> On May 27, 2019, at 1:30 AM, Peter Zijlstra <pet...@infradead.org> wrote: > > On Sat, May 25, 2019 at 01:21:59AM -0700, Nadav Amit wrote: >> cpumask_next() has no side-effects. Mark it as pure. > > It would be good to have a few word on why... because apparently you > found this makes a difference.
I see that eventually it did not make any difference. I saw in the past that some instances of the kernel are affected by it, but I will remove it from this patch-set in the next iteration.