[XLA:GPU] Emit fused reduces from batchnorm expander
authorBenjamin Kramer <kramerb@google.com>
Tue, 22 May 2018 19:34:51 +0000 (12:34 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Tue, 22 May 2018 19:37:10 +0000 (12:37 -0700)
commit67b6696f9620734369ae99e7895fa6570d7faca6
tree80ec2d10040bf926522b139010ba2f2598185027
parentb3e68e52f1f6488710c478596c30e0f0eb2dcad1
[XLA:GPU] Emit fused reduces from batchnorm expander

This is an intermediate step until we have working multi-output fusion. Once
we have it, this change should be reverted as it might interfere with fusion.

PiperOrigin-RevId: 197605814
tensorflow/compiler/xla/service/gpu/gpu_compiler.cc