Thank you! I'd be very interested to see that. I've had a couple of fun Sundays now. I'm looking forward to the next pieces- but moving the logic to the gpu (either by shaders or GPGPU) has not been something I've been planning to do.

The simulation being massively parallel, it is a perfect fit for a fragment shader. Since the interactions between the particles are mostly local, taking advantage of the fact that the GPU loads a block of neighbouring fragments in the cache to be able to compute the standard derivatives, it should run very well. I will notify you when I find some time to play with it.