[libdv-dev] Re: Should dv1394 work on PPC?

[Bill F. <bil...@mi...>]

I verified that your patch is a good workaround for non-DMA/non-DRI ATI
card, but also that it then messes up DMA/DRI enabled ATI card (as expected).
Also, the audio is still messed up (with May 17 libdv CVS), producing the
following output:

% playdv --no-mmap /dev/ieee1394/dv/host0/NTSC/in
libdv(audio.c):  Malformrmed AAUX AS? pc4.qu == 0
format 4:3
Audio is 0.0 kHz, 0 bits quantization, 0 channels, emphasis off
Xv: ATI Rage128 Video Overlay: ports 59 - 59
Xv: grabbed port 59
 Using Xv for display
soundcard doesn't support format
libdv(audio.c):  Malformrmed AAUX AS? pc4.qu == 0
libdv(audio.c):  Malformrmed AAUX AS? pc4.qu == 0
libdv(audio.c):  Malformrmed AAUX AS? pc4.qu == 0
...

Using the playdv from the 0.9 version of libdv works fine (with DMA/DRI
enabled), giving the following output:

% playdv --no-mmap /dev/ieee1394/dv/host0/NTSC/in
format 4:3
Audio is 32.0 kHz, 12 bits quantization, 2 channels, emphasis off
Xv: ATI Rage128 Video Overlay: ports 59 - 59
Xv: grabbed port 59
 Using Xv for display

There are no audio related error messages.  I did a diff between the
0.9 and May 17 CVS versions of libdv/audio.c files, but I didn't see
anything obvious.

						-Bill



On Mon, 17 Jun 2002, Bill Fink wrote:

> On Sun, 16 Jun 2002, Daniel Kobras wrote:
> 
> > On Fri, Jun 14, 2002 at 06:07:45PM -0400, Bill Fink wrote:
> > > I've run into a similar problem with my ATI graphics card.  When I don't
> > > use DRI (and thus no DMA) kino gives weird byte swapped colors, but when
> > > I enable DRI (which also enables DMA) then the colors are fine.
> > 
> > I need the attached patch to libdv (CVS) to convince playdv to produce
> > sane colours on PPC with Xv and SDL target.  GTK is fine as is.  Oh, and
> > I need to tell playdv to clamp luma; pure white appears as black
> > otherwise[1].  That was using a YUY2 (or rather 2YUY on big endian) Xv
> > display format.  I'd guess that your kino first tried the buggy 411
> > format and switched to 420 or plain RGB after you fiddled with the
> > configuration.
> 
> This is actually an ATI driver bug.  The same thing happened with the
> xine 0.9.9 logo which used YUY2 format (newer xine versions use MPEG-2
> format which doesn't exhibit the problem).  Without DMA (no DRI) the
> colors are byte swapped, but with DMA (DRI enabled) the colors are
> correct.  Using an NVIDIA card (which doesn't support DRI) gives correct
> colors.  Doing byte swapping in the application would fix the non-DRI/non-DMA
> ATI case, but would break the DRI/DMA ATI case and the NVIDIA (and presumably
> others) case (I know because I tried doing this for xine).
> 
> > To libdv maintainers:
> > The WORDS_BIGENDIAN macro disappeared at some time.  Same for
> > G_BYTE_ORDER.  I changed the instances to use BYTE_ORDER from glibc's
> > endian.h, and also whacked a works-for-me big-endian patch for the 411
> > code in YUY2.c.  I'm rather ignorant to YUV colour spaces, so I'd be
> > happy if someone more knowledgeable than me looked into the lut issue
> > mentioned above.
> 
> The playdv from the 0.9 libdv seems to work best for me on PPC.  Besides
> being a royal pain trying to build more recent CVS versions, the audio
> is broken on these newer versions, probably because of the endianness
> issues you mentioned.  You didn't mention any audio issues so I'm hoping
> your patch fixes that problem.  I'm also wondering if there's any simple
> way to have playdv use YUV instead of YUY2, since that does work fine
> (that's what xine uses for normal MPEG-2 video such as when playing DVDs).
> 
> > Regards,
> > 
> > Daniel.
> > 
> > [1] Bug in the ylut setup in dv_YUY2_init()?  The lut isn't used on x86
> >     with MMX optimisation...
> 
> 						-Thanks
> 
> 						-Bill
> 
> _______________________________________________________________
> 
> Sponsored by:
> ThinkGeek at http://www.ThinkGeek.com/
> _______________________________________________
> mailing list lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linux1394-devel