Recursion Question about future support for recursion in CUDA

I understood for one of the presenters of the current CUDA webinars that Nvidia is working on support for recursion in CUDA.
Is there a timeline for when recursion will be available yet?
And will this require new hardware?

Thanks!
Colt

It’s possible that recursion may be possible with future hardware, but the performance is too bad for it to make sense with current hardware (you can put a stack in local memory, but it’s slow).

Note that it’s always going to be faster to convert a recursive algorithm to an iterative one if you can.

Thanks!

Colt