Static slack-based instrumentation of programs

Real-time embedded programs are time sensitive and, to trace such programs, the instrumentation mechanism must honor the programs' timing constraints. We present a time-aware instrumentation technique that injects program code with slack-based conditional instrumentation. The central idea is to execute instrumentation code only when its execution does not increase the worst-case execution time beyond a program's deadline. This occurs at run-time. Unlike previous efforts, this work allows instrumenting on the path that results in the worst-case execution time of the program. We propose a software, and a hardware method of allowing for slack-based conditional instrumentation. We evaluate and compare these two alternatives using a common benchmark suite for real-time systems. Our results show that, on average, the two proposed methods achieve 57% and 80% instrumentation coverage, respectively, compared to only a 3% coverage by previous work.