On Feb 1, 2010, at 9:56 PM, Alex Harrington wrote:

>>> Quite happy to try building 1.0.1 on this machine but surely that  
>>> kind
>>> of load isn't right?
>
>> Nope, that isn't right :-/.
>
>> Maybe sync to vblank is broken and it's updating the display as fast
>> as it can? What framerate is reported if you turn on profiling? What
>> happens if you manually set the framerate to 60?
>
> I'm on a different Intel-based machine this evening with the same  
> issue running 1.0.1. If I manually set the framerate to 60 I see no  
> difference. If I set it to 6 then I see a reduced load for a short  
> period (around 80%) and then it goes to 100% solidly.

Hm. Another idea: Is antialiasing turned on (in /etc/avgrc or  
~/.avgrc)? That would explain why swap buffers is taking forever. The  
intel graphics chips have a very limited pixel fill rate.

>
> Output from the ClashText getting started example with profiling:
>
> [10-02-01 20:50:47.081] PROFILE: Profile Statistics (in us):
> [10-02-01 20:50:47.081] PROFILE: Thread main
> [10-02-01 20:50:47.081] PROFILE: Zone name                           
> Avg. time
> [10-02-01 20:50:47.081] PROFILE: ---------                           
> ---------
> [10-02-01 20:50:47.081] PROFILE: Player - Total frame  
> time              24244
> [10-02-01 20:50:47.082] PROFILE:   Player -  
> handleTimers                    5
> [10-02-01 20:50:47.082] PROFILE:   Player - dispatch  
> events                67
> [10-02-01 20:50:47.082] PROFILE:   Player -  
> PreRender                       4
> [10-02-01 20:50:47.082] PROFILE:   Player -  
> render                      24103
> [10-02-01 20:50:47.082] PROFILE:     Root node:  
> render                   1004
> [10-02-01 20:50:47.082] PROFILE:        
> pushClipRect                       115
> [10-02-01 20:50:47.082] PROFILE:        
> Words::render                      282
> [10-02-01 20:50:47.083] PROFILE:        
> popClipRect                         57
> [10-02-01 20:50:47.083] PROFILE:     Render -  
> wait                          6
> [10-02-01 20:50:47.083] PROFILE:     Render - swap  
> buffers              15702
> [10-02-01 20:50:47.083] PROFILE:   Player -  
> onFrameEnd                      5
> [10-02-01 20:50:47.083] PROFILE:
> [10-02-01 20:50:47.083] PROFILE: Framerate statistics:
> [10-02-01 20:50:47.084] PROFILE:   Total frames: 155
> [10-02-01 20:50:47.084] PROFILE:   Total time: 3.93882 seconds
> [10-02-01 20:50:47.084] PROFILE:   Framerate achieved: 39.6058
> [10-02-01 20:50:47.084] PROFILE:   Frames too late: 77
> [10-02-01 20:50:47.084] PROFILE:   Percent of time spent waiting:  
> 62.1776
> [10-02-01 20:50:47.084] PROFILE:   Framerate goal was: 60.0197
>
> Here's the script:
> #!/usr/bin/env python
> # -*- coding: utf-8 -*-
>
> from libavg import avg
>
> Player = avg.Player.get()
>
> Player.loadFile("text.avg")
> Player.stopOnEscape(True)
> Player.play()
>
> Text.avg:
> <?xml version="1.0"?>
> <avg width="640" height="480">
>  <words id="ClashText" x="10" y="10" font="arial" text="Should I  
> stay or should I go?"/>
> </avg>
>
>
>> If vsync is broken on intel, can you file a bug?
>
> Against the Kernel?


;-)

It's not vsync, so don't file a bug. But since it's possible to work  
around driver bugs, it would have been a libavg bug ;-).

Cheers,

   Uli




--
Any technology distinguishable from magic is insufficiently advanced.

Ulrich von Zadow | +49-172-7872715
Jabber: [email protected]
Skype: uzadow




_______________________________________________
libavg-users mailing list
[email protected]
https://mail.datenhain.de/mailman/listinfo/libavg-users

Reply via email to