kernel 2.6.9-rc2-mm2 vs glxgears

[Gene Heskett]

On Tuesday 28 September 2004 22:10, Dave Airlie wrote:
>> And so I'm stuck at some default that gives about 10 fps then?
>>
>> And where can this patch be obtained?
>
>Linus tree has it....
>http://linux.bkbits.net:8080/linux-2.5/diffs/drivers/char/drm/drm_dr
>v.h at 1.48?nav=index.html|src/|src/drivers|src/drivers/char|src/driver
>s/char/drm|hist/drivers/char/drm/drm_drv.h

I found this, but at a point about 39 lines of code short of the 
480th.  Applied to rc2-mm4 and building for test.

Now rebooted...
And it made no difference, although I see it listed in a cat 
of /proc/interrupts as IRQ5 which it shares with the ehci driver.
gxgears is still running at 10 fps:
51 frames in 5.0 seconds = 10.200 FPS
50 frames in 5.0 seconds = 10.000 FPS
50 frames in 5.0 seconds = 10.000 FPS
50 frames in 5.0 seconds = 10.000 FPS
44 frames in 5.0 seconds =  8.800 FPS
32 frames in 5.0 seconds =  6.400 FPS
42 frames in 5.0 seconds =  8.400 FPS
46 frames in 5.0 seconds =  9.200 FPS
49 frames in 5.0 seconds =  9.800 FPS
49 frames in 5.0 seconds =  9.800 FPS
50 frames in 5.0 seconds = 10.000 FPS

>I've no idea if it was in -rc2 or not .. or if -mm has picked it up
> I suspect Andrews tree should have it,  again I don't track Andrews
> tree and don't fix bugs in it, if you can reproduce this with
>linux-2.6.9-rc2 or linux-2.6.9-bk then I'll spend some time worrying
>about it..
>
>Andrews tree is unstable and is not meant to be used in production
>environments...

And what do you think the linus -rc2 is?  'twon't boot on a lot of 
machines.  Following the threads in the hours after the announcement, 
I skipped that one...  But, that also means I don't know for sure on 
this particular machine, so I will build it and try it and report 
back.

Rebooted to plain old -rc2, cat proc/interrupts does show it, and 
those 2 patch lines aren't in -rc2, and glxgears runs at:
937 frames in 5.0 seconds = 187.400 FPS
995 frames in 5.0 seconds = 199.000 FPS
994 frames in 5.0 seconds = 198.800 FPS
995 frames in 5.0 seconds = 199.000 FPS
994 frames in 5.0 seconds = 198.800 FPS
995 frames in 5.0 seconds = 199.000 FPS
995 frames in 5.0 seconds = 199.000 FPS
995 frames in 5.0 seconds = 199.000 FPS
868 frames in 5.0 seconds = 173.600 FPS <- I pulled it out to about 
4.5" square on my 17" 16--x1200 screen
860 frames in 5.0 seconds = 172.000 FPS
862 frames in 5.0 seconds = 172.400 FPS
861 frames in 5.0 seconds = 172.200 FPS
862 frames in 5.0 seconds = 172.400 FPS
861 frames in 5.0 seconds = 172.200 FPS
861 frames in 5.0 seconds = 172.200 FPS
---
154 frames in 5.0 seconds = 30.800 FPS <- and here I made it full 
screen.

In all cases, the cpu load was minimal, a few precent at worst.

So, while there is a speedup, its nowhere near what the MesaLibs stuff 
can do (550fps), albeit thats at 100% cpu usage.  So something in the 
-mm kernels is slowing it down hugely.

Are there any known patches I should apply to -rc2 immediately?

[...]

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
99.26% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attorneys please note, additions to this message
by Gene Heskett are:
Copyright 2004 by Maurice Eugene Heskett, all rights reserved.