On Wed, Apr 06, 2016 at 09:27:24AM +0200, Mike Galbraith wrote: > > On Tue, 2016-04-05 at 14:08 -0400, Chris Mason wrote: > > > Now, on to the patch. I pushed some code around and narrowed the > > problem down to select_idle_sibling() We have cores going into and out > > of idle fast enough that even this cut our latencies in half: > > Are you using NO_HZ? If so, you may want to try the attached.
[ nohz throttling patch ] I tested the nohz throttle two different ways, first with schbench's pipe simulation, it's easily 8% faster with messages bouncing between cpus. In production it's hard to pick a single number because the benchmarks produce latency curves as the workload scales up in RPS. The benefits range from 2-9% depending on the metric. It's a nice win, and I'd love to see it go in. -chris
