--===============0801668235==
Content-type: multipart/signed; micalg=pgp-sha1;
 protocol="application/pgp-signature"; boundary=nextPart1256941.jRPKhpZhWO
Content-transfer-encoding: 7bit

--nextPart1256941.jRPKhpZhWO
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

Op Wednesday 02 May 2007, schreef Rivo Laks:
> =DChel kenal p=E4eval (esmasp=E4ev 30 aprill 2007) kirjutas Lubos Lunak:
> >  Hello,
> >
> >  does somebody know about details about %subj%? This
> > http://lists.freedesktop.org/archives/compiz/2007-April/001965.html
> > thread discusses some internal NVidia shell variable that should improve
> > performance.
>
> Hi
>
> From what I could find in the net, it tells the driver not to "yield it's
> timeslice" which seems to mean that driver won't donate it's own execution
> time to other programs. This would result in better performance, especial=
ly
> under heavy cpu load.

It sounds like an issue with interactivity, though. As you mention below, t=
he=20
animation goes smoother, but the window doesn't draw anymore. I guess what=
=20
this does is ensure the GL part of the app gets all the CPU it can get it's=
=20
hands on. If the window isn't doing anything else (or you can prevent the=20
need for repaint by keeping the texture around), this would improve things=
=20
visually. And if the GL code takes away from the 'work' the app has to do, =
I=20
don't think that's a bad thing either - as long as you're not dealing with=
=20
things like music and video playback. I can imagine those would get hurt.

I don't know much about the details here, but playing with the scheduler is=
 a=20
dangerous thing. So be a bit carefull, you'd want sound to continue playing=
=20
if one does a desktop switch :D

> Experimenting shows that setting  __GL_YIELD=3D"NOTHING"  really does imp=
rove
> (visible) performance. Especially animations feel a lot snappier with it,
> possibly because window repaints are more delayed towards the end of
> animation. So while the animation (e.g. desktop switch) is in progress, y=
ou
> mostly see just a white/grey window background and usually the window
> repaint is completed only at the end of the animation.
> BTW, can't we just keep the window texture around (window mapped or
> whatever the correct term is) instead of repainting it every time it's
> shown (e.g. on desktop switch)?
>
> So IMHO it would probably be a good idea to use __GL_YIELD when doing GL
> compositing.
>
> Rivo
> _______________________________________________
> Kwin mailing list
> Kwin@kde.org
> https://mail.kde.org/mailman/listinfo/kwin



=2D-=20
Disclaimer:

Alles wat ik doe denk en zeg is gebaseerd op het wereldbeeld wat ik nu heb.=
=20
Ik ben niet verantwoordelijk voor wijzigingen van de wereld, of het beeld w=
at=20
ik daarvan heb, noch voor de daaruit voortvloeiende gedragingen van mezelf.=
=20
Alles wat ik zeg is aardig bedoeld, tenzij expliciet vermeld.

Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html

=A0 =A0A: Because it destroys the flow of the conversation
=A0 =A0Q: Why is top-posting bad?

--nextPart1256941.jRPKhpZhWO
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)

iD8DBQBGOIeV+wgQ1AD35iwRAjSkAJwO6nIUsjpxdZgsw1hal2zZYXV3dACfaO5W
m9tExP+cpARsZgdXOewir/E=
=aeSZ
-----END PGP SIGNATURE-----

--nextPart1256941.jRPKhpZhWO--

--===============0801668235==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Kwin mailing list
Kwin@kde.org
https://mail.kde.org/mailman/listinfo/kwin

--===============0801668235==--