We've been doing quite advanced GPU calculations using rgb textures since pi 1. We've done this through some hacks; 16 bit data is transfered to the gpu (divided into 5-6-5 bits), after which the original data is calculated from the rgb data through a shader and the result is written to another texture, etc. It really feels like a clumsy solution to do this with OpenGL ES 2.0 (although it's working just fine). But is this still the way to go? Does someone have any thoughts on whether this is something that might be more handy to do with some other method on the pi, now or in the future?