diff --git a/lite/demo/cxx/bm_demo/CMakeLists.txt b/lite/demo/cxx/bm_demo/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4119196d16e87f53e7b1d8078fde19d2a1ecb50f
--- /dev/null
+++ b/lite/demo/cxx/bm_demo/CMakeLists.txt
@@ -0,0 +1,14 @@
+cmake_minimum_required(VERSION 3.0)
+project(demo CXX C)
+
+set(TARGET mobilenet_full_api)
+# 1. lib & include
+link_directories(${PROJECT_SOURCE_DIR}/lib ${PROJECT_SOURCE_DIR}/lib/pcie ${PROJECT_SOURCE_DIR}/lib/bmcompiler)
+include_directories(${PROJECT_SOURCE_DIR}/include)
+# 2. compile options 
+add_definitions(-std=c++11 -g -O3 -DNDEBUG -pthread)
+#add_definitions(-g  )
+set(EXECUTABLE_OUTPUT_PATH ${PROJECT_SOURCE_DIR})
+# 3.add executable output
+add_executable(${TARGET} ${TARGET}.cc)
+target_link_libraries(${TARGET} -lpaddle_full_api_shared -lbmlib -lbmrt -lbmcompiler)
diff --git a/lite/demo/cxx/bm_demo/ci.sh b/lite/demo/cxx/bm_demo/ci.sh
new file mode 100755
index 0000000000000000000000000000000000000000..f18724fd662ac0bec3b239b8ee2951fff53c5318
--- /dev/null
+++ b/lite/demo/cxx/bm_demo/ci.sh
@@ -0,0 +1,20 @@
+wget http://paddle-inference-dist.bj.bcebos.com/mobilenet_v1.tar.gz
+tar -xvf mobilenet_v1.tar.gz
+
+mkdir ./lib
+cp -r /paddlelite/build.lite.bm/lite/api/libpaddle_full_api_shared.so \
+/paddlelite/third-party/bmlibs/bm_sc3_libs/lib/bmcompiler \
+/paddlelite/third-party/bmlibs/bm_sc3_libs/lib/bmnn/pcie ./lib
+
+mkdir ./include
+cp -r /paddlelite/lite/api/paddle_place.h /paddlelite/lite/api/paddle_api.h ./include
+
+mkdir ./build
+cd ./build
+cmake ..
+make
+cd ..
+rm -rf ./build
+rm -rf ./mobilenet_v1.tar.gz
+
+./mobilenet_full_api ./mobilenet_v1 224 224
diff --git a/lite/demo/cxx/bm_demo/mobilenet_full_api.cc b/lite/demo/cxx/bm_demo/mobilenet_full_api.cc
new file mode 100644
index 0000000000000000000000000000000000000000..ca557268f729bf71530a70a62311cb62cdf66359
--- /dev/null
+++ b/lite/demo/cxx/bm_demo/mobilenet_full_api.cc
@@ -0,0 +1,90 @@
+// Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <iostream>
+#include <vector>
+#include "paddle_api.h"            //NOLINT
+using namespace paddle::lite_api;  // NOLINT
+int g_batch_size = 1;
+int64_t ShapeProduction(const shape_t& shape) {
+  int64_t res = 1;
+  for (auto i : shape) res *= i;
+  return res;
+}
+
+void RunModel(std::string model_dir, int im_height, int im_width) {
+  CxxConfig config;
+  config.set_model_dir(model_dir);
+  // config.set_model_file(model_dir+"/model");
+  // config.set_param_file(model_dir+"/params");
+  config.set_valid_places({Place{TARGET(kBM), PRECISION(kFloat)},
+                           Place{TARGET(kHost), PRECISION(kFloat)}});
+  std::shared_ptr<PaddlePredictor> predictor =
+      CreatePaddlePredictor<CxxConfig>(config);
+  auto cloned_predictor = predictor->Clone();
+
+  std::unique_ptr<Tensor> input_tensor(std::move(predictor->GetInput(0)));
+  input_tensor->Resize({1, 3, im_height, im_width});
+  auto* data = input_tensor->mutable_data<float>();
+  int item_size = ShapeProduction(input_tensor->shape());
+  for (int i = 0; i < item_size; i++) {
+    data[i] = 1;
+  }
+  for (int i = 0; i < 100; i++) {
+    predictor->Run();
+  }
+
+  FILE* fp = fopen("result.txt", "wb");
+  std::unique_ptr<const Tensor> output_tensor(
+      std::move(predictor->GetOutput(0)));
+  std::cout << "Output shape " << output_tensor->shape()[1] << std::endl;
+  for (int i = 0; i < ShapeProduction(output_tensor->shape()); i++) {
+    fprintf(fp, "%f\n", output_tensor->data<float>()[i]);
+  }
+
+  fclose(fp);
+
+  std::unique_ptr<Tensor> cloned_input_tensor(
+      std::move(cloned_predictor->GetInput(0)));
+  cloned_input_tensor->Resize({1, 3, im_height, im_width});
+  auto* cloned_data = cloned_input_tensor->mutable_data<float>();
+  for (int i = 0; i < ShapeProduction(cloned_input_tensor->shape()); ++i) {
+    cloned_data[i] = 1;
+  }
+
+  for (int i = 0; i < 1; i++) {
+    cloned_predictor->Run();
+  }
+  std::unique_ptr<const Tensor> cloned_output_tensor(
+      std::move(cloned_predictor->GetOutput(0)));
+  std::cout << "cloned_Output shape " << cloned_output_tensor->shape()[1]
+            << std::endl;
+  for (int i = 0; i < ShapeProduction(cloned_output_tensor->shape());
+       i += 100) {
+    std::cout << "cloned_Output[" << i
+              << "]: " << cloned_output_tensor->data<float>()[i] << std::endl;
+  }
+}
+
+int main(int argc, char** argv) {
+  if (argc < 2) {
+    std::cerr << "[ERROR] usage: ./" << argv[0] << " naive_buffer_model_dir\n";
+    exit(1);
+  }
+  std::string model_dir = argv[1];
+  int im_height = std::stoi(argv[2]);
+  int im_width = std::stoi(argv[3]);
+  RunModel(model_dir, im_height, im_width);
+  return 0;
+}
diff --git a/lite/tools/build_bm.sh b/lite/tools/build_bm.sh
index 53f62695bcba061637956a52c1cd12c2e8b3c1c5..4d4d12a98974dd38d9bfc48c1df82527babf353f 100755
--- a/lite/tools/build_bm.sh
+++ b/lite/tools/build_bm.sh
@@ -102,7 +102,10 @@ function main {
         case $i in
             --target_name=*)
                 TARGET_NAME="${i#*=}"
-                build_bm
+                shift
+                ;;
+            --test=*)
+                WITH_TESTING=${i#*=}
                 shift
                 ;;
             *)
@@ -112,5 +115,6 @@ function main {
                 ;;
         esac
     done
+    build_bm
 }
 main $@