diff --git a/CMakeLists.txt b/CMakeLists.txt
index f4fb9f42ef39389853e654eb4dbdb6a46139e5f1..32e909e1cce001dff49a90d7753be780e14f08ea 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -17,6 +17,7 @@ endif()
 
 
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14")
+set(CMAKE_BUILD_TYPE Release)
 set(CMAKE_VERBOSE_MAKEFILE ON)
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY build)
diff --git a/src/common/enforce.h b/src/common/enforce.h
index 4b7c8dc0e267bc2862d2665e71085d679dceb1ff..51d2110e32433686d1b3353bc63b92a564a13e9d 100644
--- a/src/common/enforce.h
+++ b/src/common/enforce.h
@@ -17,7 +17,6 @@ limitations under the License. */
 #ifdef ENABLE_EXCEPTION
 #include <stdio.h>
 #include <exception>
-#include <sstream>
 #include <string>
 
 #endif
@@ -31,12 +30,11 @@ struct PaddleMobileException : public std::exception {
 
   PaddleMobileException(const char *header, const char *detail,
                         const char *file, const int line) {
-    std::stringstream ss;
-    ss << exception_prefix << "| " << header << "\n";
-    ss << "| [in file] : " << file << " \n";
-    ss << "| [on line] : " << line << " \n";
-    ss << "| [detail]  : " << detail;
-    message = ss.str();
+    char buffer[1500];
+    snprintf(buffer, sizeof(buffer),
+             "%s| %s \n| [in file] : %s\n| [on line] : %d\n| [detail]  : %s\n",
+             exception_prefix.c_str(), header, file, line, detail);
+    message = std::string(buffer);
   }
   const char *what() const noexcept { return message.c_str(); }
 };
diff --git a/src/common/type_define.h b/src/common/type_define.h
index 63665bf7933f773b2b3de40ade9c700e3e93e6a9..c26cdd91e0694d44cca9443503d3e263ee21f201 100644
--- a/src/common/type_define.h
+++ b/src/common/type_define.h
@@ -16,7 +16,6 @@ limitations under the License. */
 
 #include <map>
 #include <string>
-#include <unordered_set>
 #include <vector>
 #include "framework/attribute.h"
 #include "framework/scope.h"
@@ -40,13 +39,6 @@ using OpCreator = std::function<framework::OperatorBase<Dtype> *(
     const framework::AttributeMap & /*attrs*/,
     std::shared_ptr<framework::Scope> /*scope*/)>;
 
-using GradOpMakerFN =
-    std::function<std::vector<std::unique_ptr<framework::OpDesc>>(
-        const framework::OpDesc &,
-        const std::unordered_set<std::string> & /*no_grad_set*/,
-        std::unordered_map<std::string, std::string> * /*grad_to_var*/,
-        const std::vector<framework::BlockDesc *> &grad_block)>;
-
 using InferVarTypeFN = std::function<void(const framework::OpDesc & /*op_desc*/,
                                           framework::BlockDesc * /*block*/)>;
 
diff --git a/src/common/types.h b/src/common/types.h
index 0c0a6ac1c04cbc9e3ddb916fc656280c36e47d5f..272db3ef70f5bf923487fd29b2666f94707040bf 100644
--- a/src/common/types.h
+++ b/src/common/types.h
@@ -16,7 +16,6 @@ limitations under the License. */
 
 #include <string>
 #include <unordered_map>
-#include <utility>
 
 namespace paddle_mobile {
 enum class Precision : int { FP32 = 0 };
diff --git a/src/framework/attribute.h b/src/framework/attribute.h
index bd95326949dfb5e30be4f9739e06f592030c8e83..2eddb27ed49c82d2d5f1b3bb1ce6d3b1c65d3f3e 100644
--- a/src/framework/attribute.h
+++ b/src/framework/attribute.h
@@ -17,6 +17,7 @@ limitations under the License. */
 #include <string>
 #include <unordered_map>
 #include <vector>
+
 #include "common/enforce.h"
 #include "common/log.h"
 #include "common/variant.h"
diff --git a/src/framework/lod_tensor.h b/src/framework/lod_tensor.h
index bab3db1389610e7ed5db1a387004bdf95267867f..3b34d664d3608dd361ed7c7bb549870284adcc33 100644
--- a/src/framework/lod_tensor.h
+++ b/src/framework/lod_tensor.h
@@ -16,7 +16,6 @@ limitations under the License. */
 
 #include <memory>
 #include <string>
-#include <utility>
 #include <vector>
 #include "tensor.h"
 #include "tensor_util.h"
diff --git a/src/framework/operator.h b/src/framework/operator.h
index 17a8359d8caf8b70c4aa2b3d50c0f8cdbf78549c..d9b74e8887944774aefadf5ead0a74b1f7eac79d 100644
--- a/src/framework/operator.h
+++ b/src/framework/operator.h
@@ -16,7 +16,6 @@ limitations under the License. */
 
 #include <map>
 #include <string>
-#include <utility>
 #include <vector>
 
 #include "common/enforce.h"
diff --git a/src/framework/program/program-optimize/node.cpp b/src/framework/program/program-optimize/node.cpp
index bbec2c68b611dbf87eadc8fd274155cdccc3accd..d128d11edbcac4d1fa916feb7348970abc50147a 100644
--- a/src/framework/program/program-optimize/node.cpp
+++ b/src/framework/program/program-optimize/node.cpp
@@ -43,47 +43,6 @@ bool Node::operator==(const Node &in) {
   return true;
 }
 
-bool Node::CanSplit(std::unordered_set<std::string> complex_compute_set) {
-  bool split = false;
-  CanSplit(&split, false, 0, &complex_compute_set, this);
-  return split;
-}
-
-void Node::CanSplit(bool *split, bool spliting, int complex_count,
-                    std::unordered_set<std::string> *complex_compute_set,
-                    Node *pre_node) {
-  if (spliting) {
-    if (complex_compute_set->find(this->type_) != complex_compute_set->end()) {
-      complex_count++;
-    }
-  }
-
-  if (inputs_.size() > 1 && pre_node != inputs_.back()) {
-    return;
-  }
-  if (inputs_.size() > 1 && pre_node == inputs_.back()) {
-    if (complex_count > 1) {
-      *split = true;
-      return;
-    }
-  }
-
-  // multi output, to check
-  if (outputs_.size() > 1) {
-    spliting = true;
-    complex_compute_set = 0;
-  } else {
-    if (spliting == true && inputs_.size() > 0) {
-      spliting = false;
-    } else {
-    }
-  }
-
-  for (auto &output : outputs_) {
-    output->CanSplit(split, spliting, complex_count, complex_compute_set, this);
-  }
-}
-
 std::vector<std::shared_ptr<framework::OpDesc>> Node::OpDescs(uint size) {
   std::vector<std::shared_ptr<framework::OpDesc>> op_descs;
   OpDescs(size - 1, &op_descs);
@@ -101,107 +60,6 @@ void Node::OpDescs(uint index,
   }
 }
 
-void Node::OpDescs(std::vector<std::shared_ptr<framework::OpDesc>> *op_desc,
-                   Node *node, bool adding_thread, int thread_num) {
-  if (outputs_.size() > 1) {
-    adding_thread = false;
-  }
-
-  bool can_add_split = false;
-  // 如果当前节点有多个输出 并且 只有当前节点对应的 op_desc_ 输出数为 1 时支持
-  if (outputs_.size() > 1 &&
-      op_input_output_key[op_desc_->type_].second.size() == 1) {
-    can_add_split = true;
-
-    // 遍历当前节点的 output 节点
-    for (const auto &output : outputs_) {
-      // 不支持 output 有多个 output 的情况
-      if (output->outputs_.size() > 0) {
-        can_add_split = false;
-        break;
-      }
-
-      //与节点关联的 OpDesc
-      std::shared_ptr<framework::OpDesc> &op_desc = output->op_desc_;
-
-      //获取这个 op 的 inputs key 和 outputs key
-      auto inputs_and_outputs = op_input_output_key[op_desc->type_];
-
-      //判断现在 是否存在这个 op
-      //判断这个 output 和 input key 的 size 等于 1
-      if (op_input_output_key.find(op_desc->type_) !=
-              op_input_output_key.end() &&
-          inputs_and_outputs.first.size() == 1 &&
-          inputs_and_outputs.second.size() == 1) {
-        auto inputs_of_output = op_desc->Input(inputs_and_outputs.first[0]);
-        auto outputs_of_output = op_desc->Output(inputs_and_outputs.second[0]);
-
-        // 判断一下, 如果输入和输出没有同名, 是支持的
-        for (int i = 0; i < inputs_of_output.size(); ++i) {
-          std::string input_of_output = inputs_of_output[i];
-          for (int j = 0; j < outputs_of_output.size(); ++j) {
-            std::string output_of_output = outputs_of_output[j];
-            if (input_of_output == output_of_output) {
-              DLOG << "output的 output 包含 input" << input_of_output;
-              can_add_split = false;
-              break;
-            }
-          }
-        }
-      } else {  // 如果模型中包含没有的 op, 则不支持添加 split
-        DLOG << "找不到 这个 op 类型: " << output->op_desc_->type_;
-        can_add_split = false;
-      }
-    }
-  }
-
-  if (inputs_.size() > 1 && node != inputs_.back()) {
-    return;
-  } else if (inputs_.size() > 1 && node == inputs_.back()) {
-    adding_thread = false;
-    op_desc->push_back(this->op_desc_);
-  } else {
-    op_desc->push_back(this->op_desc_);
-  }
-  if (adding_thread) {
-    Attribute attr;
-    attr.Set<int>(thread_num);
-    this->op_desc_->attrs_["thread"] = attr;
-  }
-
-  if (can_add_split) {
-    adding_thread = true;
-    std::shared_ptr<OpDesc> split_op_desc = std::make_shared<OpDesc>();
-    split_op_desc->type_ = G_OP_TYPE_SPLIT;
-    auto outputs = this->op_desc_->Output(
-        op_input_output_key[this->op_desc_->Type()].second[0]);
-    split_op_desc->inputs_ = {
-        {op_input_output_key[G_OP_TYPE_SPLIT].first[0], outputs}};
-    auto &split_outputs =
-        split_op_desc->outputs_[op_input_output_key[G_OP_TYPE_SPLIT].second[0]];
-    for (const auto &output : outputs_) {
-      split_outputs.push_back(outputs[0]);
-    }
-    DLOG << "add split";
-    op_desc->push_back(split_op_desc);
-  }
-
-  for (int i = 0; i < outputs_.size(); ++i) {
-    auto &output = outputs_[i];
-    if (can_add_split) {
-      output->OpDescs(op_desc, this, adding_thread, i);
-    } else {
-      output->OpDescs(op_desc, this, adding_thread, thread_num);
-    }
-  }
-}
-
-std::vector<std::shared_ptr<framework::OpDesc>> Node::OpDescs() {
-  std::vector<std::shared_ptr<framework::OpDesc>> op_descs;
-  OpDescs(&op_descs, this, false, 0);
-  return op_descs;
-}
-
 std::shared_ptr<Node> Node::To(int size) {
   std::shared_ptr<Node> node = std::make_shared<Node>();
   this->To(size - 1, node);
diff --git a/src/framework/program/program-optimize/node.h b/src/framework/program/program-optimize/node.h
index 7c224295cbfda18d8122820d848775d5c515bdfe..dd19b84fd55dd37cf634c9793e35e72ee2202b19 100644
--- a/src/framework/program/program-optimize/node.h
+++ b/src/framework/program/program-optimize/node.h
@@ -16,8 +16,6 @@ limitations under the License. */
 
 #include <map>
 #include <string>
-#include <unordered_set>
-#include <utility>
 #include <vector>
 
 #include "common/log.h"
@@ -49,16 +47,10 @@ class Node {
       std::map<std::string, std::pair<std::string, std::string>> change_map,
       std::vector<std::shared_ptr<Node>> *removed_nodes);
   std::vector<std::shared_ptr<framework::OpDesc>> OpDescs(uint size);
-  std::vector<std::shared_ptr<framework::OpDesc>> OpDescs();
   std::shared_ptr<framework::OpDesc> OpDescOfNode() { return op_desc_; }
   std::string Type() { return type_; }
 
  private:
-  void CanSplit(bool *split, bool spliting, int complex_count,
-                std::unordered_set<std::string> *complex_compute_set,
-                Node *pre_node);
-  void OpDescs(std::vector<std::shared_ptr<framework::OpDesc>> *op_desc,
-               Node *node, bool adding_thread, int thread_num);
   void OpDescs(uint size,
                std::vector<std::shared_ptr<framework::OpDesc>> *op_desc);
   void To(int index, std::shared_ptr<Node>);
diff --git a/src/operators/math/pooling.cpp b/src/operators/math/pooling.cpp
index 09bc2a6ae4a7b5f2876350375a333c457c72b68a..11bce0978f789f4f02b44fbb24fdd8bd1219257e 100644
--- a/src/operators/math/pooling.cpp
+++ b/src/operators/math/pooling.cpp
@@ -15,7 +15,7 @@ limitations under the License. */
 #ifdef POOL_OP
 
 #include "pooling.h"
-#include <common/types.h>
+#include "common/types.h"
 
 namespace paddle_mobile {
 namespace operators {
diff --git a/src/operators/pool_op.h b/src/operators/pool_op.h
index 9ad0bd3e3b95503c53bbeed6a8bca7fdb48ac23e..8dc99ae686390041b3c99c2df71d91ae9801a1f2 100644
--- a/src/operators/pool_op.h
+++ b/src/operators/pool_op.h
@@ -16,11 +16,12 @@ limitations under the License. */
 
 #pragma once
 
-#include <framework/operator.h>
-#include <operators/kernel/pool_kernel.h>
-#include <operators/op_param.h>
 #include <string>
 
+#include "framework/operator.h"
+#include "operators/kernel/pool_kernel.h"
+#include "operators/op_param.h"
+
 namespace paddle_mobile {
 namespace operators {
 using framework::AttributeMap;
diff --git a/src/operators/sigmoid_op.h b/src/operators/sigmoid_op.h
index 3757e2de168e75764b9d8d1d249fe3d8c87d817a..7cdeb41af1b20ddf05ac80d7de0962c4bfe8dff4 100644
--- a/src/operators/sigmoid_op.h
+++ b/src/operators/sigmoid_op.h
@@ -16,10 +16,11 @@ limitations under the License. */
 
 #pragma once
 
-#include <framework/operator.h>
-#include <operators/op_param.h>
 #include <string>
+
+#include "framework/operator.h"
 #include "operators/kernel/sigmoid_kernel.h"
+#include "operators/op_param.h"
 
 namespace paddle_mobile {
 namespace operators {
diff --git a/src/operators/softmax_op.h b/src/operators/softmax_op.h
index 1c764248cb72f84adb7665ced0a4375a3cd79624..5cac4d8a3394b07e978ba41e18fd7fbb7f4756d7 100644
--- a/src/operators/softmax_op.h
+++ b/src/operators/softmax_op.h
@@ -16,10 +16,11 @@ limitations under the License. */
 
 #pragma once
 
-#include <framework/operator.h>
-#include <operators/op_param.h>
 #include <string>
+
+#include "framework/operator.h"
 #include "operators/kernel/softmax_kernel.h"
+#include "operators/op_param.h"
 
 namespace paddle_mobile {
 namespace operators {
diff --git a/src/operators/transpose_op.cpp b/src/operators/transpose_op.cpp
index c8e16f9e4b42037eee84dbe5cd023b67e781f2a5..3abebc77ba6733bf1bfa73846ec1a077a494de2f 100644
--- a/src/operators/transpose_op.cpp
+++ b/src/operators/transpose_op.cpp
@@ -14,9 +14,10 @@ limitations under the License. */
 
 #ifdef TRANSPOSE_OP
 
-#include "operators/transpose_op.h"
-#include <common/enforce.h>
 #include <vector>
+
+#include "common/enforce.h"
+#include "operators/transpose_op.h"
 namespace paddle_mobile {
 namespace operators {
 
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
index 7bde05f70c85b9ea363e58ff483d45436fd7abd2..124e1c2d60d81fe0d6f19ffa1936f3d1d8e7eb16 100644
--- a/test/CMakeLists.txt
+++ b/test/CMakeLists.txt
@@ -122,13 +122,9 @@ else ()
     target_link_libraries(test-enforce paddle-mobile)
 
     # gen test - test if openmp works
-    ADD_EXECUTABLE(test-openmp common/test_openmp.cc test_helper.h test_include.h executor_for_test.h)
+    ADD_EXECUTABLE(test-openmp common/test_openmp.cpp test_helper.h test_include.h executor_for_test.h)
     target_link_libraries(test-openmp paddle-mobile)
 
-    # gen test
-    ADD_EXECUTABLE(test-yolo net/test_yolo.cpp test_helper.h  test_include.h executor_for_test.h)
-    target_link_libraries(test-yolo paddle-mobile)
-
     # gen test
     ADD_EXECUTABLE(test-mobilenetssd net/test_mobilenet+ssd.cpp test_helper.h  test_include.h executor_for_test.h)
     target_link_libraries(test-mobilenetssd paddle-mobile)
diff --git a/test/common/test_lib_size.cpp b/test/common/test_lib_size.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..805668f359f0e0959ea7122f25cdaa0ad2d7ec77
--- /dev/null
+++ b/test/common/test_lib_size.cpp
@@ -0,0 +1,21 @@
+/* Copyright (c) 2018 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. */
+
+//
+// Created by liuRuiLong on 2018/6/6.
+//
+
+#include "test_lib_size.h"
+
+static test_lib_size t;
diff --git a/test/common/test_lib_size.h b/test/common/test_lib_size.h
new file mode 100644
index 0000000000000000000000000000000000000000..ab5cd359b73f1c5c5df4176ab444e48c85dbd95b
--- /dev/null
+++ b/test/common/test_lib_size.h
@@ -0,0 +1,93 @@
+/* Copyright (c) 2018 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. */
+
+//
+// Created by liuRuiLong on 2018/6/6.
+//
+
+#ifndef PADDLE_MOBILE_TEST_LIB_SIZE_H
+#define PADDLE_MOBILE_TEST_LIB_SIZE_H
+
+#include <vector>
+//#include <list>
+//#include <tuple>
+//#include <typeinfo>
+//#include <mutex>
+//#include <initializer_list>
+//#include <map>
+//#include <string>
+//#include <unordered_map>
+//#include <unordered_set>
+//#include <algorithm>
+
+//#include <iostream>
+//#include <sstream>
+#include <memory>
+//#include <stdio.h>
+//#include <cstring>
+
+void foo() {
+  //  char *str = "1234";
+  //  char dst[10];
+  //  strcpy(dst, str);
+
+  //  std::cout << "12345" << std::endl;
+  std::vector<int> vec = {1, 2, 3, 4, 5};
+
+  //  std::find(vec.begin(), vec.end(), 1);
+  //  std::find(vec.begin(), vec.end(), 1);
+
+  //  std::list<int> l;
+  //  std::mutex mutex_;
+
+  //  std::map<int, float> m;
+  //  std::unordered_map<int, float> u_m;
+  //  std::unordered_set<int> u_s;
+  //  std::string ss = "12345";
+  //  printf("%f", ss.c_str());
+
+  //  std::initializer_list<int> init_list = {1, 2};
+  //  std::tuple<int, int> t = {1, 2};
+
+  //  std::tuple_element<I, std::tuple<ARGS...>>::type
+
+  //  std::tuple<>
+
+  //  int i;
+  //  int j;
+  //  if (typeid(i) == typeid(j)){
+  //    int z = 10;
+  //  }
+
+  std::shared_ptr<int> s1 = std::make_shared<int>();
+
+  //  std::stringstream ss;
+  //  ss << "12345";
+}
+
+class test_lib_size {
+ public:
+  test_lib_size() {}
+  //  std::shared_ptr<int> Test(){
+  //    std::vector<int> vec = {1, 2, 3};
+  //    std::shared_ptr<int> si = std::make_shared<int>();
+  //    return si;
+  //  }
+
+  //  void test(){
+  //    int i = 9;
+  //  }
+};
+
+#endif  // PADDLE_MOBILE_TEST_LIB_SIZE_H
diff --git a/test/common/test_openmp.cpp b/test/common/test_openmp.cpp
index 7393810a3cd5cafcd75e3964cc21d31036d83817..c92fc919936b067ad70f40e3f5b3acd725e845e2 100644
--- a/test/common/test_openmp.cpp
+++ b/test/common/test_openmp.cpp
@@ -12,15 +12,16 @@ 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 <omp.h>
+//#include <omp.h>
 #include <iostream>
 
 int main(void) {
   #pragma omp parallel num_threads(2)
   {
-    int thread_id = omp_get_thread_num();
-    int nthreads = omp_get_num_threads();
-    std::cout << "Hello, OMP " << thread_id << "/" << nthreads << "\n";
+    //        int thread_id = omp_get_thread_num();
+    //        int nthreads = omp_get_num_threads();
+    //        std::cout << "Hello, OMP " << thread_id << "/" << nthreads <<
+    //        "\n";
   }
   return 0;
 }