Managing Compilation Overheads in a Runtime Specializer for OpenMP

While runtime compilation has in practice been largely restricted to programming languages that execute on virtual machines, such as Java and C#, parallel OpenMP programs show many promising traits for efficient and effective runtime optimization. The stOMP system for the runtime optimization of OpenMP applications is currently under development at the University of Toronto. One of the key issues in making stOMP successful is to minimize the overheads incurred when invoking compilers at runtime. This paper focuses on two techniques that have been developed to manage these overheads: a two-level, context-based hot-spot detector and a pruning mechanism that eliminates poorly behaved variables as specialization targets. A detailed presentation of both techniques is presented along with evaluations of the techniques on a subset of the SPEC OMP benchmark suite. Initial results from a prototype version of the stOMP specializer are also provided.