Analysis of the execution time unpredictability caused by dynamic branch prediction
Citations Over TimeTop 10% of 2004 papers
Abstract
This paper investigates how dynamic branch prediction in a microprocessor affects the predictability of execution time for software running on that processor. By means of experiments on a number of real processors employing various forms of branch prediction, we evaluate the impact of branch predictors on execution time predictability. The results indicate that dynamic branch predictors give a high and hard-to-predict variation in the execution time of even very simple loops, and that the execution time effects of branch mispredictions can be very large relative to the execution time of regular instructions. We have observed some cases where executing more iterations of a loop actually take less time than executing fewer iterations, due to the effect of dynamic branch predictors. We conclude that current dynamic branch predictions schemes are not suitable for use in real-time systems where execution time predictability is desired.
Related Papers
- → Experimental evaluation of code properties for WCET analysis(2004)45 cited
- → Modeling Control Speculation for Timing Analysis(2004)36 cited
- → Using Predicated Execution to Improve the Performance of a Dynamically Scheduled Machine with Speculative Execution(1996)22 cited
- → A trace based evaluation of speculative branch decoupling(2002)5 cited
- Difficulties in Computing the WCET for Processors with Speculative Execution(2002)