiler.