move REGISTER_OP_CUDA_KERNEL into cpp with eigen, test=develop (#32114)

paddle/fluid/operators/expand_as_op.cc

@@ -147,3 +147,17 @@ REGISTER_OP_CPU_KERNEL(
     ops::ExpandAsGradKernel<paddle::platform::CPUDeviceContext, int64_t>,
     ops::ExpandAsGradKernel<paddle::platform::CPUDeviceContext, float>,
     ops::ExpandAsGradKernel<paddle::platform::CPUDeviceContext, double>);
+#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_ROCM)
+REGISTER_OP_CUDA_KERNEL(
+    expand_as, ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, bool>);
+REGISTER_OP_CUDA_KERNEL(
+    expand_as_grad,
+    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, double>);
+#endif

paddle/fluid/operators/expand_as_op.cu

-/* Copyright (c) 2019 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/operators/expand_as_op.h"
-
-namespace ops = paddle::operators;
-REGISTER_OP_CUDA_KERNEL(
-    expand_as, ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, int64_t>,
-    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, bool>);
-REGISTER_OP_CUDA_KERNEL(
-    expand_as_grad,
-    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, int64_t>,
-    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, double>);

paddle/fluid/operators/expand_as_v2_op.cc

@@ -129,3 +129,18 @@ REGISTER_OP_CPU_KERNEL(
     ops::ExpandAsV2GradKernel<paddle::platform::CPUDeviceContext, int64_t>,
     ops::ExpandAsV2GradKernel<paddle::platform::CPUDeviceContext, float>,
     ops::ExpandAsV2GradKernel<paddle::platform::CPUDeviceContext, double>);
+#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_ROCM)
+REGISTER_OP_CUDA_KERNEL(
+    expand_as_v2,
+    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, double>,
+    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, bool>);
+REGISTER_OP_CUDA_KERNEL(
+    expand_as_v2_grad,
+    ops::ExpandAsV2GradKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandAsV2GradKernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::ExpandAsV2GradKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ExpandAsV2GradKernel<paddle::platform::CUDADeviceContext, double>);
+#endif

paddle/fluid/operators/expand_as_v2_op.cu

-/* Copyright (c) 2019 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/operators/expand_as_v2_op.h"
-
-namespace ops = paddle::operators;
-REGISTER_OP_CUDA_KERNEL(
-    expand_as_v2,
-    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, float>,
-    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, double>,
-    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, int>,
-    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, int64_t>,
-    ops::ExpandAsV2Kernel<paddle::platform::CUDADeviceContext, bool>);
-REGISTER_OP_CUDA_KERNEL(
-    expand_as_v2_grad,
-    ops::ExpandAsV2GradKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::ExpandAsV2GradKernel<paddle::platform::CUDADeviceContext, int64_t>,
-    ops::ExpandAsV2GradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::ExpandAsV2GradKernel<paddle::platform::CUDADeviceContext, double>);

paddle/fluid/operators/expand_op.cc

@@ -273,3 +273,21 @@ REGISTER_OP_CPU_KERNEL(
     ops::ExpandGradKernel<paddle::platform::CPUDeviceContext, double>,
     ops::ExpandGradKernel<paddle::platform::CPUDeviceContext, int>,
     ops::ExpandGradKernel<paddle::platform::CPUDeviceContext, int64_t>);
+#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_ROCM)
+REGISTER_OP_CUDA_KERNEL(
+    expand, ops::ExpandKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ExpandKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::ExpandKernel<paddle::platform::CUDADeviceContext,
+                      paddle::platform::float16>,
+    ops::ExpandKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandKernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::ExpandKernel<paddle::platform::CUDADeviceContext, bool>);
+REGISTER_OP_CUDA_KERNEL(
+    expand_grad,
+    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext,
+                          paddle::platform::float16>,
+    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, int64_t>);
+#endif

paddle/fluid/operators/expand_op.cu

-/* Copyright (c) 2016 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/operators/expand_op.h"
-
-namespace ops = paddle::operators;
-namespace plat = paddle::platform;
-
-REGISTER_OP_CUDA_KERNEL(
-    expand, ops::ExpandKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::ExpandKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::ExpandKernel<paddle::platform::CUDADeviceContext, plat::float16>,
-    ops::ExpandKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::ExpandKernel<paddle::platform::CUDADeviceContext, int64_t>,
-    ops::ExpandKernel<paddle::platform::CUDADeviceContext, bool>);
-REGISTER_OP_CUDA_KERNEL(
-    expand_grad,
-    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, plat::float16>,
-    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::ExpandGradKernel<paddle::platform::CUDADeviceContext, int64_t>);

paddle/fluid/operators/expand_v2_op.cc

@@ -278,3 +278,21 @@ REGISTER_OP_CPU_KERNEL(
     ops::ExpandV2GradKernel<paddle::platform::CPUDeviceContext, double>,
     ops::ExpandV2GradKernel<paddle::platform::CPUDeviceContext, int>,
     ops::ExpandV2GradKernel<paddle::platform::CPUDeviceContext, int64_t>);
+#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_ROCM)
+REGISTER_OP_CUDA_KERNEL(
+    expand_v2, ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, double>,
+    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext,
+                        paddle::platform::float16>,
+    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, bool>);
+REGISTER_OP_CUDA_KERNEL(
+    expand_v2_grad,
+    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext,
+                            paddle::platform::float16>,
+    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, int64_t>);
+#endif

paddle/fluid/operators/expand_v2_op.cu

-/* Copyright (c) 2016 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/operators/expand_v2_op.h"
-
-namespace ops = paddle::operators;
-namespace plat = paddle::platform;
-
-REGISTER_OP_CUDA_KERNEL(
-    expand_v2, ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, float>,
-    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, double>,
-    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, plat::float16>,
-    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, int>,
-    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, int64_t>,
-    ops::ExpandV2Kernel<paddle::platform::CUDADeviceContext, bool>);
-REGISTER_OP_CUDA_KERNEL(
-    expand_v2_grad,
-    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, plat::float16>,
-    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::ExpandV2GradKernel<paddle::platform::CUDADeviceContext, int64_t>);

paddle/fluid/operators/meshgrid_op.cc

@@ -157,3 +157,17 @@ REGISTER_OP_CPU_KERNEL(
     ops::MeshgridGradKernel<paddle::platform::CPUDeviceContext, int64_t>,
     ops::MeshgridGradKernel<paddle::platform::CPUDeviceContext, int>,
     ops::MeshgridGradKernel<paddle::platform::CPUDeviceContext, double>);
+#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_ROCM)
+REGISTER_OP_CUDA_KERNEL(
+    meshgrid, ops::MeshgridKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::MeshgridKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::MeshgridKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::MeshgridKernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::MeshgridKernel<paddle::platform::CUDADeviceContext, bool>);
+REGISTER_OP_CUDA_KERNEL(
+    meshgrid_grad,
+    ops::MeshgridGradKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::MeshgridGradKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::MeshgridGradKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::MeshgridGradKernel<paddle::platform::CUDADeviceContext, int64_t>);
+#endif

paddle/fluid/operators/meshgrid_op.cu

-// Copyright (c) 2020 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/operators/meshgrid_op.h"
-
-namespace ops = paddle::operators;
-REGISTER_OP_CUDA_KERNEL(
-    meshgrid, ops::MeshgridKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::MeshgridKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::MeshgridKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::MeshgridKernel<paddle::platform::CUDADeviceContext, int64_t>,
-    ops::MeshgridKernel<paddle::platform::CUDADeviceContext, bool>);
-REGISTER_OP_CUDA_KERNEL(
-    meshgrid_grad,
-    ops::MeshgridGradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::MeshgridGradKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::MeshgridGradKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::MeshgridGradKernel<paddle::platform::CUDADeviceContext, int64_t>);

paddle/fluid/operators/tile_op.cc

@@ -286,3 +286,20 @@ REGISTER_OP_CPU_KERNEL(
     ops::TileGradKernel<paddle::platform::CPUDeviceContext, double>,
     ops::TileGradKernel<paddle::platform::CPUDeviceContext, int>,
     ops::TileGradKernel<paddle::platform::CPUDeviceContext, int64_t>);
+#if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_ROCM)
+REGISTER_OP_CUDA_KERNEL(
+    tile, ops::TileKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::TileKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::TileKernel<paddle::platform::CUDADeviceContext,
+                    paddle::platform::float16>,
+    ops::TileKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::TileKernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::TileKernel<paddle::platform::CUDADeviceContext, bool>);
+REGISTER_OP_CUDA_KERNEL(
+    tile_grad, ops::TileGradKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::TileGradKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::TileGradKernel<paddle::platform::CUDADeviceContext,
+                        paddle::platform::float16>,
+    ops::TileGradKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::TileGradKernel<paddle::platform::CUDADeviceContext, int64_t>);
+#endif

paddle/fluid/operators/tile_op.cu

-/* Copyright (c) 2016 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/operators/tile_op.h"
-
-namespace ops = paddle::operators;
-namespace plat = paddle::platform;
-
-REGISTER_OP_CUDA_KERNEL(
-    tile, ops::TileKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::TileKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::TileKernel<paddle::platform::CUDADeviceContext, plat::float16>,
-    ops::TileKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::TileKernel<paddle::platform::CUDADeviceContext, int64_t>,
-    ops::TileKernel<paddle::platform::CUDADeviceContext, bool>);
-REGISTER_OP_CUDA_KERNEL(
-    tile_grad, ops::TileGradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::TileGradKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::TileGradKernel<paddle::platform::CUDADeviceContext, plat::float16>,
-    ops::TileGradKernel<paddle::platform::CUDADeviceContext, int>,
-    ops::TileGradKernel<paddle::platform::CUDADeviceContext, int64_t>);