Fix kernels header pos error (#44902)

* fix header pos error

* fix namespace error

paddle/phi/kernels/affine_grid_grad_kernel.h

@@ -15,7 +15,6 @@
 #pragma once
 
 #include "paddle/phi/common/int_array.h"
-#include "paddle/phi/kernels/affine_grid_impl.h"
 
 namespace phi {
 

paddle/phi/kernels/affine_grid_kernel.h

@@ -15,7 +15,6 @@
 #pragma once
 
 #include "paddle/phi/common/int_array.h"
-#include "paddle/phi/kernels/affine_grid_impl.h"
 
 namespace phi {
 

paddle/phi/kernels/cpu/affine_grid_grad_kernel.cc

@@ -13,9 +13,11 @@
 // limitations under the License.
 
 #include "paddle/phi/kernels/affine_grid_grad_kernel.h"
+
 #include "paddle/phi/backends/cpu/cpu_context.h"
 #include "paddle/phi/common/int_array.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/affine_grid_utils.h"
 
 namespace phi {
 

paddle/phi/kernels/cpu/affine_grid_kernel.cc

@@ -13,9 +13,11 @@
 // limitations under the License.
 
 #include "paddle/phi/kernels/affine_grid_kernel.h"
+
 #include "paddle/phi/backends/cpu/cpu_context.h"
 #include "paddle/phi/common/int_array.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/affine_grid_utils.h"
 
 namespace phi {
 

paddle/phi/kernels/cpu/fill_diagonal_grad_kernel.cc

@@ -15,6 +15,7 @@
 #include "paddle/phi/kernels/fill_diagonal_grad_kernel.h"
 
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/common_shape.h"
 
 namespace phi {
 
@@ -30,7 +31,7 @@ void FillDiagonalGradKernel(const Context& ctx,
     phi::Copy(ctx, out_grad, ctx.GetPlace(), false, x_grad);
 
     auto dx_dims = x_grad->dims();
-    auto strides = CalStride(dx_dims);
+    auto strides = funcs::CalStride(dx_dims);
     auto size = x_grad->numel();
     auto wrapsize = std::min(size, dx_dims[1] * dx_dims[1]);
 

paddle/phi/kernels/cpu/fill_diagonal_kernel.cc

@@ -16,6 +16,7 @@
 
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/tensor_utils.h"
+#include "paddle/phi/kernels/funcs/common_shape.h"
 
 namespace phi {
 
@@ -32,7 +33,7 @@ void FillDiagonalKernel(const Context& ctx,
   phi::Copy(ctx, x, ctx.GetPlace(), false, out);
 
   auto out_dims = out->dims();
-  auto strides = CalStride(out_dims);
+  auto strides = funcs::CalStride(out_dims);
   auto size = out->numel();
 
   // The wrap mode supported only the dims equels to 2; In wrap mode, the

paddle/phi/kernels/fill_diagonal_grad_kernel.h

@@ -16,8 +16,6 @@
 
 #include "paddle/phi/core/dense_tensor.h"
 
-#include "paddle/phi/kernels/impl/fill_diagonal_kernel_impl.h"
-
 namespace phi {
 
 template <typename T, typename Context>

paddle/phi/kernels/fill_diagonal_kernel.h

@@ -15,9 +15,6 @@
 #pragma once
 
 #include "paddle/phi/core/dense_tensor.h"
-
-#include "paddle/phi/kernels/impl/fill_diagonal_kernel_impl.h"
-
 namespace phi {
 
 template <typename T, typename Context>

paddle/phi/kernels/funcs/common_shape.h

@@ -231,5 +231,16 @@ inline DDim GetOutputDims(const DDim &s_dims, const DDim &l_dims) {
   return phi::make_ddim(shapes);
 }
 
+inline int64_t CalStride(phi::DDim dim) {
+  int rank = dim.size();
+  int64_t dimsum = 1;
+  int64_t strides = 0;
+  for (int i = rank - 1; i >= 0; i--) {
+    strides += dimsum;
+    dimsum *= dim[i];
+  }
+  return strides;
+}
+
 }  // namespace funcs
 }  // namespace phi

paddle/phi/kernels/gpu/affine_grid_grad_kernel.cu

@@ -15,6 +15,7 @@
 #pragma once
 
 #include "paddle/phi/kernels/affine_grid_grad_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_device_function.h"
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
@@ -22,6 +23,7 @@
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/int_array.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/affine_grid_utils.h"
 
 namespace phi {
 

paddle/phi/kernels/gpu/affine_grid_kernel.cu

@@ -15,6 +15,7 @@
 #pragma once
 
 #include "paddle/phi/kernels/affine_grid_kernel.h"
+
 #include "paddle/fluid/platform/device/gpu/gpu_device_function.h"
 #include "paddle/fluid/platform/device/gpu/gpu_info.h"
 #include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
@@ -22,6 +23,7 @@
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/common/int_array.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/affine_grid_utils.h"
 
 namespace phi {
 

paddle/phi/kernels/gpu/fill_diagonal_grad_kernel.cu

@@ -18,6 +18,7 @@
 #include <vector>
 
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/common_shape.h"
 
 namespace phi {
 
@@ -60,7 +61,7 @@ void FillDiagonalGradKernel(const Context& ctx,
 
   auto size = x_grad->numel();
   auto out_dims = x_grad->dims();
-  auto strides = CalStride(out_dims);
+  auto strides = funcs::CalStride(out_dims);
 
   auto wrapsize = std::min(size, out_dims[1] * out_dims[1]);
   // The wrap mode supported only the dims equels to 2; In wrap mode, the

paddle/phi/kernels/gpu/fill_diagonal_kernel.cu

@@ -19,6 +19,7 @@
 
 #include "paddle/fluid/framework/convert_utils.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/common_shape.h"
 
 namespace phi {
 
@@ -63,7 +64,7 @@ void FillDiagonalKernel(const Context& ctx,
 
   auto size = out->numel();
   auto out_dims = out->dims();
-  auto strides = CalStride(out_dims);
+  auto strides = funcs::CalStride(out_dims);
 
   // The wrap mode supported only the dims equels to 2; In wrap mode, the
   // value will be filled in cycles

paddle/phi/kernels/impl/fill_diagonal_kernel_impl.h

-// Copyright (c) 2022 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 "paddle/phi/core/dense_tensor.h"
-
-namespace phi {
-
-inline int64_t CalStride(phi::DDim dim) {
-  int rank = dim.size();
-  int64_t dimsum = 1;
-  int64_t strides = 0;
-  for (int i = rank - 1; i >= 0; i--) {
-    strides += dimsum;
-    dimsum *= dim[i];
-  }
-  return strides;
-}
-
-}  // namespace phi