the compression is really extreme unfortunately :/ nevertheless one can make out the smoother volume data quite clearly, it's good that you're not doing vol[z * (xres * yres) + y * (xres) + x]++ anymore, that must have been hell for your continuous refraction code!
a friend of mine sent me a link recently, which links to the other:
http://www.daimi.au.dk/~trier/?page_id=98http://www.vis.uni-stuttgart.de/ger/research/fields/current/spvolren/i figure getting your data into the first link's code should be really really simple; all you need to do is find a gf6/7 gpu with a decent amount of memory and you should be set for a 1000x or more speedup.
