A problem about equal-style variable when using the user-cuda


so the actual calculation of the variable must be done on the CPU since its a serial process anyway. Now if the variable itself requires the evaluation of computes those could be done on the GPU.

I am currently working on the framework to support this, turns out this is slightly more difficult than anticipated (but what is not?).

I'll let you know when I got it done, and send will send you some preview code then.

fix_move shouldn't be too hard to implement either, fix_deform might be a bit more tricky, but their is no principal reason why it should not work.


-------- Original-Nachricht --------