On Thu, 2010-11-11 at 14:54 -0800, Keith Packard wrote:
> On Thu, 11 Nov 2010 10:38:57 -0500, Adam Jackson <[email protected]> wrote:
> > The 24bpp code is very poorly tested, and virtually no modern hardware
> > implements it.  If you need a 24bpp framebuffer, use shadowfb and convert
> > in the upload hook.
> 
> I tested the 24bpp fb code when I wrote it; I guess it has rotted away
> though? There is still a huge pile of hardware which supports 24bpp, and
> some which doesn't support 32bpp well (older mga). Seems fairly harsh to
> remove this from the server.

The bugs I've seen seem like the kind of thing that can't ever have
worked (24bpp surfaces turning into a8r8g8b8 pictures, that kind of
thing).  I can dig up bz references if you like.

>From a quick scan, the following chips support 24bpp but not 32bpp:

chips (some non-HiQV variants)
cirrus alpine except 5480 and 7548
vesa (if no 32bpp modes advertised in vbios)
vmware (if the host is 24bpp)

Everything else appears to have 32bpp support (or doesn't support depth
24 period), although admittedly on older kit you may be running into
memory limits.  I still think that in any environment using Render
you're going to be better off with a blitdown in shadowfb.

Meh.  If we don't do this I'll just fix building 24bpp support out.  I'm
perfectly willing to say that 24bpp-only chips are too short to ride.

- ajax

_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel

Reply via email to