[FRONTEND] updated TensorWrapper (#299)

python/triton/code_gen.py

@@ -936,16 +936,16 @@ def next_power_of_2(n):
 
 ######
 
-
 class TensorWrapper:
-    def __init__(self, data_ptr, dtype, device):
-        self._data_ptr = data_ptr
+    def __init__(self, base, dtype):
         self.dtype = dtype
-        self.device = device
+        self.base  = base
+        self.is_cuda = base.is_cuda
+        self.device = base.device
     
     def data_ptr(self):
-        return self._data_ptr
+        return self.base.data_ptr()
 
 
 def reinterpret(tensor, dtype):
-    return TensorWrapper(tensor.data_ptr(), dtype, tensor.device)
+    return TensorWrapper(tensor, dtype)