Is there some way to estimate the overhead of dynamic instrumentation? It certainly depends on what you actually do with injected code and how often you do inject code.

But I guess there should be at least some kind of lower limit, i.e. you hook a function and do the smallest amount of code (like just jumping back?) to keep the program running.

Would this scale linearly with the number of functions? To what extent does overhead relate to the actual tool you use?

Are there ways to estimate the overhead when injected code gets more complex?

  • I worked with intel's pin, which another "dynamic instrumentation tool". In my experience there is a constant startup overhead (translation) added by this tool and then overhead gets increased further based on your instrumentation points.
    – R4444
    Nov 15, 2021 at 23:17


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.