It also sounds like it doesn’t query the power states/frequencies and instead simply assumes a recently used core will be more efficient to use than idle one?
Is that correct and if so, why? Is it slow or inefficient to query power states at regular intervals? Or would it require more bespoke per-cpu code that you don’t want in a scheduler? That type of knowledge seems it’s still inevitable to schedule between big/little cores anyway?
Is that correct and if so, why? Is it slow or inefficient to query power states at regular intervals? Or would it require more bespoke per-cpu code that you don’t want in a scheduler? That type of knowledge seems it’s still inevitable to schedule between big/little cores anyway?