Update documentation

Author: bouweandela

doc/source/hostcode.rst

@@ -15,6 +15,8 @@ There are few differences with tuning just a single CUDA or OpenCL kernel, to li
  * You have to specify the lang="C" option
  * The C function should return a ``float``
  * You have to do your own timing and error handling in C
+ * Data is not automatically copied to and from device memory. To use an array in host memory, pass in a :mod:`numpy` array. To use an array
+   in device memory, pass in a :mod:`cupy` array.
 
 You have to specify the language as "C" because the Kernel Tuner will be calling a host function. This means that the Kernel
 Tuner will have to interface with C and in fact uses a different backend. This also means you can use this way of tuning
@@ -94,7 +96,7 @@ compiled C code. This way, you don't have to compute the grid size in C, you can
 
 The filter is not passed separately as a constant memory argument, because the CudaMemcpyToSymbol operation is now performed by the C host function. Also, 
 because the code is compiled differently, we have no direct reference to the compiled module that is uploaded to the device and therefore we can not perform this 
-operation directly from Python. If you are tuning host code, you have to perform all memory allocations, frees, and memcpy operations inside the C host code, 
+operation directly from Python. If you are tuning host code, you have the option to perform all memory allocations, frees, and memcpy operations inside the C host code, 
 that's the purpose of host code after all. That is also why you have to do the timing yourself in C, as you may not want to include the time spent on memory 
 allocations and other setup into your time measurements.
 

kernel_tuner/backends/compiler.py

@@ -28,12 +28,29 @@
 
 
 def is_cupy_array(array):
-    """Check if something is a cupy array."""
+    """Check if something is a cupy array.
+
+    :param array: A Python object.
+    :type array: typing.Any
+
+    :returns: True if cupy can be imported and the object is a cupy.ndarray.
+    :rtype: bool
+    """
     return cp is not None and isinstance(array, cp.ndarray)
 
 
 def get_array_module(*args):
-    """Return the array module for arguments."""
+    """Return the array module for arguments.
+
+    This function is used to implement CPU/GPU generic code. If the cupy module can be imported
+    and at least one of the arguments is a cupy.ndarray object, the cupy module is returned.
+
+    :param args: Values to determine whether NumPy or CuPy should be used.
+    :type args: numpy.ndarray or cupy.ndarray
+
+    :returns: cupy or numpy is returned based on the types of the arguments.
+    :rtype: types.ModuleType
+    """
     return np if cp is None else cp.get_array_module(*args)
 
 
@@ -119,8 +136,8 @@ def ready_argument_list(self, arguments):
 
         :param arguments: List of arguments to be passed to the C function.
             The order should match the argument list on the C function.
-            Allowed values are np.ndarray, and/or np.int32, np.float32, and so on.
-        :type arguments: list(numpy objects)
+            Allowed values are np.ndarray, cupy.ndarray, and/or np.int32, np.float32, and so on.
+        :type arguments: list(numpy or cupy objects)
 
         :returns: A list of arguments that can be passed to the C function.
         :rtype: list(Argument)
@@ -352,8 +369,8 @@ def memset(self, allocation, value, size):
     def memcpy_dtoh(self, dest, src):
         """a simple memcpy copying from an Argument to a numpy array
 
-        :param dest: A numpy array to store the data
-        :type dest: np.ndarray
+        :param dest: A numpy or cupy array to store the data
+        :type dest: np.ndarray or cupy.ndarray
 
         :param src: An Argument for some memory allocation
         :type src: Argument
@@ -372,8 +389,8 @@ def memcpy_htod(self, dest, src):
         :param dest: An Argument for some memory allocation
         :type dest: Argument
 
-        :param src: A numpy array containing the source data
-        :type src: np.ndarray
+        :param src: A numpy or cupy array containing the source data
+        :type src: np.ndarray or cupy.ndarray
         """
         if isinstance(dest.numpy, np.ndarray) and is_cupy_array(src):
             # Implicit conversion to a NumPy array is not allowed.