How to pass OpenGL buffer into kernel? Or how to make device pointers (int type) into DeviceAllocation?

[zongyuelu]

Currently I got a OpenGL buffer object, and I did this as in docs

from pycuda import gl as cudaGL
from pycuda.compiler import SourceModule

registeredBuffer = cudaGL.RegisteredBuffer(int(glBuffer))
registeredMapping = registeredBuffer.map()
devptr, size = registeredMapping.dev_ptr_and_size()

srcCode = """
__global__ void foo(float* src) { ... }
"""
func = SourceModule(srcCode).get_function('foo')
func(devptr, block=(32, 32, 1)) # no, devptr can't be recognized as C pointer

where devptr, size here are int type, how to convert these into a DeviceAllocation so that it can be passed into a kernel? Since doc says Function.__call__ method only takes numpy.number and DeviceAllocation as inputs.

[inducer]

Use a sized integer (np.intp) or prepare the call with the appropriate type signature (P).