On 12/10/21 18:13, Rémi Bernon wrote:
Can we put a little bit more trust in the tool now? I'm not saying it's always right, and that CPU time wouldn't be spent or wasted by the game otherwise, but here, it's clearly telling us that we are adding some overhead and where it comes from.
Did I ever say that perf spot is always wrong? Your present results confirms the spot here is not random. How do I tell if that is the same in arbitrary case before checking? It seems you are not saying that it is always right as well. And while the perf was and is the first thing I run (like probably most of other people), I don't see how this case change anything in my general trust to the tool or convince me that I should stop checking those spots and stop figuring the model explaining the behaviour (tbh I have no idea how anything like that can be sanely fixed at all without the latter).