How to generate binomial random variates on the gpu in tensorflow

I would like to know how to generate an array of binomial random variates on the gpu from tensorflow. The following

with tf.device("gpu"):
  g = tf.random.get_global_generator()
  samples = g.binomial(shape = shape, counts = counts, probs = probs)

gives the error

No registered 'StatefulRandomBinomial' OpKernel for 'GPU' devices compatible with node 
    .  Registered:  device='CPU'; dtype in [DT_INT64]; T in [DT_DOUBLE]

That makes me wonder whether binomials are supported.

In addition, how can I accurately time this process? I'm trying to figure out which implementation, across multiple libraries/packages, is fastest.