tf.reduce_max inconsistent on comparison

Given a list/tensor of items, I want to check whether the max item of the whole list is the same as the max item in specific part of list:

import tensorflow as tf
a = tf.get_variable('a', (10,100))
b = tf.unstack(a,axis=1)
c = tf.reduce_max(b[0])
d = tf.reduce_max(b[0])
if c == d:
  c = tf.ones((1,100))

sess = tf.Session()
sess.run(tf.global_variables_initializer())
sess.run([c,d])

In the above example, c and d should be the same, however, when you run the model, it doesn't satisfy the condition to re-make variable c as a one vector. This is just an example to show that these types of comparison seem to be wrong. Any idea what would be the correct way to do it?