diff --git a/cinn/common/cinn_value.cc b/cinn/common/cinn_value.cc
index e5b2c56279452..e87327bf95a34 100644
--- a/cinn/common/cinn_value.cc
+++ b/cinn/common/cinn_value.cc
@@ -117,7 +117,7 @@ void CINNValuePack::AddValue(const CINNValue &value) {
   values_.push_back(value);
 }
 void CINNValuePack::Clear() { values_.clear(); }
-const char *CINNValuePack::type_info() const { return "CINNValuePack"; }
+const char *CINNValuePack::type_info() const { return __type_info__; }
 
 CINNValue &CINNValue::operator=(int32_t value) {
   *this = CINNValue(value);
diff --git a/cinn/common/cinn_value.h b/cinn/common/cinn_value.h
index 6984cd6f7178b..4ba4c5a07d294 100644
--- a/cinn/common/cinn_value.h
+++ b/cinn/common/cinn_value.h
@@ -60,10 +60,11 @@ struct CINNValuePack : public common::Object {
  private:
   CINNValuePack() = default;
   std::vector<CINNValue> values_;
+  static constexpr char* __type_info__ = "CINNValuePack";
 };
 
 struct CINNValuePackShared : public Shared<CINNValuePack> {
-  CINNValuePackShared(CINNValuePack* ptr) : Shared<CINNValuePack>(ptr) {}
+  explicit CINNValuePackShared(CINNValuePack* ptr) : Shared<CINNValuePack>(ptr) {}
 
   CINNValue& operator[](int offset) { return (*operator->())[offset]; }
   const CINNValue& operator[](int offset) const { return (*operator->())[offset]; }
diff --git a/cinn/common/graph_utils.h b/cinn/common/graph_utils.h
index 28dabcde01cff..de4b3a71f4643 100644
--- a/cinn/common/graph_utils.h
+++ b/cinn/common/graph_utils.h
@@ -35,13 +35,14 @@ class GraphEdge : public Object {
 
   GraphNode* source() const { return source_; }
   GraphNode* sink() const { return sink_; }
-  const char* type_info() const override { return "graph_edge"; }
+  const char* type_info() const override { return __type_info__; }
 
  private:
   //! Source of this edge.
   GraphNode* source_{};
   //! End of this edge.
   GraphNode* sink_{};
+  static constexpr char* __type_info__ = "graph_edge";
 };
 
 struct GraphEdgeCompare {
diff --git a/cinn/frontend/syntax.h b/cinn/frontend/syntax.h
index b171d068d9013..98a4ebb41098f 100644
--- a/cinn/frontend/syntax.h
+++ b/cinn/frontend/syntax.h
@@ -68,7 +68,7 @@ struct _Instruction_ : public common::Object {
 
   const char* type_info() const override { return __type_info__; }
 
-  static constexpr const char* __type_info__ = "cinn_frontend_instruction";
+  static constexpr char* __type_info__ = "cinn_frontend_instruction";
 };
 
 struct Instruction : public common::Shared<_Instruction_> {
diff --git a/cinn/hlir/framework/op_strategy.h b/cinn/hlir/framework/op_strategy.h
index 403a41bb6325c..08a2c9c3fb5c0 100644
--- a/cinn/hlir/framework/op_strategy.h
+++ b/cinn/hlir/framework/op_strategy.h
@@ -52,17 +52,17 @@ class OpImpl : public common::Object {
     return nullptr;
   }
 
-  const char* type_info() const override { return _type_key; }
+  const char* type_info() const override { return __type_info__; }
 
  private:
-  static constexpr char* _type_key = "OpImplementation";
+  static constexpr char* __type_info__ = "OpImplementation";
 };
 
 //! Specialized implementations for operators under certain conditions.
 class OpSpec : public common::Object {
  public:
   //! List of implementations.
-  std::vector<OpImpl*> implementations;
+  std::vector<std::shared_ptr<OpImpl>> implementations;
 
   /** \brief Condition to enable the specialization.
    *    Could be undefined to represent generic case.
@@ -71,10 +71,10 @@ class OpSpec : public common::Object {
    */
   std::string condition;
 
-  const char* type_info() const override { return _type_key; }
+  const char* type_info() const override { return __type_info__; }
 
   void AddImpl(CINNCompute fcompute, CINNSchedule fschedule, std::string name, int plevel) {
-    auto n       = make_shared<OpImpl>();
+    auto n       = std::make_shared<OpImpl>();
     n->fcompute  = fcompute;
     n->fschedule = fschedule;
     n->name      = std::move(name);
@@ -83,15 +83,15 @@ class OpSpec : public common::Object {
   }
 
  private:
-  static constexpr char* _type_key = "OpSpecialization";
+  static constexpr char* __type_info__ = "OpSpecialization";
 };
 
 //! Operator strategy class.
 class OpStrategy : public common::Object {
  public:
-  const char* type_info() const override { return "CINNOpStrategy"; }
+  const char* type_info() const override { return __type_info__; }
   //! List of operator specializations.
-  std::vector<OpSpec*> specializations;
+  std::vector<std::shared_ptr<OpSpec>> specializations;
 
   /**
    * \brief Add an implementation.
@@ -103,21 +103,40 @@ class OpStrategy : public common::Object {
   void AddImpl(CINNCompute fcompute, CINNSchedule fschedule, std::string name, int plevel) {
     //! TODO(haozech) : here curr_cond should get the condition from outside.
     //! Expected : auto curr_cond = SpecializedCondition::Current();
-    std::string curr_cond = "current_condition";
-    OpSpec* op_spec;
-    for (OpSpec* op_spec : specializations) {
-      if (op_spec->condition == curr_cond) {
+    std::string curr_condition = "default";
+    for (auto op_spec : specializations) {
+      if (op_spec->condition == curr_condition) {
         op_spec->AddImpl(fcompute, fschedule, std::move(name), plevel);
         return;
       }
     }
-    OpSpec* n    = make_shared<OpSpec>();
-    n->condition = curr_cond;
+    std::shared_ptr<OpSpec> n = std::make_shared<OpSpec>();
+    n->condition              = curr_condition;
     n->AddImpl(fcompute, fschedule, std::move(name), plevel);
     this->specializations.push_back(n);
   }
+
+ private:
+  static constexpr char* __type_info__ = "OpStrategy";
 };
 
+std::shared_ptr<OpImpl> SelectImpl(std::shared_ptr<OpStrategy> strategy) {
+  //! should get the host info from global environment.
+  std::string curr_condition  = "default";
+  std::shared_ptr<OpImpl> res = nullptr;
+  for (auto spec : strategy->specializations) {
+    if (spec->condition == "default") {
+      for (auto i : spec->implementations) {
+        if (!res || res->plevel < i->plevel) {
+          res = i;
+        }
+      }
+    }
+  }
+  CHECK(res) << "There is no available strategy implementation! SelectImpl failed!";
+  return res;
+}
+
 }  // namespace framework
 }  // namespace hlir
 }  // namespace cinn
diff --git a/cinn/hlir/framework/op_test.cc b/cinn/hlir/framework/op_test.cc
index 2cdf427cdfde0..d1684260c9192 100644
--- a/cinn/hlir/framework/op_test.cc
+++ b/cinn/hlir/framework/op_test.cc
@@ -2,15 +2,23 @@
 
 #include <gtest/gtest.h>
 
+#include <functional>
+
 #include <string>
 
 #include "cinn/cinn.h"
+
+#include "cinn/hlir/framework/node.h"
 #include "cinn/hlir/framework/op_strategy.h"
 #include "cinn/hlir/pe/broadcast.h"
 
 namespace cinn {
 namespace hlir {
 namespace framework {
+using StrategyFunction = std::function<std::shared_ptr<OpStrategy>(
+    const NodeAttr, const std::vector<ir::Tensor>, common::Type, const common::Target)>;
+
+using CCompute = std::function<std::shared_ptr<ir::Tensor>(const std::vector<ir::Tensor>)>;
 
 CINN_REGISTER_OP(add)
     .describe("test of op Add")
@@ -19,30 +27,68 @@ CINN_REGISTER_OP(add)
     .set_attr<std::string>("nick_name", "plus")
     .set_support_level(4);
 
-common::Shared<OpStrategy> GetStrategyTest() {
-  ir::PackedFunc::body_t body = [](ir::Args args, ir::RetValue* ret) {
-    Expr a = args[0];
-    Expr b = args[1];
-    (*ret) = Expr(pe::Add(a.as_tensor_ref(), b.as_tensor_ref(), "C").get());
+std::shared_ptr<OpStrategy> StrategyTest(const NodeAttr &attr,
+                                         const std::vector<ir::Tensor> &inputs,
+                                         common::Type out_type,
+                                         const common::Target &target) {
+  ir::PackedFunc::body_t compute_body = [](ir::Args args, ir::RetValue *ret) {
+    common::CINNValuePackShared a = args[0];
+    ir::Expr A                    = a[0];
+    ir::Expr B                    = a[1];
+    *ret                          = common::CINNValuePack::Make(
+        {common::CINNValue(ir::Expr(pe::Add(A.as_tensor_ref(), B.as_tensor_ref(), "C").get()))});
   };
-  ir::PackedFunc fcompute(body);
-  // TODO(haozech): fschedule should be an instance of pe::schedule...
-  ir::PackedFunc fschedule;
-  common::Shared<OpStrategy> strategy(make_shared<OpStrategy>());
-  //! To build more complex strategy, we can add more than 1
-  //! implementations to one Opstrategy, with different plevel.
-  strategy->AddImpl(fcompute, fschedule, "test.strategy", 10);
+  ir::PackedFunc fcompute(compute_body);
+
+  ir::PackedFunc::body_t schedule_body = [](ir::Args args, ir::RetValue *ret) {
+    common::CINNValuePackShared a = args[0];
+    ir::Expr A                    = a[0];
+    A.as_tensor_ref()->stage()->Vectorize(1, 16);
+    A.as_tensor_ref()->stage()->Unroll(1);
+    *ret = common::CINNValuePack::Make({common::CINNValue(A)});
+  };
+  ir::PackedFunc fschedule(schedule_body);
+
+  std::shared_ptr<OpStrategy> strategy = std::make_shared<OpStrategy>();
+
+  if (target.arch == common::Target::Arch ::X86) {
+    strategy->AddImpl(fcompute, fschedule, "test.strategy.x86", 10);
+  } else {
+    strategy->AddImpl(fcompute, fschedule, "test.strategy.else", 10);
+  }
+  strategy->AddImpl(fcompute, fschedule, "test.strategy.lowPlevel.x86", 5);
   return strategy;
 }
 
 TEST(Operator, GetAttr) {
-  auto add           = Operator::Get("add");
-  auto test_strategy = GetStrategyTest();
-  Operator temp      = *add;
-  temp.set_attr<common::Shared<OpStrategy>>("CINNStrategy", test_strategy);
+  auto add      = Operator::Get("add");
+  Operator temp = *add;
+  temp.set_attr<StrategyFunction>("CINNStrategy", StrategyTest);
   auto nick     = Operator::GetAttr<std::string>("nick_name");
-  auto strategy = Operator::GetAttr<common::Shared<OpStrategy>>("CINNStrategy");
-  ASSERT_EQ(strategy[add]->specializations[0]->implementations[0]->name, "test.strategy");
+  auto strategy = Operator::GetAttr<StrategyFunction>("CINNStrategy");
+
+  Expr M(100), N(32);
+  Placeholder<float> A("A", {M, N});
+  Placeholder<float> B("B", {M, N});
+
+  NodeAttr attr;
+  std::vector<ir::Tensor> inputs{A, B};
+  common::Type type;
+  common::Target target;
+  target.arch = common::Target::Arch::X86;
+  auto impl   = SelectImpl(strategy[add](attr, inputs, type, target));
+
+  common::CINNValuePackShared cinn_input = common::CINNValuePack::Make({common::CINNValue(A), common::CINNValue(B)});
+  common::CINNValuePackShared C          = impl->fcompute(cinn_input);
+  C                                      = impl->fschedule(C);
+  for (int i = 0; i < C.get()->size(); i++) {
+    ir::Expr temp = C[i];
+    inputs.push_back(temp.as_tensor_ref());
+  }
+  auto func = Lower("add1", inputs);
+  LOG(INFO) << "Test Strategy Codegen:\n" << func;
+
+  ASSERT_EQ(impl->name, "test.strategy.x86");
   ASSERT_EQ(add->description, "test of op Add");
   ASSERT_EQ(nick[add], "plus");
 }
diff --git a/cinn/hlir/framework/schedule.h b/cinn/hlir/framework/schedule.h
index ca675bf156c2c..f3e0085dc33cf 100644
--- a/cinn/hlir/framework/schedule.h
+++ b/cinn/hlir/framework/schedule.h
@@ -14,7 +14,7 @@ namespace framework {
  */
 class Schedule : public common::Object {
  public:
-  const char* type_info() const override { return "CINNSchedule"; }
+  const char* type_info() const override { return __type_info__; }
 
   /**
    * \brief Get the stage corresponds to the op
@@ -36,6 +36,9 @@ class Schedule : public common::Object {
 
   //! map of original operation to the stages
   std::unordered_map<std::string, ir::Tensor> stage_map;
+
+ private:
+  static constexpr char* __type_info__ = "CINNSchedule";
 };
 }  // namespace framework
 }  // namespace hlir
diff --git a/cinn/ir/node.h b/cinn/ir/node.h
index 5110095302e49..ac6a6f4cd7bce 100644
--- a/cinn/ir/node.h
+++ b/cinn/ir/node.h
@@ -141,7 +141,7 @@ class IrNode : public common::Object {
   const char* type_info() const override { return __type_info__; }
 
  protected:
-  constexpr static const char* __type_info__ = "IRNode";
+  static constexpr char* __type_info__ = "IRNode";
   Type type_;
 };
 
diff --git a/cinn/poly/stage.h b/cinn/poly/stage.h
index 86966c04f0304..1e473ee1c97b1 100644
--- a/cinn/poly/stage.h
+++ b/cinn/poly/stage.h
@@ -219,7 +219,7 @@ class Stage : public Object {
   //! Get the statements.
   std::vector<std::string> input_statements() const;
 
-  virtual const char* type_info() const { return "Status"; }
+  virtual const char* type_info() const { return __type_info__; }
 
   inline const ir::VectorizeInfo& vectorize_info() const { return vectorize_info_; }
   inline const std::set<int>& unroll_info() const { return unroll_info_; }
@@ -280,6 +280,8 @@ class Stage : public Object {
 
   std::set<int> locked_axis_;
 
+  static constexpr char* __type_info__ = "Status";
+
   friend isl_map* __isl_give GatherAccesses(Stage* stage, const std::string& tensor_name);
   friend class PolyGroupScheduler;
 };