On Tue, 22 May 2012 20:04:34 +0200, Ludwig Schmidt-Hackenberg <[email protected]> wrote: > Hi, > > I wanted to ask about the status/plans of including the new CUDA > features introduced with Kepler into PyCUDA.
It seems all of the headline features would be supported without further work on PyCUDA's side, namely nested parallelism, GPU Direct and improved queuing. There is exactly one large-ish change in CUDA 5 from the driver API side, and that is support for mip-mapped arrays. I personally don't view this as a very pressing issue for most compute codes. It'll happen eventually, but I'm not really in a hurry with it. Hope that clarifies things, Andreas
pgpiRJjhBQB3z.pgp
Description: PGP signature
_______________________________________________ PyCUDA mailing list [email protected] http://lists.tiker.net/listinfo/pycuda
