System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): Yes
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 14.04
• TensorFlow installed from (source or binary): Source
• TensorFlow version (use command below): ('v1.0.0-1783-g4c3bb1a', '1.0.0')
• Bazel version (if compiling from source): 0.4.5
• CUDA/cuDNN version: -
• GPU model and memory: -
• Exact command to reproduce: -

Problem description

I have written my custom build file to integrate Polly into XLA's LLVM build. I have included a dependency to @llvm//:polly in my tensorflow/blob/master/tensorflow/compiler/xla/service/cpu/BUILD file under the compiler_functor rule (Since this is the file with the LLVM PassManagerBuilder defined).
I also added the header

#include "external/llvm/tools/polly/include/polly/RegisterPasses.h"

in this file.
Somehow, this does not seem to be enough for XLA to recognize Polly. When I add -

polly::registerPollyPasses(...)

to this file, it fails to compile with

ERROR: Symbol polly was not declared here

I am unable to fix this (possibly linking) issue.

@hawkinsp - Could you please take a look at this?

Please refer to #10288 for more relevant information.