On Tue, 2019-06-25 at 18:32 -0400, Steven Rostedt wrote: > On Wed, 19 Jun 2019 13:52:14 +0200 > Johannes Berg <johannes@xxxxxxxxxxxxxxxx> wrote: > > > Seems to work, more or less. I now see ~6.5 seconds which is slower than > > it was, and in particular comparing to just a revert (which was ~2 > > seconds), but it's usable :-) > > I'm totally confused. Unless it is a cache miss thing, the only thing > it did was add two branches that should basically follow the old path > with the revert. > > Can you run perf comparing this patch against the revert and see where > the difference lies? I tried, but it looks the same. Then I tried to reproduce and now I see it also fairly consistently just under 2 seconds with the patch, so perhaps CPU scaling or something was throwing me a curve ball before. Sorry for the noise, patch works great. Tested-by: Johannes Berg <johannes@xxxxxxxxxxxxxxxx> johannes
|