I posted a 2013-12-07 version. Most notable difference is that the Fermi kernel now targets the Compute 2.0 architecture explicitly. Expect maybe a very small boost in hash rates. For me it went up from 228->235 with the cudaminer x64 version for my Compute 2.0 card.

Next work item on my TODO list are:

-make cudaminer suitable for >1 MHash/s operations, even with weak single or dual core CPUs
 by doing the required SHA256 hashing (for scrypt) also on the GPU

-finish and operate my 1.5 MHash/s build (3x GTX 780Ti in one mainboard)