Add reduce-precision to evaluator and add implicit broadcast remover pass.
authorMark Heffernan <meheff@google.com>
Sat, 27 Jan 2018 00:58:00 +0000 (16:58 -0800)
committerTensorFlower Gardener <gardener@tensorflow.org>
Sat, 27 Jan 2018 01:03:12 +0000 (17:03 -0800)
commitf8da6cc63ae1fd71de1ab5d9e91884872b249e55
treeee86b4a6e1dd872d2ec6b08d5c98ca08e3e62994
parentaee7f95a027accc94f1f9130f0cfaecd9399bc1d
Add reduce-precision to evaluator and add implicit broadcast remover pass.
The reduce precision support is cribbed from the CPU/GPU LLVM-emitted
implementation. The implicit broadcast pass removes any implicit broadcasts in
the module replacing them with the equivalent explicit broadcast and reshape
instructions.

PiperOrigin-RevId: 183467648
tensorflow/compiler/xla/service/BUILD
tensorflow/compiler/xla/service/hlo_evaluator.cc
tensorflow/compiler/xla/service/implicit_broadcast_remover.cc [new file with mode: 0644]
tensorflow/compiler/xla/service/implicit_broadcast_remover.h [new file with mode: 0644]
tensorflow/compiler/xla/service/implicit_broadcast_remover_test.cc [new file with mode: 0644]
tensorflow/compiler/xla/tests/BUILD
tensorflow/compiler/xla/tests/reduce_precision_test.cc