[MPlayer-dev-eng] [PATCH] Radeon 9600 Pro vidix support

[Jonas Jensen]

On Fri, 2003-12-19 at 22:09, Vladimir Mosgalin wrote:
> On Fri, 19 Dec 2003, Jonas Jensen wrote:
> JJ>driver was being used. One solution could be to just always shift the
> JJ>DGA address down by 8MB. I'm sure all Radeon cards have so much memory
> JJ>that it doesn't matter...
> 
> No. Well, if you don't need opengl, that's true, you can sacrifice 8M.
> But the problem is, it doesn't work. I haven't tested it with current
> card & driver, but with my old 8500 and 2.5.1 or 2.9.12 drivers my
> computer hung up hard instantly on launch of mplayer or any opengl app
> when limiting video memory. I believe it worked a very long time ago and
> it works on matrox, but not in this case. Well, it was a dirty solution
> anyway. That patch is a hack too, but at least it works without problems
> or limitations.

I don't see what you mean by limiting video memory. I can't say I fully
understand vidix, but the way I read it it simply tries to place its
buffer at the highest possible address in video memory, because there
will be the least chance of stepping on another application's (X's)
memory.

The patch moves the address down by 8MB, increasing the likehood of
corrupting X's cached pixmaps, but I don't see how that can crash
anything. If that crashes, playing a 1600x1200 4:2:2 movie with double
buffering (which would also take about 8MB) should crash your computer
too, right?

On my laptop's 4MB mach64 card I often step on X's cached pixmaps, or
even X's display itself, but that only causes graphics corruption, not
crashes.

-- 
Jonas Jensen <jbj at knef.dk>