[osg-users] [3rdparty] Distributed rendering (was: RE: Cloud node kit)
torben at dannhauer.info
Tue Apr 12 01:52:38 PDT 2011
with framelock I meant some kiond of "hard" lock: the whole rendering cluster waits until all includes machines have rendered their image and the swap together.
This way even in high dynamic scenes you can't see the blendzone between two channels.
Otherwise you could the the blendzone, if one machine is one or frames behind and the scene is high dynamic.
Think aof a flight simulator with artistic airplane. the can rotate with 400 degree per second: This means a rotation offset of 6.7 degree for each frame(60Hz) behind. If one machine has 4 meters channel width, your horizon has a offset of about 20 cm if your pivot point is centered on that channel!
With a simple socket communication without further frame lock implementations you have to trust that your computers are fast enough to be up to date.
Nvidias Professional cards support Framelock (http://www.nvidia.de/object/feature_frame_lock_de.html) but, I would like to get a similar feature by consumer cards...
osgVisual has a UDP implementation. I used the "eNet" package, it is a lightweight UDP package with performs all the lowlevel UDP stuff for me. It is procedural, but I implemented a class based wrapper.
The distributed rendering feature in osgVisual is only working stage, if you have design and/or code improvements, let me know, I'll be happy to integrate improvements you develop!
Read this topic online here:
More information about the osg-users