Hi what you and Łukasz Ligowski are saying is just amazing. It means, that there is generally no problems with PyPy and CUDA.
best regards -- Valery A.Khamenya On Mon, Sep 12, 2011 at 5:50 AM, Ken Watford <kwatf...@gmail.com> wrote: > I started a project called PyCL about two months ago. It's OpenCL > through ctypes, and it works with PyPy. I don't believe it currently > works with the new numpy stuff, but the standard Python array module > should work. > > It's available through the cheeseshop, and there's a repository for it > here: > https://bitbucket.org/kw/pycl > > I can't say I've had any time to work on it much in last month. Image > support wasn't quite ready yet, last I recall, but doing basic stuff > with buffers and kernels should work. Though last time I checked it > ran generally slower in PyPy than CPython. Haven't checked it since > 1.6 came out, though. > > On Sun, Sep 11, 2011 at 6:29 PM, Valery Khamenya <khame...@gmail.com> > wrote: > > Hi all, > > (replying, please, Cc to me ) > > I was quite surprised to see that cooperhead could be compiled and > installed > > OK under PyPy. > > Of course it didn't work, because micronumpy is still young: > > "AttributeError: 'module' object has no attribute 'float64' " > > Did anyone try to execute some numeric stuff using CUDA from PyPy ? > > CUDA + PyPy -- it would be just fantastic. > > best regards > > -- > > Valery A.Khamenya > > > > _______________________________________________ > > pypy-dev mailing list > > pypy-dev@python.org > > http://mail.python.org/mailman/listinfo/pypy-dev > > > > >
_______________________________________________ pypy-dev mailing list pypy-dev@python.org http://mail.python.org/mailman/listinfo/pypy-dev
