GMA500 support in Android-x86

1,573 views
Skip to first unread message

Kjaergaard

unread,
Oct 24, 2010, 4:57:08 AM10/24/10
to Android-x86
As I understand it the GMA500 used with the Atom Poulsbo chipset isn't
currently supported by Android-x86 except by running the graphics in
Vesa mode. Correct ?

Does anyone have an idea what kind of effort would be needed to
support the GMA500 ? What does the implementation of the OpenGL
acceleration means to this ?

Best regards
Jens Kristian

Doug Schaefer

unread,
Oct 24, 2010, 12:32:12 PM10/24/10
to andro...@googlegroups.com
Wikipedia actually has a good description of the issue here:

The problem is that the GMA 500 is actually a PowerVR core, the same one you see in TI's OMAP chips and such. Intel has been very good at producing open source drivers for their graphics cores. PowerVR, not so much.


--
You received this message because you are subscribed to the Google Groups "Android-x86" group.
To post to this group, send email to andro...@googlegroups.com.
To unsubscribe from this group, send email to android-x86...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/android-x86?hl=en.


okwon

unread,
Oct 24, 2010, 8:15:46 PM10/24/10
to Android-x86
GMA500 2D is supported. At least the native resolution of the LCD
panel is supported from the driver. You may refer to Viliv S5 target
in Android-x86 and following driver.

http://vilivs5.git.sourceforge.net/git/gitweb.cgi?p=vilivs5/psb-android.git;a=summary
http://android-x86.git.sourceforge.net/git/gitweb.cgi?p=android-x86/x86_hardware_menlow_psb.git;a=summary

Thanks,
Owen
> > android-x86...@googlegroups.com<android-x86%2Bunsubscribe@googlegr oups.com>
> > .

Kjaergaard

unread,
Nov 28, 2010, 4:50:34 PM11/28/10
to Android-x86
I have now tried the Viliv s5 target on my GMA500 based notebook, and
I have got it working for donut. However I would like GMA500 for
Froyo.

If I try to build the viliv s5 for Froyo the compilation of the
graphics driver fails, because of different kernel releases.

I have tried to install a fresh release of psb-kernel-source-4.42.0
and have also been able to build an image, but even when running in
vesa mode there is obvious problems. Are there any tricks or
configuration that must be applied to the psb drivers to run properly
under Android?

Have anyone else been trying and had success with running the psb
drivers with Android ?

Regards
Jens Kristian


On Oct 25, 1:15 am, okwon <pine...@gmail.com> wrote:
> GMA5002D is supported. At least the native resolution of the LCD
> panel is supported from the driver. You may refer to Viliv S5 target
> in Android-x86 and following driver.
>
> http://vilivs5.git.sourceforge.net/git/gitweb.cgi?p=vilivs5/psb-andro...http://android-x86.git.sourceforge.net/git/gitweb.cgi?p=android-x86/x...
>
> Thanks,
> Owen
>
> On Oct 25, 1:32 am, Doug Schaefer <cdtd...@gmail.com> wrote:
>
>
>
>
>
>
>
> > Wikipedia actually has a good description of the issue here:
> >    http://en.wikipedia.org/wiki/Intel_GMA#GMA_500_on_Linux
>
> > <http://en.wikipedia.org/wiki/Intel_GMA#GMA_500_on_Linux>The problem is that
> > the GMA 500 is actually a PowerVR core, the same one you see in TI's OMAP
> > chips and such. Intel has been very good at producing open source drivers
> > for their graphics cores. PowerVR, not so much.
>
> > On Sun, Oct 24, 2010 at 4:57 AM, Kjaergaard <ted...@mail.dk> wrote:
> > > As I understand it theGMA500used with the Atom Poulsbo chipset isn't
> > > currently supported by Android-x86 except by running the graphics in
> > > Vesa mode. Correct ?
>
> > > Does anyone have an idea what kind of effort would be needed to
> > > support theGMA500? What does the implementation of the OpenGL

okwon

unread,
Nov 28, 2010, 7:51:59 PM11/28/10
to Android-x86
Please try following repository. The source is for 2.6.32 kernel, but
it worked with 2.6.35 kernel as far as I remember.
http://vilivs5.git.sourceforge.net/git/gitweb.cgi?p=vilivs5/psb-android.git;a=shortlog;h=refs/heads/android-2.6.32

Or you may try with following 2.6.32 kernel.
http://vilivs5.git.sourceforge.net/git/gitweb.cgi?p=vilivs5/common_kernel.git;a=shortlog;h=refs/heads/android-2.6.32

Another way would be to using following S5 froyo-x86 repo as a base. I
have to merge this to android-x86 repos, but there is kernel
difference (32 vs 35), and I don't have much time now.
repo init -u git://vilivs5.git.sourceforge.net/gitroot/vilivs5/manifest.git
-b froyo-x86

Thanks,
Owen

On Nov 29, 6:50 am, Kjaergaard <ted...@mail.dk> wrote:
> I have now tried the Viliv s5 target on my GMA500 based notebook, and
> I have got it working for donut. However I would like GMA500 for
> Froyo.
>
> If I try to build the viliv s5 for Froyo the compilation of the
> graphics driver fails, because of different kernel releases.
>
> I have tried to install a fresh release of psb-kernel-source-4.42.0
> and have also been able to build an image, but even when running in
> vesa mode there is obvious problems. Are there any tricks or
> configuration that must be applied to the psb drivers to run properly
> under Android?
>
> Have anyone else been trying and had success with running the psb
> drivers with Android ?
>
> Regards
> Jens Kristian
>
> On Oct 25, 1:15 am, okwon <pine...@gmail.com> wrote:
>
>
>
>
>
>
>
> > GMA5002D is supported. At least the native resolution of the LCD
> > panel is supported from the driver. You may refer to Viliv S5 target
> > in Android-x86 and following driver.
>
> >http://vilivs5.git.sourceforge.net/git/gitweb.cgi?p=vilivs5/psb-andro......

Kjaergaard

unread,
Nov 29, 2010, 3:38:18 PM11/29/10
to Android-x86
Thanks for your kind advice.

I have tried to install and build from the vilivs5 repository you
mention. I had to patch 2 application files presumably due to some API
mismatch, but this shouldn't have any impact on getting the graphics
working.

I got the bootanimation screen to show up, which I interpret as the
graphics must be working to some degree. However it stops here, and
the last records on dmesg are:
[drm] fault start
this is repeated approx 100 times.

My best guess is that this is caused by some mismatch in the kernel
configuration.

So I now have this 2.6.32 kernel configuration derived from s5 that
gives me some graphics, but which stops just before the final launch
of Android, and an eeepc derived kernel configuration for 2.6.35 for
which the graphics is broken.

I will try to merge the two tracks in some way.

Thanks for help so far
Jens Kristian

On Nov 29, 1:51 am, okwon <pine...@gmail.com> wrote:
> Please try following repository. The source is for 2.6.32 kernel, but
> it worked with 2.6.35 kernel as far as I remember.http://vilivs5.git.sourceforge.net/git/gitweb.cgi?p=vilivs5/psb-andro...
>
> Or you may try with following 2.6.32 kernel.http://vilivs5.git.sourceforge.net/git/gitweb.cgi?p=vilivs5/common_ke...

Kjaergaard

unread,
Dec 6, 2010, 5:22:52 PM12/6/10
to Android-x86
I have now managed to get the 2G graphics running with the Poulsbo psb
drivers. However the graphics performance is pretty low. I have
measured the performance with 0xbenchmark and compared the performance
to an eeepc 1005HA with the i915c driver and also with an HTC Desire.
The comparison is here:
https://docs.google.com/viewer?a=v&pid=explorer&chrome=true&srcid=0BzjunbRgzF6WNWI3MGMzMzAtYzU2YS00Mjk2LWI0Y2UtNTNjNmYwOGEyMjll&hl=en&authkey=CICF0qEP

It shows that the eeepc runs about 10 times as fast as the Poulsbo
chip. I know that considerable effort has been put into optimizing the
915 drivers, but is the gain really that substantial.

Is it be possible to reuse the optimizations strategy and implement
similar optimization in the PSB drivers ?

Regards
Jens Kristian
Reply all
Reply to author
Forward
0 new messages