R6xx/R7XX kernel 2.6.33 module performance hacks

[pheldens]

note that ut2004 is difficult to benchmark, since demos involving bots are randomized, the scenes are different all the time giving differing averages +/- 50%

its however one of the few perfect working games for me on 4670
1280x960 all detail i get about 55fps average is some arbitrary demos

[Obscene_CNN]

Yes both me and my computer x11perf *is* giving better results

Could it be possible your not using KMS after all?

Regardless, I am glad you are experiencing improved performance with my patch and are happy with it.

The proof that it works is in the pudding and all that people need to do is try it or someone to benchmark it with KMS and without. I can say with out a doubt that without KMS any game where it bogs down your video card the improvement is quite apparent.

Until I can boot to a command prompt with KMS I won't be using it. I know quite a few others that feel the same. As long as compute intensive and memory intensive tasks run faster out side of X we will always startx by hand.

[xming]

Could it be possible your not using KMS after all?

Nope, I am 100% sure I am using KMS, I boot every time into 80x25 then modeprobe manually radeon modeset=1 agpmode=8 then see the console swithcing to 1600x1050 with a working frame buffer console, after that I start X manually.

[pheldens]

yes kms console works 100% here too, it one of the main benefits of kms

[Vash63]

Could it be possible your not using KMS after all?

Regardless, I am glad you are experiencing improved performance with my patch and are happy with it.

The proof that it works is in the pudding and all that people need to do is try it or someone to benchmark it with KMS and without. I can say with out a doubt that without KMS any game where it bogs down your video card the improvement is quite apparent.

Until I can boot to a command prompt with KMS I won't be using it. I know quite a few others that feel the same. As long as compute intensive and memory intensive tasks run faster out side of X we will always startx by hand.

Uh... you can boot to a command prompt with KMS. That's the only way I've ever done it since I started using it months ago on my Gentoo/Exherbo systems.

[Obscene_CNN]

After looking at some of the KMS code I have found my hacks will work there too. I have also found some other minor improvements. I will be submitting patches for those here some time this week.

[Obscene_CNN]

Updated r6xx/r7xx kernel hacks for 2.6.33 now with preliminary kms blit goodness. Back out my previous patches before applying.


http://pastebin.ca/1749087
http://pastebin.ca/1749086
http://pastebin.ca/1749084

Note: I do not know if the kms portion works at all yet as I have yet to get kms to working on my machine. The non KMS patches are a tad bit faster now and work great.

Update: I have had two success storys with the kms patch, modest gains but it works. I'll be making it better

[xming]

tested your new set of patches, work great. I had previously performance problem with glxgears, but found the problem, it was caused by kwin desktop effect, apparently that ate so much performance compared to compiz.

Now w/o your patches glxgears runs at 2000~2100 fps, with it it runs at 2300~2370. Biggest improvement comes from ut2004, unplayable framerate to somehow more playable, I need to play more maps to judge the difference.

I have produced a bz2 which contains the 3 patchs in one, a bit easier to download and apply (cleanly to 2.6.32-drm-radeon-testing) here. Obscene_CNN, I hope you don't mind.

[darkbasic]

Nexuiz timedemo - demo1
2560x1600, KMS, medium effects, Radeon HD3870
without the patch: min avg max 14/28/56
with the patch: min avg max 14/28/56

[Obscene_CNN]

Nexuiz timedemo - demo1
2560x1600, KMS, medium effects, Radeon HD3870
without the patch: min avg max 14/28/56
with the patch: min avg max 14/28/56

Updated KMS patch is coming.... It has more optimizations done to the KMS code. I have given the patches to xming and he will test them and release them probably tomorrow.