Radeon R100/R200 Mesa Driver Sees Attention

[phoronix]

Phoronix: Radeon R100/R200 Mesa Driver Sees Attention

David Airlie has made a number of commits to Mesa today that make some modest enhancements to the vintage Radeon R100/R200 classic OpenGL driver...

http://www.phoronix.com/vr.php?view=MTAyMzU

[cb88]

Cool! I'll try this out on my Lifebook P1120 when I get a chance

Whats so bad about old harddware? I rather like my little tablet quite alot and it has EXCELLENT battery life unlike my AMD heater laptop. Its not fast on the internet or anything but that said its excellent for a small portable teminal with wifi and such... just cause it's old doesn't mean it isn't good.

The P1120 has an R100 limited vram and a slow graphics bus but that said its still pretty neat. Also I have a radeon 9800SE/LE in my Tyan 2xPII@300Mhz its rather fun to boot up once in awhile as well.. just cause it's old doesn't mean someone isn't using it. The Tyan is actually rather usable even runs compiz quite nicely browsing is a tad sluggish though been waiting around awhile to see if MESA ever becomes less CPU bound as thats a major hindrance @300Mhz

[whitecat]

Nice to see improvements on these old GPU too!
At work there is still many computers with Radeon 7000 card (Pentium 4 ~3000 MHz, 512 Mo). It's sufficient for Windows XP + office suite. There will be converted in Linux thinclient by time.

[Hamish Wilson]

My brothers laptop is still running a Radeon 9200 in it as is one of our desktop machines and an old machine we are currently using with a capture card to function as a television. So I am for one happy to see theses guys still being worked on, and I still have many fond memories of my own experiences with these cards on my own desktop from about three-four years ago.

This is not a total surprise though, as it was mentioned that these cards could receive some love since one of the things that was slowing down progress was the fact they still had to support the DRI1 infrastructure. That was of course dropped a couple months ago and now they can add certain features to the KMS drivers such as colour titling. Still nice to see it actually happening though.

[PsynoKhi0]

Neat, I really appreciate that developers still devote time to those series.

[dungeon]

That is Blender 2.61, now (after these commits) works really slooooow and also menus in many games (etracer, pingus...)

But yes driver was broken earlier from rewrite (dri2 mode).

,etc...

[dungeon]

Oh these commits afects xbmc It use GL_TEXTURE_RECTANGLE_ARB on r200, so going thorough menu is very choppy and also video playback.

Please try to fix that airlie, it was been good before.

[agd5f]

Oh these commits afects xbmc It use GL_TEXTURE_RECTANGLE_ARB on r200, so going thorough menu is very choppy and also video playback.

Please try to fix that airlie, it was been good before.

Please file a bug:
https://bugs.freedesktop.org

[dungeon]

https://bugs.freedesktop.org/show_bug.cgi?id=44499

color tilling is not implemented for r200/kms, but rendering is a lot faster with enabled - hmmm.

mareko posted samo env LIBGL_SHOW_FPS, interesting

[dungeon]

I tried to fix graphics in Rayman 2 which set manual fog coordinates and reach to demos/fogcoord which also doesn't work... and succeed piglit/ext_coord is good now...
probable radeon needs this fix too.

diff a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
--- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c
+++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
@@ -79,11 +79,11 @@ static void r200_emit_vecfog(struct gl_context *ctx, struct radeon_aos *aos,
 	int size = 1;
 
 	if (stride == 0) {
-		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, count * 4, 32);
 		count = 1;
 		aos->stride = 0;
 	} else {
-		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32);
+		radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, count * 4, 32);
 		aos->stride = size;
 	}

For lighting problems i hit https://bugs.freedesktop.org/show_bug.cgi?id=22576 and related 25883, 26809, 27476, 27824, 38454, 39285... and dont't know where to look for these.
Is this problem in kernel side reg/emit?