Hi all,
I would like to use pycuda to write a few kernels to interoperate with a
larger cuda/python library I'm already using. I can get the raw device
memory address as a python int, and from searching past threads on the
mailing list it sounds like I should be able to pass this as an argument to
my kernel. However everything I try seems to generate an error -- I'm
currently doing something like this:
.
.
.
myfunc = mod.get_function("myfunc")
# myob.get() returns memory address as int
addr = np.uint32(myob.get())
myfunc(addr, block=block, grid=grid)
but I get:
TypeError: No registered converter was able to produce a C++ rvalue of type
unsigned int from this Python object of type numpy.float64
If I try just addr = myob.get() (a python int for the address), I get:
TypeError: invalid type on parameter #0 (0-based)
Is there a way to do what I'm attempting? I don't see a way to create a
DeviceAllocation from a raw address, but maybe I'm missing something?
Thanks,
-Sam
--
--------------------
J. Samuel Preston
Research Assistant
Scientific Computing and Imaging Institute
University of Utah
_______________________________________________
PyCUDA mailing list
[email protected]
http://lists.tiker.net/listinfo/pycuda
