Hi, Just a quick remark about the 20000 cores. Take it with a grain of sal as I didn't check the code. Current Nvidia GPU support 512 concurrent cores, but 16k concurrent threads. That is pretty close to 20k. So the next generation will probably bust the 20k in number of threads.
Are you sure you talk about cores and not threads? Regards Frédéric Bastien On Wed, May 18, 2011 at 8:27 AM, Tomasz Rybak <[email protected]> wrote: > Dnia 2011-05-18, śro o godzinie 08:19 +0200, Andreas Kloeckner pisze: >> On Mon, 09 May 2011 17:32:05 +0200, Tomasz Rybak <[email protected]> wrote: >> > Thanks for the tip. >> > Here is the patch. It works, and generates >> > quasi-random numbers. The only problem is when >> > with calling curandGetDirectionVectors - it generates >> > only 20000 vectors, so we will have repeated generators >> > when we have GPUs with more than 20000 cores ;-) >> > >> > BTW - there is difference in wrap_cudadrv.cpp between >> > master and curand branch. I had to fix it but >> > did not include this change in the patch. >> > >> > Please apply this patch - and I believe >> > that we could think about merging curand branch >> > into master. >> >> Landed in master with a few fixes. Let me know if I broke anything. > > After compiling master Sobol32 behaves OK. > I also do not see anything broken when looking at your changes. > >> >> Thanks for seeing this through to the end! > > Thanks. > When are you going to release final 2011.1? > I would like to know to prepare Debian package. > > Regards. > > -- > Tomasz Rybak <[email protected]> GPG/PGP key ID: 2AD5 9860 > Fingerprint A481 824E 7DD3 9C0E C40A 488E C654 FB33 2AD5 9860 > http://member.acm.org/~tomaszrybak > > _______________________________________________ > PyCUDA mailing list > [email protected] > http://lists.tiker.net/listinfo/pycuda > > _______________________________________________ PyCUDA mailing list [email protected] http://lists.tiker.net/listinfo/pycuda
