diff --git a/paddle/fluid/framework/CMakeLists.txt b/paddle/fluid/framework/CMakeLists.txt
index de19c7a0e773e31b6249f8342830170b4641989e..67073350d5a8aa3fdddf270f1b9e1f5be27d0eda 100644
--- a/paddle/fluid/framework/CMakeLists.txt
+++ b/paddle/fluid/framework/CMakeLists.txt
@@ -26,6 +26,7 @@ add_subdirectory(details)
 add_subdirectory(fleet)
 add_subdirectory(io)
 add_subdirectory(new_executor)
+add_subdirectory(paddle2cinn)
 #ddim lib
 proto_library(framework_proto SRCS framework.proto)
 proto_library(pass_desc_proto SRCS pass_desc.proto DEPS framework_proto)
diff --git a/paddle/fluid/framework/paddle2cinn/CMakeLists.txt b/paddle/fluid/framework/paddle2cinn/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8621c7363a09f1a1dff740e08cb57b2897aef8f5
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/CMakeLists.txt
@@ -0,0 +1,7 @@
+cc_library(cinn_cache_key SRCS cinn_cache_key.cc DEPS boost graph graph_helper lod_tensor proto_desc)
+cc_library(cinn_compiled_object SRCS cinn_compiled_object.cc DEPS feed_fetch_method graph lod_tensor proto_desc)
+cc_library(cinn_runner SRCS cinn_runner.cc DEPS cinn_cache_key cinn_compiled_object feed_fetch_method graph lod_tensor scope)
+
+cc_test(cinn_cache_key_test SRCS cinn_cache_key_test.cc DEPS cinn_cache_key)
+cc_test(cinn_runner_test SRCS cinn_runner_test.cc DEPS cinn_runner proto_desc)
+cc_test(cinn_compiled_object_test SRCS cinn_compiled_object_test.cc DEPS cinn_compiled_object)
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_cache_key.cc b/paddle/fluid/framework/paddle2cinn/cinn_cache_key.cc
new file mode 100644
index 0000000000000000000000000000000000000000..ac6c83be4fae3c944a285ccce99fe3285280ed09
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_cache_key.cc
@@ -0,0 +1,87 @@
+// Copyright (c) 2021 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 "paddle/fluid/framework/paddle2cinn/cinn_cache_key.h"
+
+#include <map>
+#include <string>
+
+#include "paddle/fluid/framework/ddim.h"
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/ir/graph_helper.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+CinnCacheKey::CinnCacheKey(
+    const ir::Graph& graph,
+    const std::map<std::string, const LoDTensor*>& feed_tensors) {
+  this->SetKey(graph, feed_tensors);
+}
+
+CinnCacheKey::CinnCacheKey(const ir::Graph& graph,
+                           const std::map<std::string, DDim>& feed_shapes) {
+  this->SetKey(graph, feed_shapes);
+}
+
+void CinnCacheKey::SetKey(
+    const ir::Graph& graph,
+    const std::map<std::string, const LoDTensor*>& feed_tensors) {
+  ProgramDesc program;
+  GraphToProgram(graph, &program);
+  program.Proto()->SerializeToString(&graph_serialize_str_);
+  for (const auto& name_tensor : feed_tensors) {
+    feed_shapes_[name_tensor.first] = name_tensor.second->dims();
+  }
+}
+
+void CinnCacheKey::SetKey(const ir::Graph& graph,
+                          const std::map<std::string, DDim>& feed_shapes) {
+  ProgramDesc program;
+  GraphToProgram(graph, &program);
+  program.Proto()->SerializeToString(&graph_serialize_str_);
+  feed_shapes_ = feed_shapes;
+}
+
+bool CinnCacheKey::operator!=(const CinnCacheKey& other) const {
+  return !this->operator==(other);
+}
+
+bool CinnCacheKey::operator==(const CinnCacheKey& other) const {
+  return graph_serialize_str_ == other.graph_serialize_str_ &&
+         feed_shapes_ == other.feed_shapes_;
+}
+
+size_t CinnCacheKey::Hash::hash_combine(size_t seed, size_t value) {
+  return seed ^ (value + 0x9e3779b9 + (seed << 6) + (seed >> 2));
+}
+
+size_t CinnCacheKey::Hash::operator()(const CinnCacheKey& key) const {
+  std::size_t ret = 0;
+
+  std::hash<std::string> string_hasher;
+  for (const auto& name_shape : key.feed_shapes_) {
+    ret = hash_combine(ret, string_hasher(name_shape.first));
+    ret = hash_combine(ret, string_hasher(name_shape.second.to_str()));
+  }
+
+  ret = hash_combine(ret, string_hasher(key.graph_serialize_str_));
+  return ret;
+}
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_cache_key.h b/paddle/fluid/framework/paddle2cinn/cinn_cache_key.h
new file mode 100644
index 0000000000000000000000000000000000000000..9627ae92aaba25fac9bae404d2ebfffba912db21
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_cache_key.h
@@ -0,0 +1,63 @@
+// Copyright (c) 2021 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.
+
+#pragma once
+
+#include <map>
+
+#include "paddle/fluid/framework/ddim.h"
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+// Class to store the keys for compiling CINN.
+//
+// CINN cannot handle changable shape now, so CinnRunner keeps a cache mapping
+// from CinnCacheKey to CinnCompiledObject.
+//
+// The CinnCacheKey contains a graph serialized string and the feeded tensor
+// shapes.
+class CinnCacheKey {
+ public:
+  CinnCacheKey(const ir::Graph& graph,
+               const std::map<std::string, const LoDTensor*>& feed_tensors);
+  CinnCacheKey(const ir::Graph& graph,
+               const std::map<std::string, DDim>& feed_shapes);
+
+  ~CinnCacheKey() {}
+
+  void SetKey(const ir::Graph& graph,
+              const std::map<std::string, const LoDTensor*>& feed_tensors);
+  void SetKey(const ir::Graph& graph,
+              const std::map<std::string, DDim>& feed_shapes);
+
+  bool operator==(const CinnCacheKey& other) const;
+  bool operator!=(const CinnCacheKey& other) const;
+
+  struct Hash {
+    static size_t hash_combine(size_t seed, size_t value);
+    size_t operator()(const CinnCacheKey& key) const;
+  };
+
+ private:
+  std::string graph_serialize_str_;
+  std::map<std::string, DDim> feed_shapes_;
+};
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_cache_key_test.cc b/paddle/fluid/framework/paddle2cinn/cinn_cache_key_test.cc
new file mode 100644
index 0000000000000000000000000000000000000000..a84ade26bfd1248cd44793fa0e1a12fdfea809ee
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_cache_key_test.cc
@@ -0,0 +1,101 @@
+// Copyright (c) 2021 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 <map>
+#include <unordered_set>
+
+#include "gtest/gtest.h"
+#include "paddle/fluid/framework/ddim.h"
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/paddle2cinn/cinn_cache_key.h"
+#include "paddle/fluid/framework/program_desc.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+TEST(CinnCacheKeyTest, TestAsUnorderedKey) {
+  std::unordered_set<CinnCacheKey, CinnCacheKey::Hash> test_set;
+
+  ProgramDesc empty_program;
+  ir::Graph empty_graph(empty_program);
+
+  ProgramDesc program;
+  auto *global_block = program.MutableBlock(0);
+  auto *x = global_block->Var("X");
+  x->SetType(proto::VarType::LOD_TENSOR);
+  ir::Graph graph(program);
+
+  LoDTensor tensor;
+  tensor.Resize({1, 2, 3});
+  const LoDTensor *tensor_pointer = &tensor;
+  std::map<std::string, const LoDTensor *> feed_tensors = {
+      {"X", tensor_pointer}};
+
+  DDim ddim = paddle::framework::make_ddim({1, 2, 3});
+  std::map<std::string, DDim> feed_shapes = {{"X", ddim}};
+
+  CinnCacheKey cache_key1(empty_graph, feed_tensors);
+  CinnCacheKey cache_key2(empty_graph, feed_shapes);
+  EXPECT_EQ(cache_key1, cache_key2);
+
+  CinnCacheKey cache_key3(graph, feed_shapes);
+  CinnCacheKey cache_key4(graph, feed_tensors);
+  EXPECT_EQ(cache_key3, cache_key4);
+
+  CinnCacheKey cache_key5(empty_graph,
+                          std::map<std::string, const LoDTensor *>());
+  CinnCacheKey cache_key6(empty_graph, std::map<std::string, DDim>());
+  EXPECT_EQ(cache_key5, cache_key6);
+
+  EXPECT_NE(cache_key1, cache_key3);
+  EXPECT_NE(cache_key4, cache_key2);
+
+  EXPECT_NE(cache_key3, cache_key5);
+  EXPECT_NE(cache_key6, cache_key4);
+
+  EXPECT_NE(cache_key5, cache_key1);
+  EXPECT_NE(cache_key2, cache_key6);
+
+  test_set.insert(cache_key1);
+  test_set.insert(cache_key2);
+  test_set.insert(cache_key3);
+  test_set.insert(cache_key4);
+  test_set.insert(cache_key5);
+  test_set.insert(cache_key6);
+  EXPECT_EQ(test_set.size(), 3U);
+
+  auto iter = test_set.find(cache_key1);
+  EXPECT_NE(iter, test_set.end());
+  test_set.erase(iter);
+  EXPECT_EQ(test_set.size(), 2U);
+  EXPECT_EQ(test_set.find(cache_key2), test_set.end());
+
+  iter = test_set.find(cache_key3);
+  EXPECT_NE(iter, test_set.end());
+  test_set.erase(iter);
+  EXPECT_EQ(test_set.size(), 1U);
+  EXPECT_EQ(test_set.find(cache_key4), test_set.end());
+
+  iter = test_set.find(cache_key5);
+  EXPECT_NE(iter, test_set.end());
+  test_set.erase(iter);
+  EXPECT_EQ(test_set.size(), 0U);
+  EXPECT_EQ(test_set.find(cache_key6), test_set.end());
+}
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_compiled_object.cc b/paddle/fluid/framework/paddle2cinn/cinn_compiled_object.cc
new file mode 100644
index 0000000000000000000000000000000000000000..a90494bafe9bb6fdefca6378ef03b9ee5fda1e62
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_compiled_object.cc
@@ -0,0 +1,50 @@
+// Copyright (c) 2021 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 "paddle/fluid/framework/paddle2cinn/cinn_compiled_object.h"
+
+#include <map>
+
+#include "paddle/fluid/framework/feed_fetch_type.h"
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/program_desc.h"
+#include "paddle/fluid/framework/scope.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+CinnCompiledObject::CinnCompiledObject() {
+  // TODO(zhhsplendid): complete this function after CINN interface is ready
+}
+CinnCompiledObject::~CinnCompiledObject() {
+  // TODO(zhhsplendid): complete this function after CINN interface is ready
+}
+
+void CinnCompiledObject::Compile(
+    const ir::Graph& graph,
+    std::map<std::string, const LoDTensor*>* feed_targets) {
+  // TODO(zhhsplendid): complete this function after CINN interface is ready
+}
+
+std::map<std::string, FetchType*> CinnCompiledObject::Run(
+    Scope* scope, std::map<std::string, const LoDTensor*>* feed_targets) {
+  // TODO(zhhsplendid): complete this function after CINN interface is ready
+  return std::map<std::string, FetchType*>();
+}
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_compiled_object.h b/paddle/fluid/framework/paddle2cinn/cinn_compiled_object.h
new file mode 100644
index 0000000000000000000000000000000000000000..21191d44345877dec2520d9ae848283f63017b27
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_compiled_object.h
@@ -0,0 +1,50 @@
+// Copyright (c) 2021 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.
+
+#pragma once
+
+#include <map>
+
+#include "paddle/fluid/framework/feed_fetch_type.h"
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/scope.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+// Class to store and call CINN complied object
+class CinnCompiledObject {
+ public:
+  CinnCompiledObject();
+  ~CinnCompiledObject();
+
+  // Compiles use CINN. CINN compilation needs model graph, input names, and
+  // input_shapes
+  void Compile(const ir::Graph& graph,
+               std::map<std::string, const LoDTensor*>* feed_targets);
+
+  // Feed LoDTensors to tun CINN compiled object and return fetched result
+  std::map<std::string, FetchType*> Run(
+      Scope* scope, std::map<std::string, const LoDTensor*>* feed_targets);
+
+  // Converts compiled object to Paddle Graph
+  // To be discussed
+  // ir::Graph ToGraph();
+};
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_compiled_object_test.cc b/paddle/fluid/framework/paddle2cinn/cinn_compiled_object_test.cc
new file mode 100644
index 0000000000000000000000000000000000000000..5a7861edf210c4d8cba5213d10ac6e2811e1d223
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_compiled_object_test.cc
@@ -0,0 +1,41 @@
+// Copyright (c) 2021 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 <map>
+
+#include "gtest/gtest.h"
+
+#include "paddle/fluid/framework/feed_fetch_type.h"
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/paddle2cinn/cinn_compiled_object.h"
+#include "paddle/fluid/framework/program_desc.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+TEST(CinnCompiledObjecctTest, TodoTest) {
+  ProgramDesc empty_program;
+  ir::Graph empty_graph(empty_program);
+  std::map<std::string, const LoDTensor*> empty_feed;
+  Scope empty_scope;
+
+  CinnCompiledObject compiled_obj;
+  compiled_obj.Compile(empty_graph, &empty_feed);
+  auto fetch = compiled_obj.Run(&empty_scope, &empty_feed);
+}
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_runner.cc b/paddle/fluid/framework/paddle2cinn/cinn_runner.cc
new file mode 100644
index 0000000000000000000000000000000000000000..de5af910c99add7f2947d9ea13c119dd60f6f3de
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_runner.cc
@@ -0,0 +1,46 @@
+// Copyright (c) 2021 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 "paddle/fluid/framework/paddle2cinn/cinn_runner.h"
+
+#include <map>
+
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/scope.h"
+#include "paddle/fluid/framework/tensor.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+using ir::Graph;
+
+std::map<std::string, FetchType*> CinnRunner::Run(
+    const Graph& graph, Scope* scope,
+    std::map<std::string, const LoDTensor*>* feed_targets) {
+  CinnCacheKey cur_key(graph, *feed_targets);
+  std::shared_ptr<CinnCompiledObject> obj_to_run;
+  if (cache_.find(cur_key) != cache_.end()) {
+    obj_to_run = cache_[cur_key];
+  } else {
+    obj_to_run = std::make_shared<CinnCompiledObject>();
+    obj_to_run->Compile(graph, feed_targets);
+    cache_[cur_key] = obj_to_run;
+  }
+  return obj_to_run->Run(scope, feed_targets);
+}
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_runner.h b/paddle/fluid/framework/paddle2cinn/cinn_runner.h
new file mode 100644
index 0000000000000000000000000000000000000000..5f63d64545ff75440e68ca88e5892d2e16291d26
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_runner.h
@@ -0,0 +1,55 @@
+// Copyright (c) 2021 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.
+
+#pragma once
+
+#include <map>
+#include <memory>
+#include <unordered_map>
+
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/paddle2cinn/cinn_cache_key.h"
+#include "paddle/fluid/framework/paddle2cinn/cinn_compiled_object.h"
+#include "paddle/fluid/framework/scope.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+// Entrance to run CINN.
+//
+// CINN cannot handle changable shape now, so CinnRunner keeps a cache mapping
+// from CinnCacheKey to CinnCompiledObject. If cache hits, we will re-use cache
+// stored CinnCompiledObject, otherwise we will compile again and put into
+// cache.
+class CinnRunner {
+ public:
+  CinnRunner() {}
+  ~CinnRunner() {}
+
+  // Feed LoDTensors to tun CINN compiled object and return fetched result
+  std::map<std::string, FetchType*> Run(
+      const ir::Graph& graph, Scope* scope,
+      std::map<std::string, const LoDTensor*>* feed_targets);
+
+ private:
+  std::unordered_map<CinnCacheKey, std::shared_ptr<CinnCompiledObject>,
+                     CinnCacheKey::Hash>
+      cache_;
+};
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle
diff --git a/paddle/fluid/framework/paddle2cinn/cinn_runner_test.cc b/paddle/fluid/framework/paddle2cinn/cinn_runner_test.cc
new file mode 100644
index 0000000000000000000000000000000000000000..88aca0bd66b375d14a511fa8baa91b05fb00da6f
--- /dev/null
+++ b/paddle/fluid/framework/paddle2cinn/cinn_runner_test.cc
@@ -0,0 +1,41 @@
+// Copyright (c) 2021 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 "gtest/gtest.h"
+
+#include "paddle/fluid/framework/ir/graph.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/paddle2cinn/cinn_runner.h"
+#include "paddle/fluid/framework/program_desc.h"
+#include "paddle/fluid/framework/scope.h"
+
+namespace paddle {
+namespace framework {
+namespace paddle2cinn {
+
+using ir::Graph;
+
+TEST(CinnRunnerTest, TodoTest) {
+  ProgramDesc empty_program;
+  Graph empty_graph(empty_program);
+  Scope empty_scope;
+  std::map<std::string, const LoDTensor*> empty_feed;
+
+  CinnRunner cinn_runner;
+  cinn_runner.Run(empty_graph, &empty_scope, &empty_feed);
+}
+
+}  // namespace paddle2cinn
+}  // namespace framework
+}  // namespace paddle