Re: [PATCH sched_ext/for-6.14-fixes 1/2] sched_ext: Implement auto local dispatching of migration disabled tasks

[Andrea Righi <arighi@nvidia.com>]

On Fri, Feb 07, 2025 at 12:44:07PM -1000, Tejun Heo wrote:
...
> > I'm also a bit conflicted if the default should be on or off, we're
> > changing the previous behavior, but OTOH this is going to prevent some
> > potential breakage (due to the nr_cpus_allowed mismatch) and server
> > workload is going to benefit from this, so it seems that there are more
> > pros than cons at dispatching migration_disabled tasks directly by default.
> 
> Hmm.. I didn't see a lot of migrate_disable() while testing with stress-ng
> and migrate_disable() isn't used that much in the kernel to begin with. Do
> you happen to know where migrate_disable() was coming from when you saw them
> with bpfland?

I also don't see a lot of migrate_disable() invocations, but if I do the
usual benchmark, measuring the fps while building the kernel, with the
migration_disabled tasks dispatched directly I see a significant
performance drop in the fps and the kernel build is faster, so it seems
something related to the IO? I'll investigate a bit...

-Andrea