[vlc-devel] [PATCH 04/25] cpu: add CPU feature masking
Victorien Le Couviour
victorien.lecouviour.tuffet at gmail.com
Tue Apr 14 19:07:59 CEST 2020
On Tue, Apr 14, 2020 at 6:45 PM Rémi Denis-Courmont <remi at remlab.net> wrote:
> Le tiistaina 14. huhtikuuta 2020, 19.23.18 EEST Victorien Le Couviour a
> > On Tue, Apr 14, 2020 at 5:54 PM Rémi Denis-Courmont <remi at remlab.net>
> > > Le tiistaina 14. huhtikuuta 2020, 18.42.56 EEST Victorien Le Couviour a
> > > écrit
> > >
> > > > Hi,
> > > >
> > > > Doesn't that imply I'd have to write a version per feature I want to
> > >
> > > test?
> > >
> > > No but it does imply that you don't kludge the LibVLC run-time to suit
> > > your specific test. Like everybody else who's ever written a test case
> > > here.
> > I don't see how this can work.
> So you can patch the Makefile to include select C file in your benchmark
> but you cannot (think to) patch it to have different CFLAGS and a
> vlc_CPU() implementation?
I thought about it when I did that and rejected the idea,
but maybe it's because I was benching the deinterlacers differently and
needed to link with the core for other purposes.
And so I couldn't have another implementation of vlc_CPU, as there were
already 2 with a weak attribute on one of them, IIRC.
(that was many months ago now, the reflections behind those patches aren't
so fresh in my mind anymore)
That's not the case anymore, and I don't think I need the core for anything
else here anymore anyway, since I changed the way to bench that,
so it could be fine now, I need to actually look into it again.
> That's what I call taking reviewers for fools, and it really pisses me off.
You should think about doing some meditation dude.
It's not normal the get "really pissed", as you say, over such things.
> Rémi Denis-Courmont
> Tapiola new town, Uusimaan Republic
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the vlc-devel