* Mel Gorman <mgor...@techsingularity.net> wrote: > - ttwu_stat(p, smp_processor_id(), 0); > + if (schedstat_enabled()) > + ttwu_stat(p, smp_processor_id(), 0);
Yeah. I actually like this for the cleanup factor as well: your patch makes it a _lot_ more obvious what is a statistics slowpath and what is not. We tried to achieve that effect via the _stat() naming, but in hindsight that was a failure. Thanks, Ingo