On Wed, Sep 02, 2009 at 02:54:44PM -0500, Ian Harrold wrote:
> I saw that version 1.4.2 included Marvel PXA3xx 2D graphics support and
> wanted to give it a try on my Marvel PXA310 processor.
> 
> I cross compiled the library and ran the "pxa3xx_dump" application.   I get
> the following error:
> 
> (!) Direct/Util: opening '/dev/pxa3xx_gfx' and '/dev/misc/pxa3xx_gfx' failed
>     --> No such file or directory
> 
> My /dev/fb is a link to /dev/fb0
> and according to the device directory, I have:
> /dev/fb0
> /dev/fb1
> /dev/fb2
> /dev/fb3
> /dev/fb4
> /dev/fb5
> /dev/fb6
> /dev/fb7
> 
> So what do I need to get this working?

First of all, you need to have the kernel part of that driver which is
still pending for merge. Don't know exactly why, but I will trigger it
again.

See this post:

  http://lists.arm.linux.org.uk/lurker/message/20090709.170451.47d52749.en.html

And unless you apply this to a cutting edge kernel checked out from git
repository, you might also need the other 3 patches of that series.

Also pay attention to the comment at the beginning of the driver:

/* 
 * WARNING: This controller is attached to System Bus 2 of the PXA which 
 * needs its arbiter to be enabled explictly (CKENB & 1<<9). 
 * There is currently no way to do this from Linux, so you need to teach 
 * your bootloader for now. 
 */ 

> would a "mknod -m600 /dev/pxa3xx_gfx c 10 151" or similar suffice?

Once the driver is loaded, /proc/misc will tell you which misc device is
associated to it. That number is should be the last parameter of the
command you mentioned.

> I'm running Linux Kernal 2.6.21 and busy box 1.0.1 on a CRAMFS main file
> system.   But I can recompile the kernel with these changes.

Hmm, not sure whether it's easily possible to backport the driver to
such an ancient kernel. I would consider updating rather than wasting
time on this.

> I was just
> hopeing someone might have some input rather me shooting blindly trying to
> get this working.   I am curious how much faster my application will run
> with the 2D acceleration versus the software implementation it has now.

Yes, it would be good to have more testers for this driver. There is
still one beast of a bug pending which we're currently trying to sort
out. Maybe your experiences can help us finding out.

Thanks,
Daniel
_______________________________________________
directfb-users mailing list
[email protected]
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users

Reply via email to