R500 Mesa Is Still No Match To An Old Catalyst Driver

[phoronix]

Phoronix: R500 Mesa Is Still No Match To An Old Catalyst Driver

We are in the process of conducting a set of tests looking at how the performance of Ubuntu Linux has evolved through their Long-Term Support (LTS) releases beginning with their first 6.06 "Dapper Drake" version followed by Ubuntu 8.04 "Hardy Heron", and then the Ubuntu 10.04 "Lucid Lynx" release that will be released by month's end. These benchmarks will look at how the performance of Ubuntu Linux has changed over the past four years, but first we deviated from our original plans to get a look at how the current open-source ATI R500 graphics driver in Ubuntu 10.04 provided by the Mesa stack performs against older proprietary ATI Catalyst drivers.

http://www.phoronix.com/vr.php?view=14749

[whizse]

It means that the shaders work with the Mesa GLSL parser/compiler.

It's no guarantee that individual Mesa drivers won't have bugs, but it's at least an indication that the shaders are okay and the problem you were having could be a bug in fglrx.

[Kano]

It works with that override, 2 fps only however. And what do you learn from that?

[whizse]

I can only test ATI 4550, i doubt R700 has got GLSL, i do NOT own a R500 card anyway, why should i test it?

As I mentioned previously, LIBGL_ALWAYS_SOFTWARE=1. Yes, it's dead slow, but you'll notice if the GLSL compiler barfs on it or not.

[BlackStar]

I can only test ATI 4550, i doubt R700 has got GLSL, i do NOT own a R500 card anyway, why should i test it?

Because that would help pinpoint the bug. Is it in XBMC or in fglrx?

R700 GLSL is listed as "mostly" here, so I suspect it should be working more or less. (IIRC, the Mesa is missing flow control for GLSL 1.2 but that's it).

[Kano]

I can only test ATI 4550, i doubt R700 has got GLSL, i do NOT own a R500 card anyway, why should i test it?

[nanonyme]

I only test fglrx, mesa does not interest me. Especially as you can test now also vaapi.

I suspect you missed the point. Both Mesa and fglrx have stricter compilers than nVidia. But if Mesa works and fglrx doesn't, it's very likely to be a strictly fglrx bug, not a program bug.

[Kano]

I only test fglrx, mesa does not interest me. Especially as you can test now also vaapi.

[BlackStar]

@BlackStar

Did you send a patch to the xbmc devs? Then i could test it.

Not feeling bored enough to debug (yet another) random app right now. I'll let you know, though.

A slightly more practical question: does XBMC OpenGL/GLSL output work with Mesa?

[bridgman]

The two active ledgers right now are the RadeonFeature wiki page ( http://www.x.org/wiki/RadeonFeature ) and the OpenGL 3.0 task list maintained by Brian at VMWare ( http://cgit.freedesktop.org/mesa/mesa/tree/docs/GL3.txt )

We should probably also update this one to include r6xx/7xx/EG :

http://dri.freedesktop.org/wiki/MissingFunctionality

re: distros, the open source devs tend to run whatever distro is closest to the new stuff, currently a toss-up between Lucid and F13.

I`m running Lucid right now but will probably switch to F13 for the graphics test day (Wed) and stay there for a bit. Hardware is rv570 (X1950PRO), rv620 (HD3450) and an engineering sample of HD5870 on a Phenom X4. HD5670 at home.

The proprietary driver devs tend to have a dozen or so distro images active and rotate between them depending on the specific bug fix or feature they are working on. We run a mix of PC platforms but try to match the distribution of CPUs and chipsets found in the market over the last 6 months.

We also have a small test team building and testing the latest open source driver code on a few different OSes, currently running on a 2 week cycle. I`ll dig up the last test report from them and get you the distro version and HW details as well.