Intel 945GM and XvMC?

[Chad]

Thank you for all the follow ups.

On 9/11/06, Daniel Stone <daniel at fooishbar.org> wrote:
> On Sun, Sep 10, 2006 at 11:26:49PM -0700, galenz at zinkconsulting.com wrote:
> > I have the 915G / GMA 900 chipset. The XvMC option is a bit different
> > than with the older chips, but when XVideo is enabled using the i810
> > drivers, review the startup logs for X and you'll see XVideo and
> > XVideo Motion Compensation load up. As far as I know, the GMA 950 is
> > just a faster, better version of the GMA 900.
>
> When the XVideo-MotionCompensation extension is loaded, that doesn't
> mean XvMC is actually in use: it isn't.
>
> (Confusing, I know.)
>
>

So, is there any way to get it into use?


> On Sun, Sep 10, 2006 at 11:26:49PM -0700, galenz at zinkconsulting.com wrote
> I have the 915G / GMA 900 chipset. The XvMC option is a bit different
> than with the older chips, but when XVideo is enabled using the i810
> drivers, review the startup logs for X and you'll see XVideo and
> XVideo Motion Compensation load up. As far as I know, the GMA 950 is
> just a faster, better version of the GMA 900.

This sounds promising, as well as the below information...

> Based on my tests, using VLC and other video players with XVideo
> output modules, the performance gain is very considerable. Not as
> amazing as some MPEG-2 acceleration I've seen, but 20% CPU use
> watching 720p on a Celeron D 3 GHz CPU isn't too shabby. I'm running
> FC5, VLC 0.85 and everything as x86-64 binaries.

So, with VLC and other players, do you just use the regular xv output,
or do you use some variation of xvmc output?  If it's just xv, does
that mean that something like XvMC exists in firmware in the actual
hardware itself?

> The setting under FC5 had to be changed in order to play HD, you will
> have to set the CacheLines value - the default is too low, it causes
> the players to segfault. 1100 works beautifully for 720p playback.

Thank you, I will probably get a board and start testing.

> However, I'm currently struggling to get the 1080i working properly.
> I have been experimenting with all sorts of things and it still
> causes all the players to segfault and I'm not sure why. I've been
> going back and fourth a bunch today in an earlier post regarding this
> issue - look around the list a bit and you'll see it.

Will do, a quick glance didn't reveal much which is why I posted the
question.  Of course, the usual problem of not knowing what to search
for is probably why.

Thanks again!

> Keith wrote:
> The chip supports it, but the driver code hasn't been written.

So, does that mean someone is going to, or that it doesn't need to be
done (as Galen noted it 'sorta' seems to work already)?  I'm looking
at migrating away from via-pro mini-itx boards because it's lacks a
lot of [working] driver implementation and saw the
intellinuxgraphics.org project and felt as if it were a great
alternative for a media platform.  I've found several boards with a
945GM chipset on that that have everything I think would be necessary
in a single-board type solution.  But I also found that with Via.
Hopefully with Intel seeming/appearing to do things very well from the
get-go, it won't turn into another Via situation where they promise
the world from their 'specs' but deliver a little less than par.

In a nutshell I guess:  will the driver code be written?

Thanks again everyone!

-Chad