Hi Frank On Wed, Mar 2, 2022 at 3:50 AM Frank Hofmann <fhofmann@xxxxxxxxxxxxxx> wrote: [...] > > Hi Shakeel, > > maybe it's just me, but the codechange below confuses me. I'll comment inline. > [...] > > I understand there is a memory saving from the different table > types/sizes, while the vm events per memcg is a (small) subset of vm > events. How relevant is this to the change, though? > Actually this is the whole reason behind this change i.e. reduce the size of events[_prev] arrays from ~100 to 15. I suspect updating events arrays might be the half the time the rstat flushing is spending on.
