documents compute_gradient function

cpp/field.cpp

@@ -1884,6 +1884,18 @@ double field<rnumber, be, fc>::L2norm(
     }
 }
 
+/** \brief Compute field gradient
+ *
+ *  Scalar fields are turned into vector fields
+ *
+ *  Vector fields are turned into tensor fields, with
+ *   - fastest dimension corresponding to vector component
+ *   - second fastest dimension corresponding to derivative direction
+ *  i.e. values should be accessed with something like
+ *   val(index, nabla_component, vector_component)
+ *  (either cval or rval).
+ * */
+
 template <typename rnumber,
           field_backend be,
           field_components fc1,