Message from C, C++ talks
How do you test your optimizations guys?
I mean how do you measure the impact of an optimization and see what actually is changed after it?
I was using Valgrind and run 10-20 frames of my game before and after an optimization, but it gives a bit different results each time, also it's hard to compare the results.
Now I'm making synthetic tests that will run some specific logic for specific amount of frames with fixed frame time, so it should be almost the same. But I still will compare two profiler reports by eyes, so I think it's not going to be super effective.
Do you know better ways to compare performance? What way do you use?