The Accelerated Linear Algebra (XLA) is a domain specific compiler developed by TensorFlow for optimizing its computations. By this, you will get improvements in speed, memory usage and even portability on mobile platforms.

Initially, you won't see much benefit from XLA because it's still experimental but you can try it by using the just-in-time compilation or ahead-of-time compilations.

First, we are going to briefly mention the key strengths of TensorFlow and see how TensorFlow team had the challenge to keep and increase these key strengths.