How we measure protection overhead
The Torzen team, July 8, 2026
All protection costs CPU time. The question is not whether there is overhead but where it appears and whether users notice it. Here is how we measure it, so you can repeat the measurement on your own program.
Measure functions, not the whole program
The average run time of an application hides the details. We measure every protected function separately: how often it is called in a typical scenario and how long one call takes before and after protection. The processing log of each job in the panel lists the protected functions, which you can line up with your profiler.
Look at percentiles
For rarely called functions, such as a license check, even a large slowdown goes unnoticed. For functions in the main loop, what matters is the 95th and 99th percentile of frame or response time.
Match the level to the function
- Functions called once or a few times: the highest protection level.
- Frequently called functions: mutation instead of virtualization.
- Hot loops: usually no code protection, but protect the data they use.
Per-function settings are described on the protection settings page.