Messages in this thread | | | Date | Wed, 30 Oct 2019 12:11:01 +0000 | From | Quentin Perret <> | Subject | Re: [PATCH v2] sched: rt: Make RT capacity aware |
| |
On Wednesday 30 Oct 2019 at 09:04:44 (+0100), Vincent Guittot wrote: > That's the point of my comment, choosing big cores as default and > always best choice is far from being obvious. > And this patch changes the default behavior without study of the > impact apart from stating that this should be ok
The current behaviour is totally bogus on big.LITTLE TBH, and nobody uses it as-is. Vendors hack RT a lot for that exact reason.
Right now a RT task gets a random CPU, which on big.LITTLE is effectively equivalent to selecting a random OPP on SMP. And since RT is all about predicatbility, I'd argue sticking to the big CPUs for consistency with the frequency selection policy is the only sensible default.
So +1 from me for Qais' patch :)
Thanks, Quentin
| |