add documentation of getPtr to overview.md

cuda_bindings/docs/source/overview.md

@@ -312,6 +312,23 @@ maximize performance ({numref}`Figure 1`).
 Screenshot of Nsight Compute CLI output of CUDA Python example.
 ```
 
+## Getting the address of underlying C objects from the low-level bindings
+
+Within the low-level object wrappers there are a number of cdef classes which effectively provide a PyObject interface to cuda types such as CUdevice, which will be used an example
+in this section. The definition is as follows:
+
+```cython
+cdef class CUdevice:
+    ... 
+
+    def __int__(self):
+        return <int>self._ptr[0]
+    def getPtr(self):
+        return <void_ptr>self._ptr
+```
+
+There is an important distinction between the `getPtr()` method and the behaviour of `__int__()`. If the user wants to get the address of the underlying C object, wrapped in the cdef python class, they should call `int(CUdeviceInstance)`, which returns a pointer to the object, while calling `CUdeviceInstance.getPtr()` returns the `void**` address of the pointer to the object.
+
 ## Future of CUDA Python
 
 The current bindings are built to match the C APIs as closely as possible.