COBRA: An Adaptive Runtime Binary Optimization Framework for Multithreaded Applications

This paper presents COBRA (Continuous Binary Re-Adaptation), a runtime binary optimization framework, for multithreaded applications. It is currently implemented on Itanium 2 based SMP and cc-NUMA systems. Using OpenMP NAS parallel benchmark, we show how COBRA can adaptively choose appropriate optimizations according to observed changing runtime program behavior. Coher-ent cache misses caused by true/false data sharing often limit the scalability of multithreaded applications. This pa-per shows that COBRA can significantly improve the per-formance of some applications parallelized with OpenMP, by reducing the aggressiveness of data prefetching and by using exclusive hints for prefetch instructions. For exam-ple, we show that COBRA can improve the performance of OpenMP NAS parallel benchmarks up to 68%, with an av-erage of 17.5 % on the SGI Altix cc-NUMA system.