diff --git a/paddle/operators/elementwise_add_op.h b/paddle/operators/elementwise_add_op.h
index 59abbb57d1dcfbef6ead70e4afa9a3816d60d9b5..6478e1e0c2e1cfc8a1be5e8842113ec8ca33d762 100644
--- a/paddle/operators/elementwise_add_op.h
+++ b/paddle/operators/elementwise_add_op.h
@@ -81,23 +81,6 @@ struct ElementwiseAddGradFunctor {
   }
 };
 
-template <typename T>
-struct ElementwiseAddOneGradFunctor {
-  template <typename Device, typename X, typename Y, typename Z, typename dX,
-            typename dY, typename dZ>
-  void operator()(Device d, X x, Y y, Z z, dX dx, dY dy, dZ dz) {
-    auto dz_e = framework::EigenVector<T>::Flatten(*dz);
-    if (dx) {
-      auto dx_e = framework::EigenVector<T>::Flatten(*dx);
-      dx_e.device(d) = dz_e;
-    }
-    if (dy) {
-      auto dy_e = framework::EigenVector<T>::Flatten(*dy);
-      dy_e.device(d) = dz_e.sum();
-    }
-  }
-};
-
 template <typename T>
 struct ElementwiseAddBroadCastGradFunctor {
   template <typename Device, typename X, typename Y, typename Z, typename dX,
@@ -142,7 +125,6 @@ class ElementwiseAddGradKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& ctx) const override {
     ElementwiseGradCompute<DeviceContext, T, ElementwiseAddGradFunctor<T>,
-                           ElementwiseAddOneGradFunctor<T>,
                            ElementwiseAddBroadCastGradFunctor<T>,
                            ElementwiseAddBroadCast2GradFunctor<T>>(ctx);
   }
diff --git a/paddle/operators/elementwise_div_op.h b/paddle/operators/elementwise_div_op.h
index 875abd313ffc8fdf910d461922ff41f65ef276e7..7783875e24ef0d94e50f49b3598c93ff8bd5d73f 100644
--- a/paddle/operators/elementwise_div_op.h
+++ b/paddle/operators/elementwise_div_op.h
@@ -107,7 +107,6 @@ class ElementwiseDivGradKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& ctx) const override {
     ElementwiseGradCompute<DeviceContext, T, ElementwiseDivGradFunctor<T>,
-                           ElementwiseDivGradFunctor<T>,
                            ElementwiseDivBroadCastGradFunctor<T>,
                            ElementwiseDivBroadCast2GradFunctor<T>>(ctx);
   }
diff --git a/paddle/operators/elementwise_mul_op.h b/paddle/operators/elementwise_mul_op.h
index 3ee50207c07fa2b7ccf2c002903a4f055dbfb352..0e6559eacc0f90fec78397ce66372071370291e5 100644
--- a/paddle/operators/elementwise_mul_op.h
+++ b/paddle/operators/elementwise_mul_op.h
@@ -106,7 +106,6 @@ class ElementwiseMulGradKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& ctx) const override {
     ElementwiseGradCompute<DeviceContext, T, ElementwiseMulGradFunctor<T>,
-                           ElementwiseMulGradFunctor<T>,
                            ElementwiseMulBroadCastGradFunctor<T>,
                            ElementwiseMulBroadCast2GradFunctor<T>>(ctx);
   }
diff --git a/paddle/operators/elementwise_op_function.h b/paddle/operators/elementwise_op_function.h
index 560247cb108dce5432bfe66556b9e675a3accc27..0c75276b03140473cee4b57a4022ff3c6989ab4c 100644
--- a/paddle/operators/elementwise_op_function.h
+++ b/paddle/operators/elementwise_op_function.h
@@ -311,8 +311,7 @@ EIGEN_FUNCTOR(Mul, EIGEN_MUL);
 EIGEN_FUNCTOR(Div, EIGEN_DIV);
 
 template <typename DeviceContext, typename T, typename functor,
-          typename functor1, typename broadcastfunctor,
-          typename broadcast2functor>
+          typename broadcastfunctor, typename broadcast2functor>
 void ElementwiseGradCompute(const framework::ExecutionContext& ctx) {
   using Tensor = framework::Tensor;
 
diff --git a/paddle/operators/elementwise_sub_op.h b/paddle/operators/elementwise_sub_op.h
index 66edf8672d13086f883f0a2ad7ef5802317cc79a..347e92f87c7a46d3935f69864718d63661bef05f 100644
--- a/paddle/operators/elementwise_sub_op.h
+++ b/paddle/operators/elementwise_sub_op.h
@@ -43,23 +43,6 @@ struct ElementwiseSubGradFunctor {
   }
 };
 
-template <typename T>
-struct ElementwiseSubOneGradFunctor {
-  template <typename Device, typename X, typename Y, typename Z, typename dX,
-            typename dY, typename dZ>
-  void operator()(Device d, X x, Y y, Z z, dX dx, dY dy, dZ dz) {
-    auto dz_e = framework::EigenVector<T>::Flatten(*dz);
-    if (dx) {
-      auto dx_e = framework::EigenVector<T>::Flatten(*dx);
-      dx_e.device(d) = dz_e;
-    }
-    if (dy) {
-      auto dy_e = framework::EigenVector<T>::Flatten(*dy);
-      dy_e.device(d) = (-1.0) * dz_e.sum();
-    }
-  }
-};
-
 template <typename T>
 struct ElementwiseSubBroadCastGradFunctor {
   template <typename Device, typename X, typename Y, typename Z, typename dX,
@@ -106,7 +89,6 @@ class ElementwiseSubGradKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& ctx) const override {
     ElementwiseGradCompute<DeviceContext, T, ElementwiseSubGradFunctor<T>,
-                           ElementwiseSubOneGradFunctor<T>,
                            ElementwiseSubBroadCastGradFunctor<T>,
                            ElementwiseSubBroadCast2GradFunctor<T>>(ctx);
   }