Measure the performance of the original placement to ensure that we preserve it
authorBenoit Steiner <bsteiner@google.com>
Thu, 22 Feb 2018 20:18:39 +0000 (12:18 -0800)
committerTensorFlower Gardener <gardener@tensorflow.org>
Thu, 22 Feb 2018 20:24:17 +0000 (12:24 -0800)
commit51b334875125b1e76545d02d2e8e18c7ff2be0af
treeccce37a3e5dda8cb3bb0a6ab989ab8e09b683523
parente2a9276d485ab3c3b5b0ebfbc92fc105cfa7419f
Measure the performance of the original placement to ensure that we preserve it
in case the placer isn't given enough time to find a better solution.

PiperOrigin-RevId: 186655094
tensorflow/python/grappler/graph_placer.py