Add TensorFlow custom op and run tvm in TensorFlow

Just a quick question, is there any conflicts between the TVM GPU memory allocator and TF allocator ?