[PHI decoupling] remove dependency on "paddle/fluid/operators/elementwise/xxx.h" in phi (#47870)

* rm "paddle/fluid/operators/elementwise/xxx.h" in phi

* fix bugs

* add LaunchElementwiseCudaKernel in phi

* Revert "add LaunchElementwiseCudaKernel in phi"

This reverts commit 588f45bbdad2372ec7bff0c567a29bff675d22e1.

* rm indirect dependence to "elementwise_op_impl.cu.h"

rm indirect dependence to "elementwise_op_impl.cu.h"

Revert "add LaunchElementwiseCudaKernel in phi"

This reverts commit 588f45bbdad2372ec7bff0c567a29bff675d22e1.

add LaunchElementwiseCudaKernel in phi

fix bugs

* rm LaunchSameDimsElementwiseCudaKernel and LaunchElementwiseCudaKernel in phi

paddle/phi/kernels/gpu/label_smooth_grad_kernel.cu

@@ -14,9 +14,9 @@
 
 #include "paddle/phi/kernels/label_smooth_grad_kernel.h"
 
-#include "paddle/fluid/operators/elementwise/elementwise_op_impl.cu.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/elementwise_base.h"
 
 namespace phi {
 template <typename T>
@@ -42,8 +42,7 @@ void LabelSmoothGradKernel(const Context& ctx,
   std::vector<const DenseTensor*> ins = {&out_grad};
   std::vector<DenseTensor*> outs = {label_grad};
   auto functor = LabelSmoothGradFunctor<T>(epsilon);
-  paddle::operators::LaunchSameDimsElementwiseCudaKernel<T>(
-      ctx, ins, &outs, functor);
+  phi::funcs::ElementwiseKernel<T>(ctx, ins, &outs, functor);
 }
 
 }  // namespace phi

paddle/phi/kernels/gpu/label_smooth_kernel.cu

@@ -16,9 +16,9 @@
 
 #include <vector>
 
-#include "paddle/fluid/operators/elementwise/elementwise_op_impl.cu.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/elementwise_base.h"
 
 namespace phi {
 
@@ -77,8 +77,7 @@ void LabelSmoothKernel(const Context& ctx,
     std::vector<const DenseTensor*> ins = {&label};
     std::vector<DenseTensor*> outs = {out};
     auto functor = LabelSmoothFunctor<T>(epsilon, label_dim);
-    paddle::operators::LaunchSameDimsElementwiseCudaKernel<T>(
-        ctx, ins, &outs, functor);
+    phi::funcs::ElementwiseKernel<T>(ctx, ins, &outs, functor);
   }
 }
 

paddle/phi/kernels/gpu/p_norm_kernel.cu

@@ -14,9 +14,9 @@
 
 #include "paddle/phi/kernels/p_norm_kernel.h"
 
-#include "paddle/fluid/operators/elementwise/elementwise_op_impl.cu.h"
 #include "paddle/phi/common/amp_type_traits.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/funcs/elementwise_base.h"
 #include "paddle/phi/kernels/funcs/reduce_function.h"
 #include "paddle/phi/kernels/gpu/reduce.h"
 

paddle/phi/kernels/gpu/viterbi_decode_kernel.cu

@@ -30,13 +30,14 @@ namespace cub = hipcub;
 #include <string>
 #include <vector>
 
-#include "paddle/fluid/operators/elementwise/elementwise_op_function.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/tensor_utils.h"
 #include "paddle/phi/kernels/empty_kernel.h"
+#include "paddle/phi/kernels/funcs/broadcast_function.h"
 #include "paddle/phi/kernels/funcs/compare_functors.h"
 #include "paddle/phi/kernels/funcs/concat_and_split_functor.h"
+#include "paddle/phi/kernels/funcs/elementwise_base.h"
 #include "paddle/phi/kernels/funcs/elementwise_functor.h"
 #include "paddle/phi/kernels/funcs/gather.cu.h"
 #include "paddle/phi/kernels/funcs/viterbi_decode_functor.h"
@@ -90,9 +91,8 @@ struct BinaryOperation {
                   DenseTensor* output) {
     std::vector<const DenseTensor*> ins{&lhs, &rhs};
     std::vector<DenseTensor*> outs{output};
-    paddle::operators::
-        LaunchElementwiseCudaKernel<ElementwiseType::kBinary, T, T>(
-            dev_ctx, ins, &outs, 0, BinaryFunctor<T>());
+    phi::funcs::BroadcastKernel<phi::ElementwiseType::kBinary, T, T>(
+        dev_ctx, ins, &outs, 0, BinaryFunctor<T>());
   }
 };
 
@@ -107,7 +107,7 @@ struct GetMask {
                   DenseTensor* mask) {
     std::vector<const DenseTensor*> ins = {&lhs, &rhs};
     std::vector<DenseTensor*> outs = {mask};
-    paddle::operators::LaunchSameDimsElementwiseCudaKernel<T>(
+    phi::funcs::ElementwiseKernel<T>(
         dev_ctx, ins, &outs, CompareFunctor<int64_t, T>());
   }
 };