[Pten] move operators/math/math_function_* to pten/kernels/func

paddle/fluid/distributed/ps/service/communicator/communicator.h

@@ -35,12 +35,12 @@ limitations under the License. */
 #include "paddle/fluid/framework/variable.h"
 #include "paddle/fluid/framework/variable_helper.h"
 #include "paddle/fluid/operators/math/blas.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/operators/math/selected_rows_functor.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/enforce.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/string/split.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 #include "paddle/fluid/distributed/ps/service/ps_client.h"
 
@@ -180,7 +180,7 @@ inline void MergeVars(const std::string &var_name,
 
     // set output tensor to 0.
     paddle::platform::CPUDeviceContext cpu_ctx;
-    paddle::operators::math::SetConstant<paddle::platform::CPUDeviceContext, T>
+    pten::funcs::SetConstant<paddle::platform::CPUDeviceContext, T>
         constant_functor;
     constant_functor(cpu_ctx, out_t, static_cast<T>(0));
     // sum all vars to out

paddle/fluid/distributed/ps/service/ps_service/graph_py_service.h

@@ -38,9 +38,10 @@
 #include "paddle/fluid/distributed/ps/service/ps_service/service.h"
 #include "paddle/fluid/distributed/ps/service/sendrecv.pb.h"
 #include "paddle/fluid/framework/program_desc.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/string/printf.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
+
 namespace paddle {
 namespace distributed {
 class GraphPyService {

paddle/fluid/distributed/test/brpc_service_dense_sgd_test.cc

@@ -21,8 +21,8 @@ limitations under the License. */
 #include "paddle/fluid/distributed/ps/service/brpc_ps_server.h"
 #include "paddle/fluid/framework/program_desc.h"
 #include "paddle/fluid/framework/scope.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/place.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace paddle {
 namespace distributed {
@@ -42,7 +42,6 @@ class DenseTensor;
 namespace framework = paddle::framework;
 namespace platform = paddle::platform;
 namespace operators = paddle::operators;
-namespace math = paddle::operators::math;
 namespace memory = paddle::memory;
 namespace distributed = paddle::distributed;
 

paddle/fluid/distributed/test/brpc_service_sparse_sgd_test.cc

@@ -22,8 +22,8 @@ limitations under the License. */
 #include "paddle/fluid/distributed/ps/service/brpc_ps_server.h"
 #include "paddle/fluid/distributed/ps/service/env.h"
 #include "paddle/fluid/framework/program_desc.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/place.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace paddle {
 namespace distributed {
@@ -43,7 +43,6 @@ class DenseTensor;
 namespace framework = paddle::framework;
 namespace platform = paddle::platform;
 namespace operators = paddle::operators;
-namespace math = paddle::operators::math;
 namespace memory = paddle::memory;
 namespace distributed = paddle::distributed;
 

paddle/fluid/distributed/test/brpc_utils_test.cc

@@ -17,7 +17,7 @@ limitations under the License. */
 #include "gtest/gtest.h"
 
 #include "paddle/fluid/distributed/ps/service/brpc_utils.h"
-#include "paddle/fluid/operators/math/math_function.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace paddle {
 namespace framework {
@@ -28,7 +28,6 @@ class Variable;
 namespace framework = paddle::framework;
 namespace platform = paddle::platform;
 namespace operators = paddle::operators;
-namespace math = paddle::operators::math;
 namespace memory = paddle::memory;
 namespace distributed = paddle::distributed;
 
@@ -42,7 +41,7 @@ void CreateVarsOnScope(framework::Scope* scope, platform::Place* place,
   lod1.push_back(framework::Vector<size_t>({1, 3, 8}));
   tensor1->set_lod(lod1);
   tensor1->mutable_data<float>(*place);
-  math::set_constant(ctx, tensor1, 31.9);
+  pten::funcs::set_constant(ctx, tensor1, 31.9);
 
   // var 2
   framework::Variable* var2 = scope->Var("x2");
@@ -52,7 +51,7 @@ void CreateVarsOnScope(framework::Scope* scope, platform::Place* place,
   lod2.push_back(framework::Vector<size_t>({1, 1}));
   tensor2->set_lod(lod2);
   tensor2->mutable_data<int>(*place);
-  math::set_constant(ctx, tensor2, 100);
+  pten::funcs::set_constant(ctx, tensor2, 100);
 
   // var 3
   framework::Variable* var3 = scope->Var("x3");
@@ -62,7 +61,7 @@ void CreateVarsOnScope(framework::Scope* scope, platform::Place* place,
   auto* rows = slr->mutable_rows();
   tensor3->Resize(framework::make_ddim({564, 128}));
   tensor3->mutable_data<float>(*place);
-  math::set_constant(ctx, tensor3, 32.7);
+  pten::funcs::set_constant(ctx, tensor3, 32.7);
   for (int i = 0; i < 564; ++i) rows->push_back(i);
 }
 

paddle/fluid/distributed/test/graph_node_split_test.cc

@@ -36,14 +36,13 @@ limitations under the License. */
 #include "paddle/fluid/framework/scope.h"
 #include "paddle/fluid/framework/tensor_util.h"
 #include "paddle/fluid/framework/variable.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/string/printf.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace framework = paddle::framework;
 namespace platform = paddle::platform;
 namespace operators = paddle::operators;
-namespace math = paddle::operators::math;
 namespace memory = paddle::memory;
 namespace distributed = paddle::distributed;
 

paddle/fluid/distributed/test/graph_node_test.cc

@@ -36,14 +36,13 @@ limitations under the License. */
 #include "paddle/fluid/framework/scope.h"
 #include "paddle/fluid/framework/tensor_util.h"
 #include "paddle/fluid/framework/variable.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/string/printf.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace framework = paddle::framework;
 namespace platform = paddle::platform;
 namespace operators = paddle::operators;
-namespace math = paddle::operators::math;
 namespace memory = paddle::memory;
 namespace distributed = paddle::distributed;
 

paddle/fluid/eager/grad_tensor_holder.cc

@@ -16,7 +16,7 @@
 #include "paddle/fluid/imperative/gradient_accumulator.h"
 
 #include "paddle/fluid/framework/var_type.h"
-#include "paddle/fluid/operators/math/math_function.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace egr {
 

paddle/fluid/framework/data_device_transform_test.cu

@@ -19,9 +19,9 @@ limitations under the License. */
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/framework/scope.h"
 #include "paddle/fluid/operators/elementwise/elementwise_op_function.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/init.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 #include "paddle/fluid/framework/pten_utils.h"
 

paddle/fluid/framework/data_layout_transform.cc

@@ -14,7 +14,7 @@
 
 #include "paddle/fluid/framework/data_layout_transform.h"
 
-#include "paddle/fluid/operators/math/math_function.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 #ifdef PADDLE_WITH_MKLDNN
 #include "paddle/fluid/platform/mkldnn_reuse.h"
 #endif
@@ -42,7 +42,7 @@ void CastDataLayout::apply() {
   auto place = ctx_->GetPlace();
 
   if (platform::is_cpu_place(place)) {
-    operators::math::Transpose<platform::CPUDeviceContext, T, 4> trans4;
+    pten::funcs::Transpose<platform::CPUDeviceContext, T, 4> trans4;
     auto* context = static_cast<const platform::CPUDeviceContext*>(ctx_);
     trans4(*context, in_, out_, axis_);
   } else {

paddle/fluid/framework/data_transform.h

@@ -22,10 +22,10 @@ limitations under the License. */
 #include "paddle/fluid/framework/selected_rows_utils.h"
 #include "paddle/fluid/framework/tensor.h"
 #include "paddle/fluid/framework/variable.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/macros.h"
 #include "paddle/fluid/platform/transform.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace paddle {
 namespace framework {

paddle/fluid/framework/ir/multi_devices_graph_pass/multi_devices_graph_pass.cc

@@ -33,7 +33,7 @@
 #include "paddle/fluid/framework/ir/node.h"
 #include "paddle/fluid/framework/op_info.h"
 #include "paddle/fluid/framework/scope.h"
-#include "paddle/fluid/operators/math/math_function.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 #if defined(PADDLE_WITH_DGC)
 #include "paddle/fluid/framework/details/sparse_all_reduce_op_handle.h"

paddle/fluid/imperative/basic_engine.cc

@@ -28,8 +28,8 @@
 #include "paddle/fluid/imperative/layer.h"
 #include "paddle/fluid/imperative/op_base.h"
 #include "paddle/fluid/imperative/tracer.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/profiler.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 DECLARE_bool(sort_sum_gradient);
 
@@ -103,7 +103,7 @@ void BasicEngine::Init(
     if (grad_tensor == nullptr) {
       grad_var->Resize(fwd_var.dims());
       grad_var->mutable_data(fwd_var.place(), fwd_var.type());
-      operators::math::set_constant(*dev_ctx, grad_var, 1.0);
+      pten::funcs::set_constant(*dev_ctx, grad_var, 1.0);
     } else {
       paddle::framework::TensorCopy(
           grad_tensor->Var().Get<framework::LoDTensor>(), fwd_var.place(),
@@ -156,7 +156,7 @@ void BasicEngine::CheckBackwardInputs(const OpBase& op) {
         VLOG(6) << "Set ungenerated Grad: " << var->Name()
                 << " as zero with dtype "
                 << framework::DataTypeToString(var->ForwardDataType());
-        operators::math::set_constant(*dev_ctx, tensor, 0.0);
+        pten::funcs::set_constant(*dev_ctx, tensor, 0.0);
       }
     }
   }

paddle/fluid/imperative/gradient_accumulator.cc

@@ -22,12 +22,12 @@
 #include "paddle/fluid/framework/selected_rows_utils.h"
 #include "paddle/fluid/imperative/layer.h"
 #include "paddle/fluid/operators/math/blas.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/operators/math/selected_rows_functor.h"
 #include "paddle/fluid/platform/complex.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/float16.h"
 #include "paddle/fluid/platform/profiler.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 #ifdef PADDLE_WITH_XPU
 #include "xpu/refactor/math.h"
 #endif
@@ -210,7 +210,7 @@ void TensorAddImpl(const framework::Tensor& src, framework::Tensor* dst,
   platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
   paddle::platform::DeviceContext* ctx = pool.Get(place);
   auto dev_ctx = dynamic_cast<DeviceContext*>(ctx);
-  operators::math::ElementwiseAddTo<DeviceContext, T> func;
+  pten::funcs::ElementwiseAddTo<DeviceContext, T> func;
   func(dev_ctx, src, dst);
 }
 
@@ -703,12 +703,12 @@ void EagerGradientAccumulator::SumGrad(std::shared_ptr<VariableWrapper> var,
                 << var->Var().Get<framework::LoDTensor>().dims();
         tensor->Resize(var->Var().Get<framework::LoDTensor>().dims());
         tensor->mutable_data(place, var->DataType());
-        operators::math::set_constant(*dev_ctx, tensor, 0.0);
+        pten::funcs::set_constant(*dev_ctx, tensor, 0.0);
       } else {
         auto* tensor =
             dst_var->MutableVar()->GetMutable<framework::LoDTensor>();
         tensor->mutable_data(place, var->DataType());
-        operators::math::set_constant(*dev_ctx, tensor, 0.0);
+        pten::funcs::set_constant(*dev_ctx, tensor, 0.0);
       }
     }
   }
@@ -835,12 +835,12 @@ void SortedGradientAccumulator::SumGrad(std::shared_ptr<VariableWrapper> var,
                 << var->Var().Get<framework::LoDTensor>().dims();
         tensor->Resize(var->Var().Get<framework::LoDTensor>().dims());
         tensor->mutable_data(place, var->DataType());
-        operators::math::set_constant(*dev_ctx, tensor, 0.0);
+        pten::funcs::set_constant(*dev_ctx, tensor, 0.0);
       } else {
         auto* tensor =
             dst_var->MutableVar()->GetMutable<framework::LoDTensor>();
         tensor->mutable_data(place, var->DataType());
-        operators::math::set_constant(*dev_ctx, tensor, 0.0);
+        pten::funcs::set_constant(*dev_ctx, tensor, 0.0);
       }
     }
     // looks like tmp_grad_vars will not have any member but just in case

paddle/fluid/imperative/layer.cc

@@ -20,10 +20,10 @@
 #include "paddle/fluid/imperative/op_base.h"
 #include "paddle/fluid/imperative/prepared_operator.h"
 #include "paddle/fluid/imperative/var_helper.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/enforce.h"
 #include "paddle/fluid/platform/profiler.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 #ifdef PADDLE_WITH_MKLDNN
 #include "paddle/fluid/platform/mkldnn_helper.h"
 #endif
@@ -229,7 +229,7 @@ void VarBase::ClearGradient(bool set_to_zero) {
         if (set_to_zero) {
           auto* dev_ctx =
               platform::DeviceContextPool::Instance().Get(grad_t->place());
-          operators::math::set_constant(*dev_ctx, grad_t, 0.0);
+          pten::funcs::set_constant(*dev_ctx, grad_t, 0.0);
         } else {
           grad_t->clear();
         }

paddle/fluid/imperative/partial_grad_engine.cc

@@ -28,10 +28,10 @@
 #include "paddle/fluid/imperative/layer.h"
 #include "paddle/fluid/imperative/op_base.h"
 #include "paddle/fluid/imperative/tracer.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/profiler.h"
 #include "paddle/fluid/string/string_helper.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 DECLARE_bool(sort_sum_gradient);
 
@@ -316,7 +316,7 @@ static void FillConstantLike(const VariableWrapper &ref_var,
   } else {
     dst_tensor->mutable_data(place, ref_var.DataType());
   }
-  operators::math::set_constant(*dev_ctx, dst_tensor, value);
+  pten::funcs::set_constant(*dev_ctx, dst_tensor, value);
 }
 
 /**

paddle/fluid/imperative/reducer.cc

@@ -755,7 +755,7 @@ void Reducer::MarkVarReady(const size_t var_index, const bool is_used_var) {
             {static_cast<int64_t>(length)});
       } else {
         group_tensor.Resize({static_cast<int64_t>(length)});
-        operators::math::set_constant(*dev_ctx, &group_tensor, 0.0);
+        pten::funcs::set_constant(*dev_ctx, &group_tensor, 0.0);
       }
 #endif
     }

paddle/fluid/imperative/reducer.h

@@ -29,8 +29,8 @@
 #include "paddle/fluid/framework/tensor.h"
 #include "paddle/fluid/framework/tensor_util.h"
 #include "paddle/fluid/framework/variable.h"
-#include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/for_range.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace paddle {
 namespace imperative {

paddle/fluid/imperative/tests/test_gradient_accmulator.cc

@@ -20,7 +20,7 @@
 #include "paddle/fluid/framework/variable.h"
 #include "paddle/fluid/imperative/gradient_accumulator.h"
 #include "paddle/fluid/memory/memcpy.h"
-#include "paddle/fluid/operators/math/math_function.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace imperative = paddle::imperative;
 namespace platform = paddle::platform;

paddle/fluid/operators/addmm_op.h

@@ -20,7 +20,7 @@ limitations under the License. */
 #include "paddle/fluid/framework/operator.h"
 #include "paddle/fluid/operators/eigen/eigen_function.h"
 #include "paddle/fluid/operators/math/blas.h"
-#include "paddle/fluid/operators/math/math_function.h"
+#include "paddle/pten/kernels/funcs/math_function.h"
 
 namespace ops = paddle::operators;
 namespace plat = paddle::platform;

paddle/fluid/operators/affine_grid_op.cu

@@ -170,7 +170,7 @@ class AffineGridGradOpCUDAKernel : public framework::OpKernel<T> {
       w = size_attr[3];
     }
     T* theta_grad_data = theta_grad->mutable_data<T>({n, 2, 3}, ctx.GetPlace());
-    math::SetConstant<paddle::platform::CUDADeviceContext, T>()(
+    pten::funcs::SetConstant<paddle::platform::CUDADeviceContext, T>()(
         ctx.cuda_device_context(), theta_grad, static_cast<T>(0));
 
     T h_step;