Ensure custom GPU kernels are properly registered.

This change ensures that GOOGLE_CUDA=1 when compiling .cc files with --config=cuda enabled (also includes other important copts like -fno-exceptions.)
Change: 129242223

tensorflow/contrib/quantization/kernels/quantized_pooling_ops.cc

@@ -29,11 +29,6 @@ limitations under the License.
 #include "tensorflow/core/util/padding.h"
 #include "tensorflow/core/util/tensor_format.h"
 
-#if GOOGLE_CUDA
-#include "tensorflow/core/kernels/maxpooling_op_gpu.h"
-#include "tensorflow/core/kernels/pooling_ops_common_gpu.h"
-#endif  // GOOGLE_CUDA
-
 namespace tensorflow {
 
 typedef Eigen::ThreadPoolDevice CPUDevice;

tensorflow/tensorflow.bzl

@@ -634,6 +634,7 @@ def tf_custom_op_library(name, srcs=[], gpu_srcs=[], deps=[]):
                    srcs=srcs,
                    deps=deps + if_cuda(cuda_deps),
                    data=[name + "_check_deps"],
+                   copts=tf_copts(),
                    linkshared=1,
                    linkopts = select({
                        "//conditions:default": [

tensorflow/tf_exported_symbols.lds

@@ -1 +1,2 @@
 *tensorflow*
+*perftools*gputools*

tensorflow/tf_version_script.lds

@@ -1,6 +1,7 @@
 tensorflow {
   global:
     *tensorflow*;
+    *perftools*gputools*;
   local:
     *;
 };