Causal Profiling

I came across this today, and thought it would be interesting to try it on Urho. The basic idea is that current profilers don’t produce output that always correlates well with where you should target optimizations, and this one attempts to do a better job of it. It seeks to give you an insights about “If optimized X this much, performance would increase by Y”.

If someone else wants to try it or has tried it, feel free, as I may not get to it soon.