CUDA Kernel No.52 (#2240)

* Add elementwise_grad_kernel.cu to CMakeLists.txt for GPU backend

* Fix: Replace .cu include with .h in elementwise_grad_kernel_register.cu

* fix: Correct .cu include and CMake for embedding_with_scaled_gradient_grad kernel.

* FIX(iluvatar_gpu): Correct include path for embedding_grad kernel registration file

* FIX(iluvatar_gpu): Include missing headers for embedding_grad kernel registration.

* Change included .cu files to .h

* Fix flip_kernel include path and clean up CMakeLists redundancy

---------

Co-authored-by: Yuqiang Ge <143453447+YqGe585@users.noreply.github.com>

Author: LiaoYFBH

backends/iluvatar_gpu/CMakeLists.txt

@@ -327,7 +327,6 @@ file(
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/cvm_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/embedding_with_scaled_gradient_grad_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/exponential_kernel.cu
-  ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/flip_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/gather_grad_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/gelu_grad_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/global_gather_kernel.cu
@@ -901,7 +900,6 @@ file(
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/eigh_grad_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/eigvalsh_grad_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/exponential_kernel.cu
-  ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/flip_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/gather_tree_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/graph_reindex_kernel.cu
   ${PADDLE_SOURCE_DIR}/paddle/phi/kernels/gpu/graph_sample_neighbors_kernel.cu

backends/metax_gpu/kernels/cuda_kernels/flip_kernel_register.cu

@@ -13,7 +13,7 @@
 // limitations under the License.
 
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/gpu/flip_kernel.cu"  //NOLINT
+#include "paddle/phi/kernels/flip_kernel.h"  //NOLINT
 PD_CUSTOM_KERNEL_REGISTER(flip,
                           metax_gpu,
                           ALL_LAYOUT,