diff --git a/paddle/phi/kernels/xpu/adam_kernel.cc b/paddle/phi/kernels/xpu/adam_kernel.cc
index 8f55a1a03af4b6..7bddb0b44bc365 100644
--- a/paddle/phi/kernels/xpu/adam_kernel.cc
+++ b/paddle/phi/kernels/xpu/adam_kernel.cc
@@ -78,7 +78,7 @@ void AdamDenseKernel(
   const float* beta1_const_pow_ptr = nullptr;
   DenseTensor xpu_beta1_pow;
   if (beta1_pow.place() == CPUPlace()) {
-    phi::Copy(dev_ctx, beta1_pow, dev_ctx.GetPlace(), false, &xpu_beta1_pow);
+    Copy(dev_ctx, beta1_pow, dev_ctx.GetPlace(), false, &xpu_beta1_pow);
     if (xpu_beta1_pow.dtype() == DataType::FLOAT16)
       funcs::GetDataPointer<Context, float>(
           xpu_beta1_pow, &beta1_pow_ptr, dev_ctx, &RAII_GUARD);
@@ -96,7 +96,7 @@ void AdamDenseKernel(
   const float* beta2_const_pow_ptr = nullptr;
   DenseTensor xpu_beta2_pow;
   if (beta2_pow.place() == CPUPlace()) {
-    phi::Copy(dev_ctx, beta2_pow, dev_ctx.GetPlace(), false, &xpu_beta2_pow);
+    Copy(dev_ctx, beta2_pow, dev_ctx.GetPlace(), false, &xpu_beta2_pow);
     if (xpu_beta2_pow.dtype() == DataType::FLOAT16)
       funcs::GetDataPointer<Context, float>(
           xpu_beta2_pow, &beta2_pow_ptr, dev_ctx, &RAII_GUARD);
@@ -145,12 +145,12 @@ void AdamDenseKernel(
 
   if (skip_update_) {
     VLOG(4) << "Adam skip update";
-    phi::Copy(dev_ctx, param, dev_ctx.GetPlace(), false, param_out);
-    phi::Copy(dev_ctx, moment1, dev_ctx.GetPlace(), false, moment1_out);
-    phi::Copy(dev_ctx, moment2, dev_ctx.GetPlace(), false, moment2_out);
+    Copy(dev_ctx, param, dev_ctx.GetPlace(), false, param_out);
+    Copy(dev_ctx, moment1, dev_ctx.GetPlace(), false, moment1_out);
+    Copy(dev_ctx, moment2, dev_ctx.GetPlace(), false, moment2_out);
     if (!use_global_beta_pow) {
-      phi::Copy(dev_ctx, beta1_pow, beta1_pow.place(), false, beta1_pow_out);
-      phi::Copy(dev_ctx, beta2_pow, beta2_pow.place(), false, beta2_pow_out);
+      Copy(dev_ctx, beta1_pow, beta1_pow.place(), false, beta1_pow_out);
+      Copy(dev_ctx, beta2_pow, beta2_pow.place(), false, beta2_pow_out);
     }
     return;
   }
@@ -305,7 +305,7 @@ void MergedAdamKernel(
   DenseTensor lr_host;
   lr_host.Resize(learning_rate[0]->dims());
   dev_ctx.template HostAlloc<float>(&lr_host);
-  phi::Copy(dev_ctx, *learning_rate[0], CPUPlace(), false, &lr_host);
+  Copy(dev_ctx, *learning_rate[0], CPUPlace(), false, &lr_host);
   float lr_ = *(lr_host.template data<float>());
 
   float beta1_pow_data;
@@ -315,7 +315,7 @@ void MergedAdamKernel(
     DenseTensor beta1_pow_host;
     beta1_pow_host.Resize(beta1_pow[0]->dims());
     dev_ctx.template HostAlloc<float>(&beta1_pow_host);
-    phi::Copy(dev_ctx, *beta1_pow[0], CPUPlace(), false, &beta1_pow_host);
+    Copy(dev_ctx, *beta1_pow[0], CPUPlace(), false, &beta1_pow_host);
     beta1_pow_data = *(beta1_pow_host.template data<float>());
   }
 
@@ -326,7 +326,7 @@ void MergedAdamKernel(
     DenseTensor beta2_pow_host;
     beta2_pow_host.Resize(beta2_pow[0]->dims());
     dev_ctx.template HostAlloc<float>(&beta2_pow_host);
-    phi::Copy(dev_ctx, *beta2_pow[0], CPUPlace(), false, &beta2_pow_host);
+    Copy(dev_ctx, *beta2_pow[0], CPUPlace(), false, &beta2_pow_host);
     beta2_pow_data = *(beta2_pow_host.template data<float>());
   }
 
@@ -450,9 +450,9 @@ void MergedAdamKernel(
 
   // update param, moment1, moment2
   for (int i = 0; i < param_num; i++) {
-    phi::Copy(dev_ctx, *param[i], dev_ctx.GetPlace(), false, param_out[i]);
-    phi::Copy(dev_ctx, *moment1[i], dev_ctx.GetPlace(), false, moment1_out[i]);
-    phi::Copy(dev_ctx, *moment2[i], dev_ctx.GetPlace(), false, moment2_out[i]);
+    Copy(dev_ctx, *param[i], dev_ctx.GetPlace(), false, param_out[i]);
+    Copy(dev_ctx, *moment1[i], dev_ctx.GetPlace(), false, moment1_out[i]);
+    Copy(dev_ctx, *moment2[i], dev_ctx.GetPlace(), false, moment2_out[i]);
   }
 
   if (!use_global_beta_pow) {
diff --git a/paddle/phi/kernels/xpu/adamw_kernel.cc b/paddle/phi/kernels/xpu/adamw_kernel.cc
index bc9a3adaca3176..e7074fcee5dfa2 100644
--- a/paddle/phi/kernels/xpu/adamw_kernel.cc
+++ b/paddle/phi/kernels/xpu/adamw_kernel.cc
@@ -100,12 +100,12 @@ void AdamwDenseKernelKL3(const Context& dev_ctx,
   // skip_update=true, just copy input to output
   if (skip_update_) {
     VLOG(4) << "Adamw skip update";
-    phi::Copy(dev_ctx, param, dev_ctx.GetPlace(), false, param_out);
-    phi::Copy(dev_ctx, moment1, dev_ctx.GetPlace(), false, moment1_out);
-    phi::Copy(dev_ctx, moment2, dev_ctx.GetPlace(), false, moment2_out);
+    Copy(dev_ctx, param, dev_ctx.GetPlace(), false, param_out);
+    Copy(dev_ctx, moment1, dev_ctx.GetPlace(), false, moment1_out);
+    Copy(dev_ctx, moment2, dev_ctx.GetPlace(), false, moment2_out);
     if (!use_global_beta_pow) {
-      phi::Copy(dev_ctx, beta1_pow, beta1_pow.place(), false, beta1_pow_out);
-      phi::Copy(dev_ctx, beta2_pow, beta2_pow.place(), false, beta2_pow_out);
+      Copy(dev_ctx, beta1_pow, beta1_pow.place(), false, beta1_pow_out);
+      Copy(dev_ctx, beta2_pow, beta2_pow.place(), false, beta2_pow_out);
     }
     return;
   }
@@ -646,12 +646,12 @@ void AdamwDenseKernel(
   }
   if (skip_update_) {
     VLOG(4) << "Adamw skip update";
-    phi::Copy(dev_ctx, param, dev_ctx.GetPlace(), false, param_out);
-    phi::Copy(dev_ctx, moment1, dev_ctx.GetPlace(), false, moment1_out);
-    phi::Copy(dev_ctx, moment2, dev_ctx.GetPlace(), false, moment2_out);
+    Copy(dev_ctx, param, dev_ctx.GetPlace(), false, param_out);
+    Copy(dev_ctx, moment1, dev_ctx.GetPlace(), false, moment1_out);
+    Copy(dev_ctx, moment2, dev_ctx.GetPlace(), false, moment2_out);
     if (!use_global_beta_pow) {
-      phi::Copy(dev_ctx, beta1_pow, beta1_pow.place(), false, beta1_pow_out);
-      phi::Copy(dev_ctx, beta2_pow, beta2_pow.place(), false, beta2_pow_out);
+      Copy(dev_ctx, beta1_pow, beta1_pow.place(), false, beta1_pow_out);
+      Copy(dev_ctx, beta2_pow, beta2_pow.place(), false, beta2_pow_out);
     }
     return;
   }
@@ -665,8 +665,8 @@ void AdamwDenseKernel(
   DenseTensor xpu_beta1_pow;
   DenseTensor xpu_beta2_pow;
   if (beta1_pow.place() == CPUPlace() && beta2_pow.place() == CPUPlace()) {
-    phi::Copy(dev_ctx, beta1_pow, dev_ctx.GetPlace(), false, &xpu_beta1_pow);
-    phi::Copy(dev_ctx, beta2_pow, dev_ctx.GetPlace(), false, &xpu_beta2_pow);
+    Copy(dev_ctx, beta1_pow, dev_ctx.GetPlace(), false, &xpu_beta1_pow);
+    Copy(dev_ctx, beta2_pow, dev_ctx.GetPlace(), false, &xpu_beta2_pow);
     dev_ctx.Wait();
     beta1_pow_ptr = xpu_beta1_pow.template data<float>();
     beta2_pow_ptr = xpu_beta2_pow.template data<float>();
diff --git a/paddle/phi/kernels/xpu/argsort_grad_kernel.cc b/paddle/phi/kernels/xpu/argsort_grad_kernel.cc
index c71a266fbda476..33b90aa16c791e 100644
--- a/paddle/phi/kernels/xpu/argsort_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/argsort_grad_kernel.cc
@@ -43,7 +43,7 @@ void ArgsortGradKernel(const Context& dev_ctx,
   if (out_grad.numel() == 0) return;
 
   if (rank == 0) {
-    phi::Copy<Context>(dev_ctx, out_grad, dev_ctx.GetPlace(), false, in_grad);
+    Copy<Context>(dev_ctx, out_grad, dev_ctx.GetPlace(), false, in_grad);
     return;
   }
 
diff --git a/paddle/phi/kernels/xpu/argsort_kernel.cc b/paddle/phi/kernels/xpu/argsort_kernel.cc
index 23b53712f7abc2..3806034eacc1bb 100644
--- a/paddle/phi/kernels/xpu/argsort_kernel.cc
+++ b/paddle/phi/kernels/xpu/argsort_kernel.cc
@@ -117,7 +117,7 @@ void ArgsortKernel(const Context& dev_ctx,
   auto indices_data = dev_ctx.template Alloc<int64_t>(indices);
 
   if (rank == 0) {
-    phi::Copy<Context>(dev_ctx, input, dev_ctx.GetPlace(), false, output);
+    Copy<Context>(dev_ctx, input, dev_ctx.GetPlace(), false, output);
     funcs::set_constant(dev_ctx, indices, static_cast<int64_t>(0));
     return;
   }
diff --git a/paddle/phi/kernels/xpu/c_softmax_with_cross_entropy_grad_kernel.cc b/paddle/phi/kernels/xpu/c_softmax_with_cross_entropy_grad_kernel.cc
index 92e18e773bbb06..fde40ed1c668ff 100644
--- a/paddle/phi/kernels/xpu/c_softmax_with_cross_entropy_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/c_softmax_with_cross_entropy_grad_kernel.cc
@@ -44,7 +44,7 @@ void CSoftmaxWithCrossEntropyGradKernel(const Context& dev_ctx,
   const DenseTensor* softmax = &softmax_in;
 
   if (logit_grad != softmax) {
-    phi::Copy(dev_ctx, *softmax, dev_ctx.GetPlace(), false, logit_grad);
+    Copy(dev_ctx, *softmax, dev_ctx.GetPlace(), false, logit_grad);
   }
   const auto softmax_dims = softmax->dims();
   const int axis = softmax_dims.size() - 1;
diff --git a/paddle/phi/kernels/xpu/cast_kernel.cc b/paddle/phi/kernels/xpu/cast_kernel.cc
index c65d12656ef698..58b396b1f2dde6 100644
--- a/paddle/phi/kernels/xpu/cast_kernel.cc
+++ b/paddle/phi/kernels/xpu/cast_kernel.cc
@@ -98,7 +98,7 @@ void CastKernel(const Context& dev_ctx,
       return;
     }
     if (!out->IsSharedWith(x)) {
-      phi::Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+      Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     }
     return;
   }
@@ -167,7 +167,7 @@ void CastKernel<phi::complex64, XPUContext>(const XPUContext& dev_ctx,
       return;
     }
     if (!out->IsSharedWith(x)) {
-      phi::Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+      Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     }
     return;
   }
diff --git a/paddle/phi/kernels/xpu/eig_kernel.cc b/paddle/phi/kernels/xpu/eig_kernel.cc
index 988ddba293ed63..d2c82a29327fe7 100644
--- a/paddle/phi/kernels/xpu/eig_kernel.cc
+++ b/paddle/phi/kernels/xpu/eig_kernel.cc
@@ -108,8 +108,8 @@ void EigKernel(const Context& dev_ctx,
   }
 
   // copy result from cpu to xpu tensor
-  phi::Copy(dev_ctx, out_w_cpu, phi::XPUPlace(), false, out_w);
-  phi::Copy(dev_ctx, out_v_cpu, phi::XPUPlace(), false, out_v);
+  Copy(dev_ctx, out_w_cpu, phi::XPUPlace(), false, out_w);
+  Copy(dev_ctx, out_v_cpu, phi::XPUPlace(), false, out_v);
 }
 
 }  // namespace phi
diff --git a/paddle/phi/kernels/xpu/embedding_grad_kernel.cc b/paddle/phi/kernels/xpu/embedding_grad_kernel.cc
index 1fb73d692db47f..7872412888245f 100644
--- a/paddle/phi/kernels/xpu/embedding_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/embedding_grad_kernel.cc
@@ -88,7 +88,7 @@ void EmbeddingSparseGradKernel(const Context& dev_ctx,
   ids_cpu.Resize(input.dims());
   dev_ctx.HostAlloc(&ids_cpu, input.dtype(), input.numel() * sizeof(int64_t));
   if (input.dtype() == phi::DataType::INT64) {
-    phi::Copy(dev_ctx, input, CPUPlace(), false, &ids_cpu);
+    Copy(dev_ctx, input, CPUPlace(), false, &ids_cpu);
 
     ids = CopyIdsToVector<int64_t, int64_t>(ids_cpu);
 
diff --git a/paddle/phi/kernels/xpu/flip_kernel.cc b/paddle/phi/kernels/xpu/flip_kernel.cc
index 9afe3e5042dd0e..5f20721cd26a69 100644
--- a/paddle/phi/kernels/xpu/flip_kernel.cc
+++ b/paddle/phi/kernels/xpu/flip_kernel.cc
@@ -37,7 +37,7 @@ void FlipKernel(const Context& dev_ctx,
     return;
   }
   if (formatted_axis.size() == 0) {
-    phi::Copy<Context>(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+    Copy<Context>(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     return;
   }
   std::vector<int64_t> x_shape = common::vectorize(x.dims());
diff --git a/paddle/phi/kernels/xpu/index_add_grad_kernel.cc b/paddle/phi/kernels/xpu/index_add_grad_kernel.cc
index 0e52d62ee884e8..8e4d44bb5dc6c7 100644
--- a/paddle/phi/kernels/xpu/index_add_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/index_add_grad_kernel.cc
@@ -47,7 +47,7 @@ void IndexAddGradKernel(const Context& dev_ctx,
   }
 
   if (x_grad) {
-    phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
   }
   if (add_value_grad) {
     phi::IndexSelectKernel<T, Context>(
diff --git a/paddle/phi/kernels/xpu/index_add_kernel.cc b/paddle/phi/kernels/xpu/index_add_kernel.cc
index 78f30bb2e6223f..7538e2f736d866 100644
--- a/paddle/phi/kernels/xpu/index_add_kernel.cc
+++ b/paddle/phi/kernels/xpu/index_add_kernel.cc
@@ -43,7 +43,7 @@ void IndexAddKernel(const Context& dev_ctx,
     return;
   }
   if (index.numel() == 0) {
-    phi::Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+    Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     return;
   }
 
diff --git a/paddle/phi/kernels/xpu/index_elementwise_put_grad_kernel.cc b/paddle/phi/kernels/xpu/index_elementwise_put_grad_kernel.cc
index 0ef76bb3cccc43..95acf5515d3d18 100644
--- a/paddle/phi/kernels/xpu/index_elementwise_put_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/index_elementwise_put_grad_kernel.cc
@@ -144,7 +144,7 @@ void LaunchIndexElementwisePutWithTensorGradXPUKernel(
     DenseTensor* value_grad,
     DenseTensor* x_grad) {
   if (x_grad && !value_grad) {
-    phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
 
     XPUIndexElementwisePutGradKernel<T, Context, int64_t>(dev_ctx,
                                                           out_grad,
@@ -158,7 +158,7 @@ void LaunchIndexElementwisePutWithTensorGradXPUKernel(
                                                           value_grad);
   } else if (value_grad) {
     if (x_grad) {
-      phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+      Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     }
     if (value_grad->numel() == 1) {
       DenseTensor tmp_value_grad(value_grad->dtype());
@@ -248,7 +248,7 @@ void LaunchIndexElementwisePutGradXPUKernel(
     const int64_t slice_offset,
     DenseTensor* x_grad) {
   if (x_grad) {
-    phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
 
     XPUIndexElementwisePutGradKernel<T, Context, int64_t>(dev_ctx,
                                                           out_grad,
@@ -288,7 +288,7 @@ void IndexElementwisePutGradKernel(
   std::vector<DenseTensor> tmp_args;
   if (indices.empty()) {
     if (x_grad) {
-      phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+      Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     }
     return;
   }
@@ -330,7 +330,7 @@ void IndexElementwisePutWithTensorGradKernel(
   std::vector<DenseTensor> tmp_args;
   if (indices.empty()) {
     if (x_grad) {
-      phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+      Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     }
     if (value_grad) {
       FullKernel<T, Context>(dev_ctx,
diff --git a/paddle/phi/kernels/xpu/index_elementwise_put_kernel.cc b/paddle/phi/kernels/xpu/index_elementwise_put_kernel.cc
index 4973de181a0e24..220aa480d0f0c3 100644
--- a/paddle/phi/kernels/xpu/index_elementwise_put_kernel.cc
+++ b/paddle/phi/kernels/xpu/index_elementwise_put_kernel.cc
@@ -40,7 +40,7 @@ void XPUIndexElementwisePutWithTensorKernel(
     is_same_place = (input.place() == output->place());
   }
   if (!is_initialized || !is_same_place) {
-    phi::Copy(dev_ctx, input, dev_ctx.GetPlace(), false, output);
+    Copy(dev_ctx, input, dev_ctx.GetPlace(), false, output);
   }
 
   int64_t num_indices = 0;
@@ -143,7 +143,7 @@ void XPUIndexElementwisePutKernel(const Context& dev_ctx,
     is_same_place = (input.place() == output->place());
   }
   if (!is_initialized || !is_same_place) {
-    phi::Copy(dev_ctx, input, dev_ctx.GetPlace(), false, output);
+    Copy(dev_ctx, input, dev_ctx.GetPlace(), false, output);
   }
 
   int64_t num_indices = 0;
@@ -254,7 +254,7 @@ void IndexElementwisePutWithTensorKernel(
   }
   if (index.empty()) {
     if (!out->initialized()) {
-      phi::Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+      Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     }
     return;
   }
@@ -298,7 +298,7 @@ void IndexElementwisePutKernel(const Context& dev_ctx,
   }
   if (index.empty()) {
     if (!out->initialized()) {
-      phi::Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+      Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     }
     return;
   }
diff --git a/paddle/phi/kernels/xpu/index_put_grad_kernel.cc b/paddle/phi/kernels/xpu/index_put_grad_kernel.cc
index 2b10d9cdde633c..0fa637302c3354 100644
--- a/paddle/phi/kernels/xpu/index_put_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/index_put_grad_kernel.cc
@@ -56,7 +56,7 @@ void IndexPutGradKernel(const Context& dev_ctx,
 
   if (int_indices_v.empty()) {
     if (x_grad) {
-      phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+      Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     }
     if (value_grad) {
       FullKernel<T, Context>(dev_ctx,
@@ -88,7 +88,7 @@ void IndexPutGradKernel(const Context& dev_ctx,
   int ret = 0;
   using XPUType = typename XPUTypeTrait<T>::Type;
   if (x_grad) {
-    phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     if (!accumulate) {
       DenseTensor zero_tensor(x_grad->dtype());
       FullKernel<T, Context>(
diff --git a/paddle/phi/kernels/xpu/index_put_kernel.cc b/paddle/phi/kernels/xpu/index_put_kernel.cc
index eb9124a841c127..fdea3d251f0381 100644
--- a/paddle/phi/kernels/xpu/index_put_kernel.cc
+++ b/paddle/phi/kernels/xpu/index_put_kernel.cc
@@ -54,7 +54,7 @@ void IndexPutKernel(const Context& dev_ctx,
       funcs::DealWithBoolIndices<T, Context>(dev_ctx, indices, &tmp_args);
   if (int_indices_v.empty()) {
     if (!out->initialized()) {
-      phi::Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+      Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     }
     return;
   }
diff --git a/paddle/phi/kernels/xpu/interpolate_grad_kernel.cc b/paddle/phi/kernels/xpu/interpolate_grad_kernel.cc
index 27ea33ccd0cdf0..3fb9d9d217c9cd 100644
--- a/paddle/phi/kernels/xpu/interpolate_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/interpolate_grad_kernel.cc
@@ -128,7 +128,7 @@ void InterpolateGradKernel(
   PADDLE_ENFORCE_XDNN_SUCCESS(r, "constant");
 
   if (in_h == out_h && in_w == out_w) {
-    phi::Copy<Context>(dev_ctx, output_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy<Context>(dev_ctx, output_grad, dev_ctx.GetPlace(), false, x_grad);
     return;
   }
 
diff --git a/paddle/phi/kernels/xpu/interpolate_kernel.cc b/paddle/phi/kernels/xpu/interpolate_kernel.cc
index d655b03a035bb6..c173455adc74cb 100644
--- a/paddle/phi/kernels/xpu/interpolate_kernel.cc
+++ b/paddle/phi/kernels/xpu/interpolate_kernel.cc
@@ -133,7 +133,7 @@ void InterpolateKernel(
   dev_ctx.template Alloc<T>(output);
 
   if (in_h == out_h && in_w == out_w) {
-    phi::Copy<Context>(dev_ctx, x, dev_ctx.GetPlace(), false, output);
+    Copy<Context>(dev_ctx, x, dev_ctx.GetPlace(), false, output);
     return;
   }
   bool nearest = "nearest" == interp_method;
diff --git a/paddle/phi/kernels/xpu/multiclass_nms3_kernel.cc b/paddle/phi/kernels/xpu/multiclass_nms3_kernel.cc
index 3403c9da8b9399..10baa2f42dde92 100644
--- a/paddle/phi/kernels/xpu/multiclass_nms3_kernel.cc
+++ b/paddle/phi/kernels/xpu/multiclass_nms3_kernel.cc
@@ -64,11 +64,11 @@ void MultiClassNMSKernel(const Context& dev_ctx,
       rois_num_host.Resize(rois_num.get_ptr()->dims());
       if (rois_num.get_ptr()->dtype() == phi::DataType::INT64) {
         dev_ctx.template HostAlloc<int64_t>(&rois_num_host);
-        phi::Copy(dev_ctx,
-                  *rois_num.get_ptr(),
-                  rois_num_host.place(),
-                  false,
-                  &rois_num_host);
+        Copy(dev_ctx,
+             *rois_num.get_ptr(),
+             rois_num_host.place(),
+             false,
+             &rois_num_host);
         n = rois_num.get_ptr()->numel();
         for (int64_t i = 0; i < n; i++) {
           rois_num_vec.push_back(rois_num_host.data<int64_t>()[i]);
@@ -76,11 +76,11 @@ void MultiClassNMSKernel(const Context& dev_ctx,
         }
       } else if (rois_num.get_ptr()->dtype() == phi::DataType::INT32) {
         dev_ctx.template HostAlloc<int>(&rois_num_host);
-        phi::Copy(dev_ctx,
-                  *rois_num.get_ptr(),
-                  rois_num_host.place(),
-                  false,
-                  &rois_num_host);
+        Copy(dev_ctx,
+             *rois_num.get_ptr(),
+             rois_num_host.place(),
+             false,
+             &rois_num_host);
         n = rois_num.get_ptr()->numel();
         for (int64_t i = 0; i < n; i++) {
           rois_num_vec.push_back(rois_num_host.data<int>()[i]);
@@ -209,8 +209,7 @@ void MultiClassNMSKernel(const Context& dev_ctx,
     for (int64_t i = 1; i <= n; i++) {
       nms_rois_num_cpu_data[i - 1] = batch_starts[i] - batch_starts[i - 1];
     }
-    phi::Copy(
-        dev_ctx, nms_rois_num_cpu, nms_rois_num->place(), true, nms_rois_num);
+    Copy(dev_ctx, nms_rois_num_cpu, nms_rois_num->place(), true, nms_rois_num);
   }
   LegacyLoD lod;
   if (num_kept == 0) {
diff --git a/paddle/phi/kernels/xpu/scatter_kernel.cc b/paddle/phi/kernels/xpu/scatter_kernel.cc
index d5d3faa28e4a3b..674742314bbeeb 100644
--- a/paddle/phi/kernels/xpu/scatter_kernel.cc
+++ b/paddle/phi/kernels/xpu/scatter_kernel.cc
@@ -29,7 +29,7 @@ void ScatterKernel(const Context &dev_ctx,
                    DenseTensor *out) {
   if (index.numel() == 0) {
     dev_ctx.template Alloc<T>(out);
-    phi::Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+    Copy(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     return;
   }
   if (out && out->numel() == 0) {
@@ -91,7 +91,7 @@ void ScatterKernel(const Context &dev_ctx,
   int64_t dim1 = common::product(common::slice_ddim(x_dims, 1, x_dims.size()));
 
   DenseTensor indices_cpu(index.type());
-  phi::Copy(dev_ctx, index, CPUPlace(), true, &indices_cpu);
+  Copy(dev_ctx, index, CPUPlace(), true, &indices_cpu);
 
   int r = 0;
   if (index_type == phi::DataType::INT32) {
diff --git a/paddle/phi/kernels/xpu/slice_kernel.cc b/paddle/phi/kernels/xpu/slice_kernel.cc
index 520f51f8b38669..d010f317cdb779 100644
--- a/paddle/phi/kernels/xpu/slice_kernel.cc
+++ b/paddle/phi/kernels/xpu/slice_kernel.cc
@@ -63,7 +63,7 @@ void SliceKernel(const Context& dev_ctx,
       }
     }
     if (is_same) {
-      phi::Copy<Context>(dev_ctx, input, dev_ctx.GetPlace(), false, out);
+      Copy<Context>(dev_ctx, input, dev_ctx.GetPlace(), false, out);
       return;
     }
   }
@@ -175,7 +175,7 @@ void SliceKernel<phi::complex64, XPUContext>(
       }
     }
     if (is_same) {
-      phi::Copy<XPUContext>(dev_ctx, input, dev_ctx.GetPlace(), false, out);
+      Copy<XPUContext>(dev_ctx, input, dev_ctx.GetPlace(), false, out);
       return;
     }
   }
diff --git a/paddle/phi/kernels/xpu/tile_grad_kernel.cc b/paddle/phi/kernels/xpu/tile_grad_kernel.cc
index 790cb387235883..e7a84d62542bc8 100644
--- a/paddle/phi/kernels/xpu/tile_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/tile_grad_kernel.cc
@@ -67,7 +67,7 @@ void TileGradKernel(const Context& dev_ctx,
   }
   // no need reduce, just copy
   if (just_copy) {
-    phi::Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     // TensorCopy may change the dims of dx
     x_grad->Resize(x_dims);
   } else {
diff --git a/paddle/phi/kernels/xpu/top_k_grad_kernel.cc b/paddle/phi/kernels/xpu/top_k_grad_kernel.cc
index c72aa24431ba89..38e1fb22f4b697 100644
--- a/paddle/phi/kernels/xpu/top_k_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/top_k_grad_kernel.cc
@@ -66,7 +66,7 @@ void TopkGradKernel(const Context& dev_ctx,
   const int64_t* indices_data = indices.data<int64_t>();
 
   if (in_dims.size() == 0) {
-    phi::Copy<Context>(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy<Context>(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     return;
   }
 
diff --git a/paddle/phi/kernels/xpu/top_k_kernel.cc b/paddle/phi/kernels/xpu/top_k_kernel.cc
index a9bc691ed78b3a..4cc02a3abd0000 100644
--- a/paddle/phi/kernels/xpu/top_k_kernel.cc
+++ b/paddle/phi/kernels/xpu/top_k_kernel.cc
@@ -41,7 +41,7 @@ void TopkKernel(const Context& dev_ctx,
 
   const auto& in_dims = x.dims();
   if (in_dims.size() == 0) {
-    phi::Copy<Context>(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+    Copy<Context>(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     dev_ctx.template Alloc<int64_t>(indices);
     funcs::set_constant(dev_ctx, indices, static_cast<int64_t>(0));
     return;
diff --git a/paddle/phi/kernels/xpu/transpose_grad_kernel.cc b/paddle/phi/kernels/xpu/transpose_grad_kernel.cc
index 501e3eda4d2037..427c4ac37717a8 100644
--- a/paddle/phi/kernels/xpu/transpose_grad_kernel.cc
+++ b/paddle/phi/kernels/xpu/transpose_grad_kernel.cc
@@ -33,7 +33,7 @@ void TransposeGradKernel(const Context& dev_ctx,
 
   size_t axis_size = axis.size();
   if (axis_size == 0) {
-    phi::Copy<Context>(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy<Context>(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     return;
   }
 
@@ -75,7 +75,7 @@ void TransposeGradKernel<phi::complex64, XPUContext>(
 
   size_t axis_size = axis.size();
   if (axis_size == 0) {
-    phi::Copy<XPUContext>(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
+    Copy<XPUContext>(dev_ctx, out_grad, dev_ctx.GetPlace(), false, x_grad);
     return;
   }
 
diff --git a/paddle/phi/kernels/xpu/transpose_kernel.cc b/paddle/phi/kernels/xpu/transpose_kernel.cc
index 17148ba39cb842..28f682fbfca280 100644
--- a/paddle/phi/kernels/xpu/transpose_kernel.cc
+++ b/paddle/phi/kernels/xpu/transpose_kernel.cc
@@ -40,7 +40,7 @@ void TransposeKernel(const Context& dev_ctx,
     return;
   }
   if (formatted_axis.size() == 0) {
-    phi::Copy<Context>(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+    Copy<Context>(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     return;
   }
 
@@ -73,7 +73,7 @@ void TransposeKernel<phi::complex64, XPUContext>(const XPUContext& dev_ctx,
     return;
   }
   if (formatted_axis.size() == 0) {
-    phi::Copy<XPUContext>(dev_ctx, x, dev_ctx.GetPlace(), false, out);
+    Copy<XPUContext>(dev_ctx, x, dev_ctx.GetPlace(), false, out);
     return;
   }