[Savage40] Xv on savage 2000

Alex Deucher agd5f at yahoo.com
Wed Feb 9 12:41:04 PST 2005


--- Felix Kühling <fxkuehl at gmx.de> wrote:

> Am Mittwoch, den 09.02.2005, 08:07 -0800 schrieb Alex Deucher:
> > I recently got Xv on savage 2000 working, but it has a couple
> problems:
> > 
> > - The video is half as wide as it should be (squashed horizontally
> by
> > 1/2 exactly) the other half is either garbage or the squashed image
> > repeated
> > - colors are wrong.  Looks like only the UV data is being shown.
> > - horizontal downscaling is funky (may be a hardware limitation)
> > 
> > On the plus side it actually displays something and scaling works.
> > If anyone has any ideas please let me know.  At the moment I'm at a
> > loss.  Stride and offset seemed like the logical places to look,
> but
> > messing with those seem to produce predicatable stride/offset
> related
> > behavior.  Maybe the hardware expects the data in a different order
> or
> > something?
> 
> If colors are wrong it could mean that you're supplying a different
> color format than what the hardware expects. Which color format are
> you
> giving to the hardware. Maybe one can deduce from the resulting
> image,
> what color format the hardware is assuming. I know from some reading
> of
> savage_video.c that there is a planar format where chrominance and
> luminance are stored separately. Maybe it's something like this, like
> the left half is the chrominance and the right half is luminance, or
> the
> other way around.

Yeah, I've tried most non-RGB formats as per setblend2000()  They all
seem to produce roughly the same result.  FWIW, it looks like the older
savages should be using format type 4 rather than 1 for YV12 and I420,
but they seem to work.  I guess the data is reformatted beforehand?  If
you use type 4 you are also supposed to set the addresses and strides
of the different planes (you can see the commented out code).  no
difference either way.

> 
> Anyway, getting the colors right first might help finding the right
> strides and offsets.
> 
> > 
> > patch against xorg cvs:
> > http://www.botchco.com/alex/xorg/s2kXv.diff
> 
> Is this the same patch you sent me some time ago (your description
> sounds like it)?

roughly.  although I have have scaling working better.

Alex

> 
> > 
> > Thanks,
> > 
> > Alex
> > 
> 
> -- 
> | Felix Kühling <fxkuehl at gmx.de>                     http://fxk.de.vu
> |
> | PGP Fingerprint: 6A3C 9566 5B30 DDED 73C3  B152 151C 5CC1 D888 E595
> |
> 
> _______________________________________________
> Savage40 mailing list
> Savage40 at probo.com
> http://www.probo.com/mailman/listinfo/savage40
> 


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


More information about the Savage40 mailing list