Learning Bayesian Networks for Improved Instruction Cache Analysis

Abstract

Abstract-As modern processors can execute instructions at far greater rates than these instructions can be retrieved from main memory, computer systems commonly include caches that speed up access times. While these improve average execution times, they introduce additional complexity in determining the Worst Case Execution Times crucial for Real-Time Systems. In this paper, an approach is presented that utilises Bayesian Networks in order to more accurately estimate the worst-case caching behaviour of programs. With this method, a Bayesian Network is learned from traces of program execution that allows both constructive and destructive dependencies between instructions to be determined and a joint distribution over the number of cache hits to be found. Attention is given to the question of how the accuracy of the network depends on both the number of observations used for learning and the cardinality of the set of potential parents considered by the learning algorithm.

Probabilistic Instruction Cache Analysis Using Bayesian Networks

“…previously published version of the technique [6] is modified to allow it to be used in WCET analysis. The earlier model mixed the prediction of control flow and cache analysis.…”

“…The technique presented in this paper is a modified version of the cache model proposed in [6]. However, as presented, that technique allowed only for the distribution of cache hits or misses on any unknown execution of the program to be determined.…”

“…To make this paper self-contained, and as the technique in [6] has not previously been published in the Real-Time Systems community, we first outline that technique here. The problems with this are then made explicit before the changes necessary are presented.…”