Yeah, performance is my biggest issue with it to be honest. It's probably down to some lack of experience on my part . At the moment, depending on how many render targets your card supports it will render up to 4(!) extra passes for each mesh excluding alpha blended meshes, which is quite a performance hit. If I could integrate the rendering better into the engine so that halos shaders do most of the work for me that would help, but doing that without breaking something else would be difficult :P.