On Mon, 7 Nov 2011 14:17:03 +1300, Igor <rych...@gmail.com> wrote: > I found a place to patch in pycuda-2011.1.2\pycuda\gpuarray.py, > class vec: > ... > for base_name, base_type, counts in [ > ('char', np.int8, [1,2,3,4]), > ('uchar', np.uint8, [1,2,3,4]), > ('short', np.int16, [1,2,3,4]), > ('ushort', np.uint16, [1,2,3,4]), > ('int', np.uint32, [1,2,3,4]), > ('uint', np.uint32, [1,2,3,4]), > ('long', long_dtype, [1,2,3,4]), > ('ulong', ulong_dtype, [1,2,3,4]), > ('longlong', np.int64, [1,2]), > ('ulonglong', np.uint64, [1,2]), > ('float', np.float32, [1,2,3,4]), > ('ulonglong', np.float64, [1,2]), > ]: > ... > Can I just change the second occurrence of ulonglong to double? Do I > need to recompile anything then?
Thanks for the report--I've fixed this upstream. Hope this solves your issue. Best, Andreas
pgpdOYBRUoJ9v.pgp
Description: PGP signature
_______________________________________________ PyCUDA mailing list PyCUDA@tiker.net http://lists.tiker.net/listinfo/pycuda