I'm always working on a lot of optimizations but I feel that this one is ready to be tested by the community. This feature speeds up rendering by merging all submodels into one big model while letting the GPU do all the work into positioning and orienting them for the final render. This cuts down CPU overhead by significantly reducing the number of draw calls needed to render a model, especially if the model has a lot of submodels.
This shifts a bit of extra overhead on the GPU though. I'm essentially communicating with the video card through floating point textures and constantly have to update those textures per frame. Also, there's an additional cost in the vertex shader of having to fetch the necessary transform data from those textures and doing an extra matrix multiply. Your mileage may vary with this build.
This requires an OpenGL 3.0 compliant card. I should mention that this build is based off of Trunk Revision 8627.
To switch on and off this feature, load up the builds using the Launcher, go to the "Features" tab, open the "Experimental" drop down selection, and enable "Enable merged index buffers"
Please give feedback of what your performance is with and without this feature enabled. Also be sure to indicate what CPU and GPU you're running on.
EDIT: Please make sure you don't have any *.sdr files in your "effects" folder! This test build has new shader files built to get this working.
http://www.mediafire.com/?mv9gfwq50by8zgg