Yeah, been there, tried that. Turns out I kick started a cuda version that got embraced by eurekafag, which was pretty cool. But make no mistake, every time someone builds a slightly better version, the tendency is to keep it to themselves, unless they lack the hardware to use it to their own profit. Which accounts for the rest of us, I guess
Anyway, I've been moving my attention elsewhere as the amount of interest I got from the community to pursue this was so small I feel I'm just doing this for myself. Which is fine, except I have other things I'd rather do right now.
But if someone preps up a clean implementation against the default client I'll certainly help with all I learned in cuda (and to a lesser extent in opencl) and be a very avid tested!