Text archives Help
- From: Carson Brownlee <brownlee@cs.utah.edu>
- To: Abe Stephens <abe@sci.utah.edu>
- Cc: MANTA <manta@sci.utah.edu>
- Subject: Re: [Manta] vsync?
- Date: Sun, 06 Jan 2008 22:09:54 -0700
Vysinc is disabled by default on macs, or at least it was. ATI used to
make a control panel for this kind of thing, but I don't think they ever
updated it after the intel switch and I don't think nvidia ever made
anything like it.
I'm still getting the same problem on hex. you're pretty sure the
framerate text is solely to blame?
Carson
On Sun, 2008-01-06 at 20:55 -0700, Abe Stephens wrote:
>
Thanks for the tip--I think I was able to resolve this by updating the
>
framerate text in a GUI idle callback, instead of directly from the gl
>
swapbuffer related code. While swapping buffers is pretty low cost, I
>
guess the flurry of refresh messages sent by the gui when you change
>
some text is too much to handle at high rates.
>
>
On another note, I've noticed the empty scene frame rate in the viewer
>
tops out at 60hz on my mac, so maybe there is some sort of sync going
>
on (supposedly vsync is disabled by default).
>
>
Abe
>
>
>
On Jan 6, 2008, at 7:21 PM, R. Keith Morley wrote:
>
>
>
>
> We had this issue in a c++ wx based manta gui. The status bar
>
> update was being updated
>
> once per frame, and after 50 fps it was hogging the display thread
>
> -- i just throttled the status bar
>
> update so that it only updates every 0.02s or something and the
>
> problem went away.
>
>
>
> k
>
>
>
> Abe Stephens wrote:
>
>>
>
>> For everyone's benefit: the issue Carson is describing is that on
>
>> certain systems the synchronization between the Manta display
>
>> thread and the GUI thread (wxPython in this case) preempts and
>
>> starves the GUI event handling loop when the frame rate exceeds
>
>> 30-60 fps. Often this happens during a visualization when the user
>
>> accidentally moves the camera away from the dataset--the frame rate
>
>> goes berserk and since the GUI becomes unresponsive it's impossible
>
>> to move the dataset back into view.
>
>>
>
>> I'm not sure if this is a vsync issue--as I think it can occur
>
>> (occasionally) well below the refresh rate of our displays. James
>
>> might know better, since he wrote the python/manta interaction
>
>> code, but I'd imagine we can address the starvation issue directly
>
>> somewhere in there. This problem didn't arise with the X11, Fox, or
>
>> Performer GUIs because those approaches rendered completely
>
>> asynchronously.
>
>>
>
>> Abe
>
>>
>
>>
>
>> On Jan 6, 2008, at 6:39 PM, Carson Brownlee wrote:
>
>>
>
>>> Manta doesn't have any sort of vsync right? The only way I know
>
>>> how to do this is through vendor specific OGL calls (CGL, WGL,
>
>>> etc.). Is there a better way to do this (other than if delta <
>
>>> time then delay (time-delta))? The problem I'm having is that
>
>>> when Manta is rendering nothing on certain systems it seems slow
>
>>> down wxPython windows.
>
>>> thanks,
>
>>> Carson
>
>
Archive powered by MHonArc 2.6.16.