diff --git a/tensorflow/compiler/tests/BUILD b/tensorflow/compiler/tests/BUILD
index ef19e238589..4bbb2767ac0 100644
--- a/tensorflow/compiler/tests/BUILD
+++ b/tensorflow/compiler/tests/BUILD
@@ -523,8 +523,12 @@ cuda_py_test(
 # --dump_graph_dir, and the config file was written by hand.
 #
 # Run the following to build a minimal benchmark of the computation on Android:
-# $ bazel build -c opt --config=android_arm \
-#       third_party/tensorflow/compiler/tests:lstm_layer_inference_benchmark
+# $ bazel build -c opt --cxxopt='-std=c++11' --linkopt='-lm' \
+#   --cpu=armeabi-v7a \
+#   --host_crosstool_top=@bazel_tools//tools/cpp:toolchain \
+#   --crosstool_top=//external:android/crosstool \
+#   //tensorflow/compiler/tests:lstm_layer_inference_benchmark
+
 #
 # Currently the resulting binary size is ~190KB
 tf_library(