Tensorflow XLAによる高速化、メモリ効率化、サイズの最適化によるポータビリティ向上について ref: http://qiita.com/GushiSnow/items/a373b8d5d904566f65fd

file0.txt

Do you wish to build TensorFlow with the XLA just-in-time compiler (experimental)? [y/N]

file1.txt

# Config to turn on JIT compilation
config = tf.ConfigProto()
config.graph_options.optimizer_options.global_jit_level = tf.OptimizerOptions.ON_1

sess = tf.Session(config=config)

file2.txt

    jit_scope = tf.contrib.compiler.jit.experimental_jit_scope

    x = tf.placeholder(np.float32)
    with jit_scope():
      y = tf.add(x, x)  # The "add" will be compiled with XLA.

file3.txt

with tf.device("/job:localhost/replica:0/task:0/device:XLA_GPU:0"):
  output = tf.add(input1, input2)