On 18/06/24(Tue) 08:52, Stuart Henderson wrote:
> On 2024/06/17 22:11, Renato Aguiar wrote:
> > 
> > On Sat, Jun 15 2024, Kirill A. Korinsky wrote:
> > 
> > >
> > > This work is based on Solene's patch, which she announced at
> > > https://marc.info/?l=openbsd-tech&m=163259444331471&w=2 , which she
> > > benchmarked, and the results of which she published on her blog
> > > https://dataswamp.org/~solene/2021-09-26-openbsd-power-usage.html .
> > >
> > > With two changes:
> > >
> > > 1. Unused CPU cores are halted and moved to deepest C-state to save some
> > >    power.
> > > 2. It uses powersavelimit settings instead of 100 as active CPU performance.
> > >
> > 
> > I think a patch to just disable specific cpu cores from userspace,
> > e.g. via sysctl, is more likely to be accepted.
> 
> Is that really a good thing to make available to userland?

Not it's not.

> > If there were such
> > mechanism, you could implement your power saving policy in userspace,
> > like obsdfreqd does.
> 
> ugh

That's completely opposed to what has been discussed in the previous
mails.  Instead of having another piece of software with its own
heuristic that do clever tricks to reduce power consumption we want this
logic to be integrated in the scheduler.