Gentle ping, in case you have time to look at this.
On Thu, 21 Nov 2019 at 18:05, Andrei Vagin <avagin@xxxxxxxxx> wrote:
>
> Hi Thomas,
>
> What is your plan on this series? We know you are probably busy with
> the next merge window. We just want to check that this is still in your
> TODO list.
>
> On Tue, Nov 12, 2019 at 01:26:49AM +0000, Dmitry Safonov wrote:
> >
> > v7..v8 Changes:
> > * Fix compile-time errors:
> > - on architectures without the support of time namespaces.
> > - when CONFIG_POSIX_TIMERS isn't set.
> > * Added checks in selftests for CONFIG_POSIX_TIMERS.
> > * Inline do_hres and do_coarse.
> > (And added Tested-by Vincenzo - thanks!)
> > * Make TIME_NS depends on GENERIC_VDSO_TIME_NS and set it per-arch.
> >
> > [v1..v7 Changelogs is at the very bottom here]
> >
> > Our performance measurements show that the price of VDSO's clock_gettime()
> > in a child time namespace is about 8% with a hot CPU cache and about 90%
>
> Here is a typo. The price of VDSO's clock_gettime() in a child time
> namespace is about 12% with a cold CPU cache. The table with
> measurements for a cold CPU cache contains correct data.
>
> > with a cold CPU cache. There is no performance regression for host
> > processes outside time namespace on those tests.
> >
>
> ....
>
> >
> > Cold CPU cache (lesser tsc per cycle - the better):
> >
> > | before | CONFIG_TIME_NS=n | host | inside timens
> > --------------------------------------------------------------
> > tsc | 476 | 480 | 487 | 531
> > stdev(tsc) | 0.6 | 1.3 | 4.3 | 5.7
> > diff (%) | 100 | 100.9 | 102 | 112
> >
Thanks,
Dmitry
