Unless I messed things up again, patch 1 and 2 are for perf/urgent and the rest
can wait.
I would still like to relax the HT scheduling constraint in case there are no
funny events at all, but I've not yet found a nice way to do that.
Peter,
On Thu, May 21, 2015 at 4:17 AM, Peter Zijlstra <[email protected]> wrote:
>
> Unless I messed things up again, patch 1 and 2 are for perf/urgent and the rest
> can wait.
>
> I would still like to relax the HT scheduling constraint in case there are no
> funny events at all, but I've not yet found a nice way to do that.
>
Are you talking about the HT bug workaround? If there is no
corrupting event across HT threads, then scheduling is not impacted.
On Thu, May 21, 2015 at 04:48:16AM -0700, Stephane Eranian wrote:
> Peter,
>
> On Thu, May 21, 2015 at 4:17 AM, Peter Zijlstra <[email protected]> wrote:
> >
> > Unless I messed things up again, patch 1 and 2 are for perf/urgent and the rest
> > can wait.
> >
> > I would still like to relax the HT scheduling constraint in case there are no
> > funny events at all, but I've not yet found a nice way to do that.
> >
> Are you talking about the HT bug workaround? If there is no
> corrupting event across HT threads, then scheduling is not impacted.
It is, commit c02cdbf60b51 ("perf/x86/intel: Limit to half counters when
the HT workaround is enabled, to avoid exclusive mode starvation") is
active irrespective of any active events.