Hi Claudio, If you run osgviewer and then press 's' a couple of times to full stats you see the various phases of operation, the last phase is the orange bar which represents the time on the GPU (this is only available on NVidia hardware though due to use of extensions to get this time). The GPU stats position is an best estimate (the extension doesn't provide the absolute timing), but the length of the bar is accurate.
Now try toggle the threading model by pressing 'm', SingleThreaded will actually give you the lowest latency if everything is rendering comfortable within one frame. Threading models like DrawThreadPerContext scale very well so can handle big models better but it does add a frame latency due to the double buffering of the rendering backend. Robert. On 7/13/07, Claudio Arduino <[EMAIL PROTECTED]> wrote:
I have a question about the latency between when an instruction changing the scene graph is issued and when the effect of that instruction is visible on the screen. I'm using a very simple application (OSG 2.0) with a very simple scene graph made by a square with a white face and a black face, running at 75 fps with vsync enabled on the graphic card; ever every 30 frames two instruction are issued: 1 a rotation instruction provoking a color change from black to white on the whole screen 2 a character is sent on the serial port I'm measuring whith an oscilloscope the time elapsed from the beginning of the electric signal on the serial port and the pattern change in the electric signal on the VGA Red signal line due to the black to white transition; the result is that it takes two frames for a change to be visible on the screen (the measured time is around 26 ms) when OSG viewer is configured in single thread. Does someone know why there are two frames of delay instead of one? (please note that we are not using triple buffering on the graphic card) _______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
_______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
