!18919 Add TensorCopySlices op

Merge pull request !18919 from caifubi/master-tensor-copy-slices

mindspore/ccsrc/backend/kernel_compiler/common_utils.cc

@@ -856,5 +856,78 @@ bool GetShapeSize(const std::vector<size_t> &shape, const TypePtr &type_ptr, int
   size_i[0] = LongMulWithOverflowCheck(size_i[0], SizeToInt(type_byte));
   return true;
 }
+
+void CastShapeSizeToLong(const std::vector<size_t> &shape, std::vector<int64_t> *long_shape) {
+  MS_EXCEPTION_IF_NULL(long_shape);
+  std::transform(shape.begin(), shape.end(), std::back_inserter(*long_shape), SizeToLong);
+}
+
+void CheckSliceValid(const std::vector<int64_t> &start, const std::vector<int64_t> &stop,
+                     const std::vector<int64_t> &step, const std::vector<int64_t> &input_shape) {
+  if (start.size() != stop.size() || start.size() != step.size() || start.size() > input_shape.size()) {
+    MS_LOG(EXCEPTION)
+      << "TensorCopySlices requires the length of begin, stride and end must be equal and less than input dimension.";
+  }
+
+  size_t size = start.size();
+  for (size_t i = 0; i < size; ++i) {
+    if (stop[i] <= start[i]) {
+      MS_LOG(EXCEPTION) << "Invalid slice: (" << start[i] << ", " << stop[i] << " ," << step[i] << ")";
+    }
+    // Operator need to be generalized in the future. Only support to copy continuous memory now.
+    if (step[i] != 1) {
+      MS_LOG(EXCEPTION) << "The element in step only support 1, but got:" << step;
+    }
+  }
+
+  size_t slice_pos = size;
+  for (size_t i = 0; i < size; ++i) {
+    if (stop[i] - start[i] > 1) {
+      slice_pos = i;
+      break;
+    }
+  }
+
+  for (size_t i = slice_pos + 1; i < size; ++i) {
+    if (stop[i] - start[i] != input_shape[i]) {
+      MS_LOG(EXCEPTION) << "Only support copy continuous memory now. For example tensor[0, 0:100] is fine, "
+                           "but tensor[0:100, 0] is not supported.";
+    }
+  }
+}
+
+size_t GetCopySize(const std::vector<int64_t> &dim_offset, const std::vector<int64_t> &start,
+                   const std::vector<int64_t> &stop) {
+  for (size_t i = 0; i < start.size(); ++i) {
+    if (stop[i] - start[i] != 1) {
+      return (stop[i] - start[i]) * dim_offset[i];
+    }
+  }
+  return dim_offset[start.size() - 1];
+}
+
+std::vector<int64_t> CalDimOffset(const std::vector<int64_t> &input_shape) {
+  std::vector<int64_t> dim_offset;
+  int64_t offset = 1;
+  for (auto iter = input_shape.rbegin(); iter != input_shape.rend(); ++iter) {
+    dim_offset.push_back(offset);
+    offset = offset * (*iter);
+  }
+  std::reverse(dim_offset.begin(), dim_offset.end());
+  return dim_offset;
+}
+
+size_t CalOffset(const std::vector<int64_t> &start, const std::vector<int64_t> &stop, const std::vector<int64_t> &step,
+                 const std::vector<int64_t> &dim_offset) {
+  size_t size = start.size();
+  size_t offset = 0;
+  for (size_t i = 0; i < size; ++i) {
+    offset += dim_offset[i] * start[i];
+    if (stop[i] - start[i] != 1) {
+      break;
+    }
+  }
+  return offset;
+}
 }  // namespace kernel
 }  // namespace mindspore

mindspore/ccsrc/backend/kernel_compiler/common_utils.h

@@ -133,6 +133,15 @@ inline T ComputeLerp(T top_left, T top_right, T bottom_left, T bottom_right, T x
   T bottom = bottom_left + (bottom_right - bottom_left) * x_lerp;
   return top + (bottom - top) * y_lerp;
 }
+
+void CastShapeSizeToLong(const std::vector<size_t> &shape, std::vector<int64_t> *long_shape);
+void CheckSliceValid(const std::vector<int64_t> &start, const std::vector<int64_t> &stop,
+                     const std::vector<int64_t> &step, const std::vector<int64_t> &input_shape);
+size_t CalOffset(const std::vector<int64_t> &start, const std::vector<int64_t> &stop, const std::vector<int64_t> &step,
+                 const std::vector<int64_t> &dim_offset);
+std::vector<int64_t> CalDimOffset(const std::vector<int64_t> &input_shape);
+size_t GetCopySize(const std::vector<int64_t> &dim_offset, const std::vector<int64_t> &start,
+                   const std::vector<int64_t> &stop);
 }  // namespace kernel
 }  // namespace mindspore
 

mindspore/ccsrc/backend/kernel_compiler/cpu/tensor_copy_slices_cpu_kernel.cc

@@ -0,0 +1,70 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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 "backend/kernel_compiler/cpu/tensor_copy_slices_cpu_kernel.h"
+
+#include <functional>
+#include <unordered_map>
+#include "abstract/utils.h"
+#include "backend/kernel_compiler/common_utils.h"
+#include "runtime/device/cpu/cpu_device_address.h"
+
+namespace mindspore {
+namespace kernel {
+void TensorCopySlicesCPUKernel::InitKernel(const CNodePtr &kernel_node) {
+  auto input_shape = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 0);
+  auto update_shape = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 1);
+  auto output_shape = AnfAlgo::GetOutputInferShape(kernel_node, 0);
+
+  CastShapeSizeToLong(input_shape, &input_shape_);
+  CastShapeSizeToLong(update_shape, &update_shape_);
+  CastShapeSizeToLong(output_shape, &output_shape_);
+
+  auto begin = AnfAlgo::GetNodeAttr<std::vector<int64_t>>(kernel_node, BEGIN);
+  auto end = AnfAlgo::GetNodeAttr<std::vector<int64_t>>(kernel_node, END);
+  auto stride = AnfAlgo::GetNodeAttr<std::vector<int64_t>>(kernel_node, STRIDES);
+  CheckSliceValid(begin, end, stride, input_shape_);
+
+  data_type_ = AnfAlgo::GetInputDeviceDataType(kernel_node, 0);
+  auto dim_offset = CalDimOffset(input_shape_);
+  auto type_size = abstract::TypeIdSize(data_type_);
+  offset_ = CalOffset(begin, end, stride, dim_offset) * type_size;
+  copy_size_ = GetCopySize(dim_offset, begin, end) * type_size;
+}
+
+bool TensorCopySlicesCPUKernel::Launch(const std::vector<kernel::AddressPtr> &inputs,
+                                       const std::vector<kernel::AddressPtr> & /*workspace*/,
+                                       const std::vector<kernel::AddressPtr> &outputs) {
+  if (inputs.size() != 2 || outputs.size() != 1) {
+    MS_LOG(ERROR) << "TensorCopySlices requires 1 input and 1 output, but got " << inputs.size() << " input and "
+                  << outputs.size() << " output.";
+    return false;
+  }
+
+  auto input_addr = reinterpret_cast<uint8_t *>(inputs[0]->addr);
+  auto update_addr = reinterpret_cast<uint8_t *>(inputs[1]->addr);
+  auto output_addr = reinterpret_cast<uint8_t *>(outputs[0]->addr);
+
+  if (memcpy_s(output_addr, outputs[0]->size, input_addr, inputs[0]->size) != EOK) {
+    MS_LOG(EXCEPTION) << "TensorCopySlices memcpy input failed";
+  }
+  if (memcpy_s(output_addr + offset_, copy_size_, update_addr, copy_size_) != EOK) {
+    MS_LOG(EXCEPTION) << "TensorCopySlices memcpy update failed";
+  }
+  return true;
+}
+}  // namespace kernel
+}  // namespace mindspore

mindspore/ccsrc/backend/kernel_compiler/cpu/tensor_copy_slices_cpu_kernel.h

@@ -0,0 +1,51 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#ifndef MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_TENSOR_COPY_SLICES_CPU_KERNEL_H_
+#define MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_TENSOR_COPY_SLICES_CPU_KERNEL_H_
+
+#include <vector>
+#include <memory>
+
+#include "backend/kernel_compiler/cpu/cpu_kernel.h"
+#include "backend/kernel_compiler/cpu/cpu_kernel_factory.h"
+#include "nnacl/fp32/strided_slice_fp32.h"
+
+namespace mindspore {
+namespace kernel {
+class TensorCopySlicesCPUKernel : public CPUKernel {
+ public:
+  TensorCopySlicesCPUKernel() = default;
+  ~TensorCopySlicesCPUKernel() override = default;
+
+  void InitKernel(const CNodePtr &kernel_node) override;
+  bool Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &workspace,
+              const std::vector<AddressPtr> &outputs) override;
+
+ private:
+  TypeId data_type_;
+  size_t offset_;
+  size_t copy_size_;
+  std::vector<int64_t> input_shape_;
+  std::vector<int64_t> update_shape_;
+  std::vector<int64_t> output_shape_;
+};
+
+MS_REG_CPU_KERNEL(TensorCopySlices, KernelAttr(), TensorCopySlicesCPUKernel);
+}  // namespace kernel
+}  // namespace mindspore
+
+#endif  // MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_CPU_TENSOR_COPY_SLICES_CPU_KERNEL_H_

mindspore/ccsrc/backend/kernel_compiler/gpu/arrays/tensor_copy_slices_gpu_kernel.cc

@@ -0,0 +1,50 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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 "backend/kernel_compiler/gpu/arrays/tensor_copy_slices_gpu_kernel.h"
+
+namespace mindspore {
+namespace kernel {
+MS_REG_GPU_KERNEL_ONE(
+  TensorCopySlices,
+  KernelAttr().AddInputAttr(kNumberTypeFloat64).AddInputAttr(kNumberTypeFloat64).AddOutputAttr(kNumberTypeFloat64),
+  TensorCopySlicesGpuKernel, double)
+MS_REG_GPU_KERNEL_ONE(
+  TensorCopySlices,
+  KernelAttr().AddInputAttr(kNumberTypeFloat32).AddInputAttr(kNumberTypeFloat32).AddOutputAttr(kNumberTypeFloat32),
+  TensorCopySlicesGpuKernel, float)
+MS_REG_GPU_KERNEL_ONE(
+  TensorCopySlices,
+  KernelAttr().AddInputAttr(kNumberTypeFloat16).AddInputAttr(kNumberTypeFloat16).AddOutputAttr(kNumberTypeFloat16),
+  TensorCopySlicesGpuKernel, half)
+MS_REG_GPU_KERNEL_ONE(
+  TensorCopySlices,
+  KernelAttr().AddInputAttr(kNumberTypeInt64).AddInputAttr(kNumberTypeInt64).AddOutputAttr(kNumberTypeInt64),
+  TensorCopySlicesGpuKernel, int64_t)
+MS_REG_GPU_KERNEL_ONE(
+  TensorCopySlices,
+  KernelAttr().AddInputAttr(kNumberTypeInt32).AddInputAttr(kNumberTypeInt32).AddOutputAttr(kNumberTypeInt32),
+  TensorCopySlicesGpuKernel, int)
+MS_REG_GPU_KERNEL_ONE(
+  TensorCopySlices,
+  KernelAttr().AddInputAttr(kNumberTypeInt8).AddInputAttr(kNumberTypeInt8).AddOutputAttr(kNumberTypeInt8),
+  TensorCopySlicesGpuKernel, char)
+MS_REG_GPU_KERNEL_ONE(
+  TensorCopySlices,
+  KernelAttr().AddInputAttr(kNumberTypeUInt8).AddInputAttr(kNumberTypeUInt8).AddOutputAttr(kNumberTypeUInt8),
+  TensorCopySlicesGpuKernel, uchar)
+}  // namespace kernel
+}  // namespace mindspore

mindspore/ccsrc/backend/kernel_compiler/gpu/arrays/tensor_copy_slices_gpu_kernel.h

@@ -0,0 +1,140 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#ifndef MINDSPORE_MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_ARRAYS_TENSOR_STRIDE_UPDATE_GPU_KERNEL_H_
+#define MINDSPORE_MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_ARRAYS_TENSOR_STRIDE_UPDATE_GPU_KERNEL_H_
+
+#include <algorithm>
+#include <vector>
+#include "backend/kernel_compiler/gpu/gpu_kernel.h"
+#include "backend/kernel_compiler/gpu/gpu_kernel_factory.h"
+#include "backend/kernel_compiler/common_utils.h"
+
+namespace mindspore {
+namespace kernel {
+template <typename T>
+class TensorCopySlicesGpuKernel : public GpuKernel {
+ public:
+  TensorCopySlicesGpuKernel() : input_size_(0), update_size_(0), output_size_(0), offset_(0), copy_size_(0) {}
+  ~TensorCopySlicesGpuKernel() {}
+
+  bool Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &workspace,
+              const std::vector<AddressPtr> &outputs, void *stream_ptr) override {
+    T *input_addr = GetDeviceAddress<T>(inputs, 0);
+    T *update_addr = GetDeviceAddress<T>(inputs, 1);
+    T *output_addr = GetDeviceAddress<T>(outputs, 0);
+
+    if (inputs[1]->size != copy_size_) {
+      MS_LOG(EXCEPTION) << "Invalid update size:" << inputs[1]->size << " copy_size_:" << copy_size_;
+    }
+
+    CHECK_CUDA_RET_WITH_EXCEPT(kernel_node_,
+                               cudaMemcpyAsync(output_addr, input_addr, inputs[0]->size, cudaMemcpyDeviceToDevice,
+                                               reinterpret_cast<cudaStream_t>(stream_ptr)),
+                               "TensorCopySlices cudaMemcpyAsync outputs failed");
+
+    CHECK_CUDA_RET_WITH_EXCEPT(kernel_node_,
+                               cudaMemcpyAsync(output_addr + offset_, update_addr, inputs[1]->size,
+                                               cudaMemcpyDeviceToDevice, reinterpret_cast<cudaStream_t>(stream_ptr)),
+                               "TensorCopySlices cudaMemcpyAsync outputs failed");
+
+    return true;
+  }
+
+  const std::vector<size_t> &GetInputSizeList() const override { return input_size_list_; }
+  const std::vector<size_t> &GetOutputSizeList() const override { return output_size_list_; }
+  const std::vector<size_t> &GetWorkspaceSizeList() const override { return workspace_size_list_; }
+
+  bool Init(const CNodePtr &kernel_node) override {
+    kernel_node_ = kernel_node;
+
+    size_t input_num = AnfAlgo::GetInputTensorNum(kernel_node);
+    if (input_num != 2) {
+      MS_LOG(ERROR) << "Input number is " << input_num << ", but TensorCopySlices needs 2 inputs.";
+      return false;
+    }
+
+    size_t output_num = AnfAlgo::GetOutputTensorNum(kernel_node);
+    if (output_num != 1) {
+      MS_LOG(ERROR) << "Output number is " << output_num << ", but TensorCopySlices has 1 output.";
+      return false;
+    }
+
+    auto input_shapes = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 0);
+    auto update_shapes = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 1);
+    auto output_shapes = AnfAlgo::GetOutputInferShape(kernel_node, 0);
+
+    CastShapeSizeToLong(input_shapes, &input_shapes_);
+    CastShapeSizeToLong(update_shapes, &update_shapes_);
+    CastShapeSizeToLong(output_shapes, &output_shapes_);
+
+    GetSize();
+    InitSizeLists();
+
+    auto begin = GetAttr<std::vector<int64_t>>(kernel_node, kAttrBegin);
+    auto end = GetAttr<std::vector<int64_t>>(kernel_node, kAttrEnd);
+    auto strides = GetAttr<std::vector<int64_t>>(kernel_node, kAttrStrides);
+
+    CheckSliceValid(begin, end, strides, input_shapes_);
+    auto dim_offset = CalDimOffset(input_shapes_);
+    offset_ = CalOffset(begin, end, strides, dim_offset);
+    copy_size_ = GetCopySize(dim_offset, begin, end) * sizeof(T);
+    return true;
+  }
+
+ protected:
+  void GetSize() {
+    input_size_ = sizeof(T);
+    for (size_t i = 0; i < input_shapes_.size(); i++) {
+      input_size_ *= LongToSize(input_shapes_[i]);
+    }
+
+    update_size_ = sizeof(T);
+    for (size_t i = 0; i < update_shapes_.size(); i++) {
+      update_size_ *= LongToSize(update_shapes_[i]);
+    }
+    output_size_ = sizeof(T);
+    for (size_t i = 0; i < output_shapes_.size(); i++) {
+      output_size_ *= LongToSize(output_shapes_[i]);
+    }
+  }
+
+  void InitSizeLists() override {
+    input_size_list_.push_back(input_size_);
+    input_size_list_.push_back(update_size_);
+    output_size_list_.push_back(output_size_);
+    return;
+  }
+
+ private:
+  std::vector<size_t> input_size_list_;
+  std::vector<size_t> output_size_list_;
+  std::vector<size_t> workspace_size_list_;
+
+  std::vector<int64_t> input_shapes_;
+  std::vector<int64_t> update_shapes_;
+  std::vector<int64_t> output_shapes_;
+
+  size_t input_size_;
+  size_t update_size_;
+  size_t output_size_;
+
+  size_t offset_;
+  size_t copy_size_;
+};
+}  // namespace kernel
+}  // namespace mindspore
+#endif  // MINDSPORE_MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_ARRAYS_TENSOR_STRIDE_UPDATE_GPU_KERNEL_H_

mindspore/ccsrc/backend/kernel_compiler/rts/memcpy_async.cc

@@ -154,10 +154,10 @@ device::DynamicKernelPtr MemCpyAsyncKernel::GenDynamicKernel(const CNodePtr &cno
                                                   kernel_inputs[0]->size);
 }
 
-const std::vector<TypeId> data_type_list{kNumberTypeInt,     kNumberTypeInt8,    kNumberTypeInt16, kNumberTypeInt32,
-                                         kNumberTypeInt64,   kNumberTypeUInt,    kNumberTypeUInt8, kNumberTypeUInt16,
-                                         kNumberTypeUInt32,  kNumberTypeUInt64,  kNumberTypeFloat, kNumberTypeFloat16,
-                                         kNumberTypeFloat32, kNumberTypeFloat64, kNumberTypeBool};
+const std::vector<TypeId> data_type_list = {
+  kNumberTypeInt,   kNumberTypeInt8,    kNumberTypeInt16,   kNumberTypeInt32,   kNumberTypeInt64,
+  kNumberTypeUInt,  kNumberTypeUInt8,   kNumberTypeUInt16,  kNumberTypeUInt32,  kNumberTypeUInt64,
+  kNumberTypeFloat, kNumberTypeFloat16, kNumberTypeFloat32, kNumberTypeFloat64, kNumberTypeBool};
 const std::vector<std::string> format_list = {kOpFormat_DEFAULT,  kOpFormat_NCHW,   kOpFormat_NHWC,
                                               kOpFormat_NC1HWC0,  kOpFormat_FRAC_Z, kOpFormat_NC1KHKWHWC0,
                                               kOpFormat_C1HWNCoC0};

mindspore/ccsrc/backend/kernel_compiler/rts/tensor_copy_slices.cc

@@ -0,0 +1,187 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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 "backend/kernel_compiler/rts/tensor_copy_slices.h"
+#include <memory>
+#include <numeric>
+#include <functional>
+#include <string>
+#include "abstract/utils.h"
+#include "backend/session/anf_runtime_algorithm.h"
+#include "backend/kernel_compiler/common_utils.h"
+#include "common/trans.h"
+#include "runtime/mem.h"
+#include "runtime/device/kernel_runtime.h"
+#include "utils/ms_context.h"
+
+using mindspore::ge::model_runner::MemcpyAsyncTaskInfo;
+namespace mindspore {
+namespace kernel {
+TensorCopySlices::TensorCopySlices() {}
+
+TensorCopySlices::~TensorCopySlices() {}
+
+bool TensorCopySlices::Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> & /*workspace*/,
+                              const std::vector<AddressPtr> &outputs, void *stream_ptr) {
+  if (inputs.size() != 2) {
+    MS_LOG(ERROR) << "inputs size is not 2";
+    return false;
+  }
+  if (outputs.size() != 1) {
+    MS_LOG(ERROR) << "outputs size is not 1";
+    return false;
+  }
+  if (outputs[0]->size != inputs[0]->size) {
+    MS_LOG(ERROR) << "TensorCopySlices destMax > src size";
+    return false;
+  }
+
+  auto status = rtMemcpyAsync(outputs[0]->addr, outputs[0]->size, inputs[0]->addr, inputs[0]->size,
+                              RT_MEMCPY_DEVICE_TO_DEVICE, stream_ptr);
+  if (status != RT_ERROR_NONE) {
+    MS_LOG(ERROR) << "MemCpyAsync op rtMemcpyAsync failed!";
+    return false;
+  }
+  status = rtMemcpyAsync(VoidPointerOffset(outputs[0]->addr, offset_), copy_size_, inputs[1]->addr, copy_size_,
+                         RT_MEMCPY_DEVICE_TO_DEVICE, stream_ptr);
+  if (status != RT_ERROR_NONE) {
+    MS_LOG(ERROR) << "MemCpyAsync op rtMemcpyAsync failed!";
+    return false;
+  }
+  return true;
+}
+
+bool TensorCopySlices::Init(const mindspore::AnfNodePtr &anf_node) {
+  MS_EXCEPTION_IF_NULL(anf_node);
+  GetInputOutputInfo(anf_node);
+  GetInputOutputTotalCount(anf_node);
+
+  auto begin = AnfAlgo::GetNodeAttr<std::vector<int64_t>>(anf_node, kAttrBegin);
+  auto end = AnfAlgo::GetNodeAttr<std::vector<int64_t>>(anf_node, kAttrEnd);
+  auto strides = AnfAlgo::GetNodeAttr<std::vector<int64_t>>(anf_node, kAttrStrides);
+
+  CheckSliceValid(begin, end, strides, input_shape_);
+  auto dim_offset = CalDimOffset(input_shape_);
+  offset_ = CalOffset(begin, end, strides, dim_offset) * abstract::TypeIdSize(input_type_id_);
+  copy_size_ = GetCopySize(dim_offset, begin, end) * abstract::TypeIdSize(input_type_id_);
+  return true;
+}
+
+void TensorCopySlices::GetInputOutputInfo(const AnfNodePtr &anf_node) {
+  MS_EXCEPTION_IF_NULL(anf_node);
+  size_t input_size = AnfAlgo::GetInputTensorNum(anf_node);
+  if (input_size != 2) {
+    MS_LOG(EXCEPTION) << "TensorCopySlices input size is not 2";
+  }
+  input_type_id_ = AnfAlgo::GetPrevNodeOutputDeviceDataType(anf_node, 0);
+  update_type_id_ = AnfAlgo::GetPrevNodeOutputDeviceDataType(anf_node, 0);
+  output_type_id_ = AnfAlgo::GetOutputDeviceDataType(anf_node, 0);
+  if (input_type_id_ != output_type_id_ || input_type_id_ != update_type_id_) {
+    MS_LOG(EXCEPTION) << "Input and output of TensorCopySlices is not same, input type:" << input_type_id_
+                      << " output_type_id_:" << output_type_id_;
+  }
+
+  auto input_shape = AnfAlgo::GetPrevNodeOutputInferShape(anf_node, 0);
+  auto update_shape = AnfAlgo::GetPrevNodeOutputInferShape(anf_node, 1);
+  auto output_shape = AnfAlgo::GetOutputDeviceShape(anf_node, 0);
+  CastShapeSizeToLong(input_shape, &input_shape_);
+  CastShapeSizeToLong(update_shape, &update_shape_);
+  CastShapeSizeToLong(output_shape, &output_shape_);
+}
+
+void *TensorCopySlices::VoidPointerOffset(void *ptr, size_t offset) {
+  return reinterpret_cast<uint8_t *>(ptr) + offset;
+}
+
+void TensorCopySlices::GetInputOutputTotalCount(const AnfNodePtr &anf_node) {
+  MS_EXCEPTION_IF_NULL(anf_node);
+  size_t input_size = AnfAlgo::GetInputTensorNum(anf_node);
+  if (input_size != 2) {
+    MS_LOG(EXCEPTION) << "TensorCopySlices input size is not 2";
+  }
+
+  auto input_shape = AnfAlgo::GetInputDeviceShape(anf_node, 0);
+  size_t total_size = std::accumulate(input_shape.begin(), input_shape.end(), 1, std::multiplies<>());
+  total_size *= abstract::TypeIdSize(input_type_id_);
+  MS_LOG(INFO) << "TensorCopySlices size[" << total_size << "]";
+  // Shape and DType of input0 and output0 are same.
+  input_size_list_.emplace_back(total_size);
+  output_size_list_.emplace_back(total_size);
+
+  auto update_shape = AnfAlgo::GetInputDeviceShape(anf_node, 1);
+  size_t update_size = std::accumulate(update_shape.begin(), update_shape.end(), 1, std::multiplies<>());
+  update_size *= abstract::TypeIdSize(update_type_id_);
+  input_size_list_.emplace_back(update_size);
+}
+
+std::vector<TaskInfoPtr> TensorCopySlices::GenTask(const std::vector<AddressPtr> &inputs,
+                                                   const std::vector<AddressPtr> &,
+                                                   const std::vector<AddressPtr> &outputs, uint32_t stream_id) {
+  if (inputs.size() != 2) {
+    MS_LOG(EXCEPTION) << "inputs size is not 2.";
+  }
+  if (outputs.size() != 1) {
+    MS_LOG(EXCEPTION) << "outputs size is not 1.";
+  }
+  if (outputs[0]->size != inputs[0]->size) {
+    MS_LOG(EXCEPTION) << "TensorCopySlices input size and output size not equal.";
+  }
+
+  stream_id_ = stream_id;
+  std::shared_ptr<MemcpyAsyncTaskInfo> task_info_ptr1 =
+    std::make_shared<MemcpyAsyncTaskInfo>(kernel_name_, stream_id, outputs[0]->addr, outputs[0]->size, inputs[0]->addr,
+                                          inputs[0]->size, RT_MEMCPY_DEVICE_TO_DEVICE, NeedDump());
+  std::shared_ptr<MemcpyAsyncTaskInfo> task_info_ptr2 = std::make_shared<MemcpyAsyncTaskInfo>(
+    kernel_name_, stream_id, VoidPointerOffset(outputs[0]->addr, offset_), copy_size_, inputs[1]->addr, copy_size_,
+    RT_MEMCPY_DEVICE_TO_DEVICE, NeedDump());
+  return {task_info_ptr1, task_info_ptr2};
+}
+
+const std::vector<TypeId> data_type_list = {
+  kNumberTypeInt,   kNumberTypeInt8,    kNumberTypeInt16,   kNumberTypeInt32,   kNumberTypeInt64,
+  kNumberTypeUInt,  kNumberTypeUInt8,   kNumberTypeUInt16,  kNumberTypeUInt32,  kNumberTypeUInt64,
+  kNumberTypeFloat, kNumberTypeFloat16, kNumberTypeFloat32, kNumberTypeFloat64, kNumberTypeBool};
+// If input's format is 5D, we will insert TransData before TensorCopySlices.
+const std::vector<std::string> format_list = {kOpFormat_DEFAULT, kOpFormat_NCHW, kOpFormat_NHWC};
+
+TensorCopySlicesDesc::TensorCopySlicesDesc() {}
+
+TensorCopySlicesDesc::~TensorCopySlicesDesc() {}
+
+// TensorCopySlices Register
+std::vector<std::shared_ptr<kernel::KernelBuildInfo>> TensorCopySlicesDesc::GetKernelInfo() {
+  std::vector<std::shared_ptr<kernel::KernelBuildInfo>> tensor_copy_slices_build_info{};
+  for (const auto &format : format_list) {
+    for (const auto &type : data_type_list) {
+      auto builder = KernelBuildInfo::KernelBuildInfoBuilder();
+      vector<string> input_format{format, format};
+      vector<TypeId> input_type{type, type};
+      vector<string> output_format{format};
+      vector<TypeId> output_type{type};
+      builder.SetInputsFormat(input_format);
+      builder.SetInputsDeviceType(input_type);
+      builder.SetOutputsFormat(output_format);
+      builder.SetOutputsDeviceType(output_type);
+      builder.SetProcessor(AICORE);
+      builder.SetKernelType(RT_KERNEL);
+      builder.SetFusionType(OPAQUE);
+      tensor_copy_slices_build_info.emplace_back(builder.Build());
+    }
+  }
+  return tensor_copy_slices_build_info;
+}
+}  // namespace kernel
+}  // namespace mindspore

mindspore/ccsrc/backend/kernel_compiler/rts/tensor_copy_slices.h

@@ -0,0 +1,66 @@
+/**
+ * Copyright 2021 Huawei Technologies Co., Ltd
+ *
+ * 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.
+ */
+
+#ifndef MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_RTS_TENSOR_COPY_SLICES_H
+#define MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_RTS_TENSOR_COPY_SLICES_H
+
+#include <vector>
+#include <memory>
+#include "backend/kernel_compiler/rts/rt_kernel.h"
+#include "backend/kernel_compiler/rts/rt_kernel_info.h"
+
+namespace mindspore {
+namespace kernel {
+class TensorCopySlices : public RtKernel {
+ public:
+  TensorCopySlices();
+  ~TensorCopySlices() override;
+
+  bool Init(const AnfNodePtr &anf_node) override;
+  bool Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &workspace,
+              const std::vector<AddressPtr> &outputs, void *stream_ptr) override;
+  std::vector<TaskInfoPtr> GenTask(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &workspace,
+                                   const std::vector<AddressPtr> &outputs, uint32_t stream_id) override;
+
+ private:
+  void GetInputOutputInfo(const AnfNodePtr &anf_node);
+  void GetInputOutputTotalCount(const AnfNodePtr &anf_node);
+  void *VoidPointerOffset(void *ptr, size_t offset);
+
+  std::vector<int64_t> input_shape_;
+  std::vector<int64_t> update_shape_;
+  std::vector<int64_t> output_shape_;
+  TypeId input_type_id_{};
+  TypeId output_type_id_{};
+  TypeId update_type_id_{};
+
+  size_t offset_{0};
+  size_t copy_size_{0};
+};
+
+class TensorCopySlicesDesc : public RtKerDesc {
+ public:
+  TensorCopySlicesDesc();
+  ~TensorCopySlicesDesc() override;
+  std::vector<std::shared_ptr<kernel::KernelBuildInfo>> GetKernelInfo() override;
+};
+
+MS_REG_RTKERNEL_DESC(tensorcopyslices, TensorCopySlicesDesc);
+MS_REG_RTKERNEL(tensorcopyslices, TensorCopySlices);
+}  // namespace kernel
+}  // namespace mindspore
+
+#endif  // MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_RTS_TENSOR_COPY_SLICES_H

mindspore/ccsrc/backend/optimizer/common/const_input_to_attr_registry.cc

@@ -66,6 +66,7 @@ ConstInputToAttrInfoRegistry::ConstInputToAttrInfoRegistry() {
   Register(kStridedSliceAssignOpName, {1, 2, 3});
   Register(kStridedSliceOpName, {1, 2, 3});
   Register(kStridedSliceGradOpName, {1, 2, 3, 4});
+  Register(kTensorCopySlicesOpName, {2, 3, 4});
   Register(kFlattenGradOpName, {1});
   Register(kExpandDimsOpName, {1});
   Register(kSplitOpName, {0});

mindspore/ccsrc/utils/utils.h

@@ -215,6 +215,7 @@ constexpr auto kSparseApplyRMSPropOpName = "SparseApplyRMSProp";
 constexpr auto kSparseApplyAdadeltaOpName = "SparseApplyAdadelta";
 constexpr auto kApplyAdamWithAmsgradOpName = "ApplyAdamWithAmsgrad";
 constexpr auto kTensorMoveOpName = "TensorMove";
+constexpr auto kTensorCopySlicesOpName = "TensorCopySlices";
 constexpr auto kTensorScatterUpdateOpName = "TensorScatterUpdate";
 constexpr auto kScatterNdUpdateOpName = "ScatterNdUpdate";
 constexpr auto kPushOpName = "Push";
@@ -406,6 +407,7 @@ constexpr auto kAttrPaddings = "paddings";
 constexpr auto kAttrNumSegments = "num_segments";
 constexpr auto kAttrStackOpName = "stack_op_name";
 constexpr auto kAttrBegin = "begin";
+constexpr auto kAttrEnd = "end";
 constexpr auto kAttrSize = "size";
 constexpr auto kAttrIsDynamicShape = "is_dynamic_shape";
 constexpr auto kAttrInputIsDynamicShape = "input_is_dynamic_shape";

mindspore/core/abstract/infer_functions.h

@@ -279,6 +279,8 @@ AbstractBasePtr InferImplMaskedSelect(const AnalysisEnginePtr &, const Primitive
                                       const AbstractBasePtrList &args_spec_list);
 AbstractBasePtr InferImplTransData(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
                                    const AbstractBasePtrList &args_spec_list);
+AbstractBasePtr InferImplTensorCopySlices(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
+                                          const AbstractBasePtrList &args_spec_list);
 template <typename T>
 AbstractBasePtr InferTupleOrListOrDictLen(const std::string &op_name, const AbstractBasePtrList &args_spec_list) {
   // Inputs: a tuple or list or dict.

mindspore/core/abstract/prim_arrays.cc

@@ -1214,5 +1214,14 @@ AbstractBasePtr InferImplDynamicStitch(const AnalysisEnginePtr &, const Primitiv
   return std::make_shared<AbstractTensor>(infer_type,
                                           std::make_shared<abstract::Shape>(out_shape, min_shape, max_shape));
 }
+
+AbstractBasePtr InferImplTensorCopySlices(const AnalysisEnginePtr &, const PrimitivePtr &primitive,
+                                          const AbstractBasePtrList &args_spec_list) {
+  auto &op_name = primitive->name();
+  CheckArgsSize(op_name, args_spec_list, 5);
+  AbstractTensorPtr input = CheckArg<AbstractTensor>(op_name, args_spec_list, 0);
+  return std::make_shared<AbstractTensor>(input->element(), input->shape());
+}
+
 }  // namespace abstract
 }  // namespace mindspore

mindspore/core/abstract/primitive_infer_map.cc

@@ -107,6 +107,7 @@ PrimitiveEvalImplMap &GetPrimitiveToEvalImplMap() {
     {prim::kPrimSequenceMask, {InferImplSequenceMask, nullptr, true}},
     {prim::kPrimSort, {InferImplSort, nullptr, true}},
     {prim::kPrimMaskedSelect, {InferImplMaskedSelect, nullptr, true}},
+    {prim::kPrimTensorCopySlices, {InferImplTensorCopySlices, nullptr, true}},
     // Structure
     {prim::kPrimMakeTuple, {InferImplMakeTuple, nullptr, true}},
     {prim::kPrimMakeList, {InferImplMakeList, nullptr, true}},

mindspore/core/base/core_ops.h

@@ -212,6 +212,7 @@ inline const PrimitivePtr kPrimDynamicGRUV2 = std::make_shared<Primitive>("Dynam
 inline const PrimitivePtr kPrimDynamicGRUV2Grad = std::make_shared<Primitive>("DynamicGRUV2Grad");
 inline const PrimitivePtr kPrimScatterAdd = std::make_shared<Primitive>("ScatterAdd");
 inline const PrimitivePtr kPrimScatterUpdate = std::make_shared<Primitive>("ScatterUpdate");
+inline const PrimitivePtr kPrimTensorCopySlices = std::make_shared<Primitive>("TensorCopySlices");
 inline const PrimitivePtr kPrimMapUniform = std::make_shared<Primitive>("MapUniform");
 inline const PrimitivePtr kPrimSplit = std::make_shared<Primitive>("Split");
 inline const PrimitivePtr kPrimSequenceMask = std::make_shared<Primitive>("SequenceMask");

mindspore/ops/_grad_experimental/__init__.py

@@ -15,5 +15,6 @@
 
 """grad experimental impl."""
 from .._grad.grad_base import get_bprop_fn
+from . import grad_inner_ops
 
 __all__ = ['get_bprop_fn']

mindspore/ops/_grad_experimental/grad_inner_ops.py

@@ -0,0 +1,32 @@
+# Copyright 2021 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+
+"""inner_ops"""
+
+from .._grad.grad_base import bprop_getters
+from ..operations import _inner_ops as inner
+from .. import functional as F
+
+@bprop_getters.register(inner.TensorCopySlices)
+def get_bprop_tensor_copy_slices(self):
+    """Generate bprop for TensorCopySlices"""
+    tensor_copy_slices = inner.TensorCopySlices()
+
+    def bprop(x, update, begin, end, stride, out, dout):
+        x_grad = tensor_copy_slices(dout, F.zeros_like(update))
+        update_grad = F.strided_slice(dout, begin, end, stride)
+        return x_grad, update_grad, F.zeros_like(begin), F.zeros_like(end), F.zeros_like(stride)
+
+    return bprop

mindspore/ops/_op_impl/cpu/__init__.py

@@ -61,3 +61,4 @@ from .add import _add_cpu
 from .one_hot import _one_hot_cpu
 from .pad import _pad_cpu
 from .range import _range_cpu
+from .tensor_copy_slices import _tensor_copy_slices_cpu

mindspore/ops/_op_impl/cpu/tensor_copy_slices.py

@@ -0,0 +1,41 @@
+# Copyright 2021 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+
+"""TensorCopySlices op"""
+from mindspore.ops.op_info_register import op_info_register, CpuRegOp, DataType
+
+tensor_copy_slices_op_info = CpuRegOp("TensorCopySlices") \
+    .input(0, "x", "required") \
+    .input(1, "value", "required") \
+    .output(0, "y", "required") \
+    .dtype_format(DataType.U8_Default, DataType.U8_Default, DataType.U8_Default) \
+    .dtype_format(DataType.U16_Default, DataType.U16_Default, DataType.U16_Default) \
+    .dtype_format(DataType.U32_Default, DataType.U32_Default, DataType.U32_Default) \
+    .dtype_format(DataType.U64_Default, DataType.U64_Default, DataType.U64_Default) \
+    .dtype_format(DataType.I8_Default, DataType.I8_Default, DataType.I8_Default) \
+    .dtype_format(DataType.I16_Default, DataType.I16_Default, DataType.I16_Default) \
+    .dtype_format(DataType.I32_Default, DataType.I32_Default, DataType.I32_Default) \
+    .dtype_format(DataType.I64_Default, DataType.I64_Default, DataType.I64_Default) \
+    .dtype_format(DataType.F16_Default, DataType.F16_Default, DataType.F16_Default) \
+    .dtype_format(DataType.F32_Default, DataType.F32_Default, DataType.F32_Default) \
+    .dtype_format(DataType.F64_Default, DataType.F64_Default, DataType.F64_Default) \
+    .dtype_format(DataType.BOOL_Default, DataType.BOOL_Default, DataType.BOOL_Default) \
+    .get_op_info()
+
+
+@op_info_register(tensor_copy_slices_op_info)
+def _tensor_copy_slices_cpu():
+    """TensorCopySlices cpu register"""
+    return

mindspore/ops/operations/_inner_ops.py

@@ -1154,3 +1154,42 @@ class DynamicBroadcastGradientArgs(Primitive):
     @prim_attr_register
     def __init__(self):
         """Init BroadcastGradientArgs"""
+
+
+class TensorCopySlices(Primitive):
+    """
+    Copy continues memory.
+
+    Inputs:
+        - **x** (Tensor) - The target Tensor.
+        - **value** (Tensor) - The tensor to update x.
+        - **begin** (tuple[int]) - A tuple which represents the location where to start. Only
+          constant value is allowed.
+        - **end** (tuple[int]) - A tuple or which represents the maximum location where to end.
+          Only constant value is allowed.
+        - **strides** (tuple[int]) - A tuple which represents the stride is continuously added
+          before reaching the maximum location. Only constant value is allowed.
+
+    Outputs:
+        - **y** (Tensor), has the same shape and data type of x.
+
+    Examples:
+        >>> import numpy as np
+        >>> from mindspore.ops.operations import _inner_ops
+        >>> copy_slices = _inner_ops.TensorCopySlices()
+        >>> out = copy_slices(Tensor(np.zeros((5, 5))), Tensor(np.ones((2, 5))), (3, 0), (5, 5), (1, 1))
+        >>> print(out)
+            [[1., 1., 1., 1., 1.],
+             [1., 1., 1., 1., 1.],
+             [1., 1., 1., 1., 1.],
+             [0., 0., 0., 0., 0.],
+             [0., 0., 0., 0., 0.]]
+
+    Supported Platforms:
+        ``Ascend`` ``GPU`` ``CPU``
+    """
+
+    @prim_attr_register
+    def __init__(self):
+        """Initialize TensorScatterUpdate"""
+        self.init_prim_io_names(inputs=['x', 'value', 'begin', 'end', 'strides'], outputs=['y'])

tests/st/ops/ascend/test_aicpu_ops/test_tensor_copy_slices.py

@@ -0,0 +1,105 @@
+# Copyright 2021 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+
+import numpy as np
+import pytest
+import mindspore.context as context
+import mindspore.nn as nn
+from mindspore import Tensor
+from mindspore.ops.operations import _inner_ops
+
+class Net(nn.Cell):
+    def __init__(self):
+        super(Net, self).__init__()
+        self.copy_slices = _inner_ops.TensorCopySlices()
+
+    def construct(self, input_x, update, begin, end, strides):
+        return self.copy_slices(input_x, update, begin, end, strides)
+
+def convert_begin_end_strides_to_slice(begin, end, strides):
+    result = []
+    for x, y, z in zip(begin, end, strides):
+        result.append(slice(x, y, z))
+    return tuple(result)
+
+def test_tensor_copy_slices_net(input_shape, update_shape, begin, end, strides, dtype):
+    input_np = np.zeros(input_shape, dtype)
+    update_np = np.ones(update_shape, dtype)
+    input_tensor = Tensor(input_np)
+    update = Tensor(update_np)
+    net = Net()
+    output = net(input_tensor, update, begin, end, strides)
+    slices = convert_begin_end_strides_to_slice(begin, end, strides)
+    input_np[slices] = update_np
+    assert (output.asnumpy() == input_np).all()
+
+def test_tensor_copy_slices_net_many_dtype(input_shape, update_shape, begin, end, strides, dtypes):
+    for dtype in dtypes:
+        test_tensor_copy_slices_net(input_shape, update_shape, begin, end, strides, dtype)
+
+support_dtype = (np.int64, np.int32, np.float64, np.float32)
+
+def test_tensor_copy_slices():
+    test_tensor_copy_slices_net_many_dtype((10,), (5,), (0,), (5,), (1,), support_dtype)
+    test_tensor_copy_slices_net_many_dtype((10,), (5,), (5,), (10,), (1,), support_dtype)
+    test_tensor_copy_slices_net_many_dtype((10, 10), (5, 10), (0,), (5,), (1,), support_dtype)
+    test_tensor_copy_slices_net_many_dtype((10, 10), (5, 10), (5,), (10,), (1,), support_dtype)
+    test_tensor_copy_slices_net_many_dtype((10, 10), (5,), (9, 5), (10, 10), (1, 1), support_dtype)
+    test_tensor_copy_slices_net_many_dtype((10, 10, 10), (5, 10), (0, 5, 0), (1, 10, 10), (1, 1, 1,), support_dtype)
+    test_tensor_copy_slices_net_many_dtype((10, 10, 10), (5, 10), (9, 5,), (10, 10,), (1, 1,), support_dtype)
+
+@pytest.mark.level0
+@pytest.mark.platform_arm_ascend_training
+@pytest.mark.platform_x86_ascend_training
+@pytest.mark.env_onecard
+def test_tensor_copy_slices_ascend_graph():
+    context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
+    test_tensor_copy_slices()
+
+@pytest.mark.level0
+@pytest.mark.platform_arm_ascend_training
+@pytest.mark.platform_x86_ascend_training
+@pytest.mark.env_onecard
+def test_tensor_copy_slices_ascend_pynative():
+    context.set_context(mode=context.PYNATIVE_MODE, device_target="Ascend")
+    test_tensor_copy_slices()
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_tensor_copy_slices_gpu_graph():
+    context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
+    test_tensor_copy_slices()
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_tensor_copy_slices_gpu_pynative():
+    context.set_context(mode=context.PYNATIVE_MODE, device_target="GPU")
+    test_tensor_copy_slices()
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_tensor_copy_slices_cpu_graph():
+    context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
+    test_tensor_copy_slices()
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_cpu_training
+@pytest.mark.env_onecard
+def test_tensor_copy_slices_cpu_pynative():
+    context.set_context(mode=context.PYNATIVE_MODE, device_target="CPU")
+    test_tensor_copy_slices()

tests/st/ops/ascend/test_aicpu_ops/test_tensor_setitem.py

@@ -0,0 +1,84 @@
+# Copyright 2021 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+
+import numpy as np
+import mindspore.context as context
+import mindspore.nn as nn
+from mindspore import Tensor
+
+
+class Net(nn.Cell):
+    def construct(self, input_x, update, slice_tuple):
+        input_x[slice_tuple] = update
+        return input_x
+
+
+def test_tensor_setitem_net(input_shape, update_shape, slice_tuple, dtype):
+    input_np = np.zeros(input_shape, dtype)
+    update_np = np.ones(update_shape, dtype)
+    input_tensor = Tensor(input_np)
+    update = Tensor(update_np)
+    net = Net()
+    output = net(input_tensor, update, slice_tuple)
+    input_np[slice_tuple] = update_np
+    assert (output.asnumpy() == input_np).all()
+
+
+def test_tensor_setitem_net_many_dtype(input_shape, update_shape, slice_tuple, dtypes):
+    for dtype in dtypes:
+        test_tensor_setitem_net(input_shape, update_shape, slice_tuple, dtype)
+
+
+support_dtype = (np.int64, np.int32, np.float64, np.float32)
+
+
+def test_tensor_setitem_all():
+    test_tensor_setitem_net_many_dtype((10,), (5,), (slice(0, 5),), support_dtype)
+    test_tensor_setitem_net_many_dtype((10,), (5,), (slice(5, 10),), support_dtype)
+    test_tensor_setitem_net_many_dtype((10, 10), (5, 10), (slice(0, 5),), support_dtype)
+    test_tensor_setitem_net_many_dtype((10, 10), (5, 10), (slice(5, 10),), support_dtype)
+    test_tensor_setitem_net_many_dtype((10, 10), (5,), (9, slice(5, 10)), support_dtype)
+    test_tensor_setitem_net_many_dtype((10, 10, 10), (5, 10), (0, slice(5, 10)), support_dtype)
+    test_tensor_setitem_net_many_dtype((10, 10, 10), (5, 10), (9, slice(5, 10)), support_dtype)
+
+
+def test_tensor_copy_slices_ascend_graph():
+    context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
+    test_tensor_setitem_all()
+
+
+def test_tensor_copy_slices_ascend_pynative():
+    context.set_context(mode=context.PYNATIVE_MODE, device_target="Ascend")
+    test_tensor_setitem_all()
+
+
+def test_tensor_copy_slices_gpu_graph():
+    context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
+    test_tensor_setitem_all()
+
+
+def test_tensor_copy_slices_gpu_pynative():
+    context.set_context(mode=context.PYNATIVE_MODE, device_target="GPU")
+    test_tensor_setitem_all()
+
+
+def test_tensor_copy_slices_cpu_graph():
+    context.set_context(mode=context.GRAPH_MODE, device_target="CPU")
+    test_tensor_setitem_all()
+
+
+def test_tensor_copy_slices_cpu_pynative():
+    context.set_context(mode=context.PYNATIVE_MODE, device_target="CPU")
+    test_tensor_setitem_all()