On Samstag 09 Januar 2010, David Dreisigmeyer wrote:
> The examples run now but I'm getting:
> 
> [snip]
> ImportError: dlopen(/Library/Python/2.6/
> site-packages/pycuda-0.94beta-py2.6-macosx-10.6-i386.egg/pycuda/_driver.so,
> 2): no suitable image found.  Did find:
> /Library/Python/2.6/site-packages/pycuda-0.94beta-py2.6-macosx-10.6-
> i386.egg/pycuda/_driver.so: mach-o, but wrong architecture

In what sense do the examples run? Do all of them give the above error?
Only some? 

I don't quite understand the error message either--it seems to say that
you're on an i386 Python, but _driver.so (.so? really?  shouldn't that
be .dylib on Mac?) somehow managed to build for a non-matching
architecture (x86_64 then?). PyCUDA's setup script explicitly requests
i386, because I think Mac CUDA is 32-bit only.  (confirmation?)

Can one of the Mac gurus take a look?

Andreas

Attachment: signature.asc
Description: This is a digitally signed message part.

_______________________________________________
PyCUDA mailing list
[email protected]
http://tiker.net/mailman/listinfo/pycuda_tiker.net

Reply via email to