Hi Tomasz, On Sat, 13 Aug 2011 12:41:06 +0200, Tomasz Rybak <[email protected]> wrote: > I have been trying to finally fix problem with CURAND 4 wrapper. > I am sending patch adding classes wrapping generators added in CURAND 4, > i.e. Sobol64 and ScrambledSobol{32, 64}. > I have fixed warnings related to block size in prepare_call. > > I have also tried to fix problem with lack of resources > when creating Sobol64 classes. I have decided to decrease number > of threads in constructor for both 1.x and 2.x devices. > Unfortunately currently I have access to ION (1.1 device), > so I cannot test it on Fermi device - can someone test > it on Tesla C20x0 or GeForce 4x0 or 5x0? > > If this patch is OK, then the only missing part of CURAND 4 > will be log_* generators.
just in case you're wondering: I've received these, and they're just waiting for me to find some time to take a look. Likely by end of the month. Andreas
pgp1FsHvdpuna.pgp
Description: PGP signature
_______________________________________________ PyCUDA mailing list [email protected] http://lists.tiker.net/listinfo/pycuda
