Performance of Runtime Optimization on BLAST

Abstract

Optimization of a real world application BLAST is used to demonstrate the limitations of static and profile-guided optimizations and to highlight the potential of runtime optimization systems. We analyze the performance profile of this application to determine performance bottlenecks and evaluate the effect of aggressive compiler optimizations on BLAST. We find that applying common optimizations (e.g. O3) can degrade performance. Profile guided optimizations do not show much improvement across the board, as current implementations do not address critical performance bottlenecks in BLAST. In some cases, these optimizations lower performance significantly due to unexpected secondary effects of aggressive optimizations. We also apply runtime optimization to BLAST using the ADORE framework. ADORE speeds up some queries by as much as 58% using data cache prefetching. Branch mispredictions can also be significant for some input sets. Dynamic optimization techniques to improve branch prediction accuracy are described and examined for the application. We find that the primary limitation to the application of runtime optimization for branch misprediction is the tight coupling between data and dependent branch. With better hardware support for influencing branch prediction, a runtime optimizer may deploy optimizations to reduce branch misprediction stalls.