[4.9-RC] What happened to ``options HTT'' ?

[Helge Oldach]

Oliver Fromme:
>Brooks Davis <brooks at one-eyed-alien.net> wrote:
> > On Tue, Oct 28, 2003 at 11:42:57PM +0200, Jaco van Tonder wrote:
> > > I just cvsup'ed to that latest -STABLE and saw that option HTT
> > > is no longer availible in LINT. Does anyone know why this is no
> > > longer supported? I searched the stable@ list but cant seem to
> > > find anything on this.
> > It is no longer necessicary as HTT logical CPUs are now
> > activated by default. If you wish to use them, set the sysctl
> > machdep.hlt_logical_cpus to 0.
>Can that sysctl be changed at runtime, or is it only possible to do it
>in the loader? The UPDATING entry isn't completely clear about that.

I can confirm from experience that changing this sysctl to 0 immediately
started the second logical CPU. Turning it back to 1 apparently stops
processes on the second logical CPU to start, instead everything runs
on CPU0. However processes running on the second CPU remain there until
they terminate.

>By the way, are there any "real world" experiences whether it's worth
>it to use the logical CPUs, say, on a PostgreSQL or Apache server, or
>with Java stuff (jakarta/tomcat etc.)?

That would interest me as well. In particular I would be interested
where improvements might be expected - in CPU-intensive environments,
interrupt-intensive enviroments, and I/O-intensive environments.

I am running a system that walks a couple of thousand files every minute
- would that benefit from HTT? (I seem to recall that UFS as part of the
kernel is restricted to a single CPU?)

Helge