nv 2.1.x: why no acceleration with RandR rotation enabled and screen not rotated?

[Alex Deucher]

On 9/25/07, Giuseppe Bilotta <giuseppe.bilotta at gmail.com> wrote:
> On 9/25/07, Alex Deucher <alexdeucher at gmail.com> wrote:
> > On 9/25/07, Giuseppe Bilotta <giuseppe.bilotta at gmail.com> wrote:
> > > Hello all,
> > >
> > > I just dist-upgraded my Debian unstable and got the new X.org,
> > > including nv 2.1.3. Browsing the man page I found out that it
> > > was possible to enable RandR rotation with an appropriate option.
> > > As documented in the man page, enabling this disables acceleration
> > > (and this is indeed quite perceptible).
> > >
> > > So I was wondering if there are technical reasons for the
> > > acceleration to be disabled by just enabling the option, even
> > > when the screen is not actually rotated. I don't know the
> > > implementation details, but I would imagine it to be possible to
> > > select the accelerated path when the screen isn't rotated, and
> > > only fall back to unaccelerated when the screen actually gets
> > > rotated. Or is it just not possible with the current RnR and
> > > acceleration? Is there something I can read on the matter?
> >
> > Acceleration in rotated modes requires exa composite support which is
> > not available in nv at the moment.
>
> Thanks, this explains why I shouldn't get accelerated output when the
> screen is rotated, but why does nv disable acceleration tout court?
> couldn't it just disable it just when the screen is actually rotated?

It should just work; does on other drivers.  You'd have to ask Aaron
why nv works that way.

Alex