aclnnAddcmul&aclnnInplaceAddcmul

Supported Products

Product	Supported
Atlas A3 training products/Atlas A3 inference products	√
Atlas A2 training products/Atlas A2 inference products	√
Atlas 200I/500 A2 inference products	×
Atlas inference products	×
Atlas training products	√

Function Description

Description: Performs element-wise multiplication of tensor1 and tensor2, multiplies the result by the scalar value, and then performs element-wise addition of this scaled result with the input self or selfRef.
Formula: $out = self + value \times tensor1 \times tensor2$ When aclnnAddcmul is used, self and out in the formula correspond to those in the first-phase API. When aclnnInplaceAddcmul is used, both self and out in the formula correspond to selfRef in the first-phase API.

Prototype

aclnnAddcmul and aclnnInplaceAddcmul implement the same function in different ways. Select a proper operator based on your requirements.
- aclnnAddcmul: An output tensor object needs to be created to store the computation result.
- aclnnInplaceAddcmul: No output tensor object needs to be created, and the computation result is written in place to the input tensor's memory.

Each operator has two-phase API calls. First, aclnnAddcmulGetWorkspaceSize or aclnnInplaceAddcmulGetWorkspaceSize is called to obtain the workspace size required for computation and the executor covering the operator computation process. Then, aclnnAddcmul or aclnnInplaceAddcmul is called to perform computation.

aclnnStatus aclnnAddcmulGetWorkspaceSize(
  const aclTensor* self, 
  const aclTensor* tensor1, 
  const aclTensor* tensor2,  
  const aclScalar* value, 
  aclTensor*       out, 
  uint64_t*        workspaceSize, 
aclOpExecutor**    executor)

aclnnStatus aclnnAddcmul(
  void*          workspace, 
  uint64_t       workspaceSize, 
  aclOpExecutor* executor, 
  aclrtStream    stream)

aclnnStatus aclnnInplaceAddcmulGetWorkspaceSize(
  const aclTensor* selfRef, 
  const aclTensor* tensor1, 
  const aclTensor* tensor2,  
  const aclScalar* value, 
  uint64_t*        workspaceSize, 
  aclOpExecutor**  executor)

aclnnStatus aclnnInplaceAddcmul(
  void*          workspace, 
  uint64_t       workspaceSize, 
  aclOpExecutor* executor, 
  aclrtStream    stream)

aclnnAddcmulGetWorkspaceSize

Parameters:

Name	Input/Output	Description	Usage Notes	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
self	Input	`self` in the formula.	The data types of `self`, `tensor1`, and `tensor2` follow the type promotion rules (see type promotion rules), and the promoted type must be within the supported input types. The shapes of `self`, `tensor1`, and `tensor2` follow the broadcast relationship.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	ND	0 to 8	√
tensor1	Input	Input `tensor1` in the formula.	The data types of `tensor1`, `self`, and `tensor2` follow the type promotion rules (see type promotion rules), and and the promoted type must be within the supported input types. The shapes of `tensor1`, `self`, and `tensor2` follow the broadcast relationship.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	ND	0 to 8	√
tensor2	Input	Input `tensor2` in the formula.	The data types of `tensor2`, `self`, and `tensor1` follow the type promotion rules (see type promotion rules), and and the promoted type must be within the supported input types.. The shapes of `tensor2`, `self`, and `tensor1` follow the broadcast relationship.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	ND	0 to 8	√
value	Input	Input `value` in the formula.	Its data type can be cast to the data type promoted from `self`, `tensor1`, and `tensor2`. For details, see type casting rules.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	-	-	-
out	Output	`out` in the formula.	The data type promoted from `self`, `tensor1`, and `tensor2` can be cast to the data type of `out`. For details, see type casting rules. The shape of `out` is the same as the shape obtained by broadcasting `self`, `tensor1`, and `tensor2`.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	ND	0 to 8	√
workspaceSize	Output	Size of the workspace required to be allocated on the device.	-	-	-	-	-
executor	Output	Operator executor, covering the operator computation process.	-	-	-	-	-

Atlas training products: The data type cannot be BFLOAT16.

Returns:

aclnnStatus: status code. For details, see aclnn Return Codes.

The first-phase API implements input parameter verification. The following errors may be thrown.

Return	Error Code	Description
ACLNN_ERR_PARAM_NULLPTR	161001	The input `self`, `tensor1`, `tensor2`, `value`, or `out` is a null pointer.
ACLNN_ERR_PARAM_INVALID	161002	The data types or formats of `self`, `tensor1`, and `tensor2` are not supported.
		`self`, `tensor1`, and `tensor2` do not meet the type promotion rules.
		The data type promoted from `self`, `tensor1`, and `tensor2` is not supported.
		The data type promoted from `self`, `tensor1`, and `tensor2` cannot be cast to the specified type of `out`.
		The shape of `self`, `tensor1`, or `tensor2` has more than 8 dimensions.
		The shapes of `self`, `tensor1`, and `tensor2` do not follow the broadcast relationship.
		The shape of `out` is different from the shape obtained by broadcasting `self`, `tensor1`, and `tensor2`

aclnnAddcmul

Parameters:

Name	Input/Output	Description
workspace	Input	Address of the workspace to be allocated on the device.
workspaceSize	Input	Size of the workspace to be allocated on the device, obtained by calling the first-phase API `aclnnAddcmulGetWorkspaceSize`.
executor	Input	Operator executor, covering the operator computation process.
stream	Input	Stream for executing the task.

Returns:

aclnnStatus: status code. For details, see aclnn Return Codes.

aclnnInplaceAddcmulGetWorkspaceSize

Parameters:

Name	Input/Output	Description	Usage Notes	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
selfRef	Input/Output	`self` or `out` in the formula.	The data types of `selfRef`, `tensor1`, and `tensor2` follow the type promotion rules (see type promotion rules), and the promoted data type can be cast to the data type of `selfRef` (see type casting rules). In addition, the promoted type must be within the supported input types. The shapes of selfRef, tensor1, and tensor2 meet the broadcast Broadcast Relationships requirements. The shape must be the same as that obtained after broadcasting selfRef, tensor1, and tensor2.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	ND	0 to 8	√
tensor1	Input	Input `tensor1` in the formula.	The data types of `tensor1`, `selfRef`, and `tensor2` follow the type promotion rules (see type promotion rules), and the promoted type must be within the supported input types. The shapes of `tensor1`, `selfRef`, and `tensor2` follow the broadcast relationship.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	ND	0 to 8	√
tensor2	Input	Input `tensor2` in the formula.	The data types of `tensor2`, `selfRef`, and `tensor1` follow the type promotion rules (see type promotion rules), and and the promoted type must be within the supported input types. The shapes of `tensor2`, `selfRef`, and `tensor1` follow the broadcast relationship.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	ND	0 to 8	√
value	Input	Input `value` in the formula.	Its data type can be cast to the data type promoted from `selfRef`, `tensor1`, and `tensor2`. For details, see type casting rules.	FLOAT, FLOAT16, DOUBLE, BFLOAT16, INT32, INT64, INT8, and UINT8	-	-	-
workspaceSize	Output	Size of the workspace required to be allocated on the device.	-	-	-	-	-
executor	Output	Operator executor, covering the operator computation process.	-	-	-	-	-

Atlas training products: The data type cannot be BFLOAT16.

Returns:

aclnnStatus: status code. For details, see aclnn Return Codes.

The first-phase API implements input parameter verification. The following errors may be thrown.

Return	Error Code	Description
ACLNN_ERR_PARAM_NULLPTR	161001	The input `selfRef`, `tensor1`, `tensor2`, or `value` is a null pointer.
ACLNN_ERR_PARAM_INVALID	161002	The data types or formats of `selfRef`, `tensor1` and `tensor2` are not supported.
		`selfRef`, `tensor1`, and `tensor2` do not meet the type promotion rules.
		The data type promoted from `selfRef`, `tensor1`, and `tensor2` is not supported.
		The data type promoted from the inputs `selfRef`, `tensor1`, and `tensor2` cannot be cast to the specified type of output `selfRef`.
		The shape of `selfRef`, `tensor1`, or `tensor2` has more than 8 dimensions.
		The shapes of `selfRef`, `tensor1`, and `tensor2` do not follow the broadcast relationship.
		The shape of `selfRef` does not match the shape obtained by broadcasting the inputs `selfRef`, `tensor1`, and `tensor2`.

aclnnInplaceAddcmul

Parameters:

Name	Input/Output	Description
workspace	Input	Address of the workspace to be allocated on the device.
workspaceSize	Input	Size of the workspace to be allocated on the device, obtained by calling the first-phase API `aclnnInplaceAddcmulGetWorkspaceSize`.
executor	Input	Operator executor, covering the operator computation process.
stream	Input	Stream for executing the task.

Returns:

aclnnStatus: status code. For details, see aclnn Return Codes.

Constraints

Deterministic computation:
- aclnnAddcmul and aclnnInplaceAddcmul each default to a deterministic implementation.

Example

The following example is for reference only. For details, see Compilation and Running Sample.

aclnnAddcmul

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_addcmul.h"

#define CHECK_RET(cond, return_expr) \
  do {                               \
    if (!(cond)) {                   \
      return_expr;                   \
    }                                \
  } while (0)

#define LOG_PRINT(message, ...)     \
  do {                              \
    printf(message, ##__VA_ARGS__); \
  } while (0)

int64_t GetShapeSize(const std::vector<int64_t>& shape) {
  int64_t shapeSize = 1;
  for (auto i : shape) {
    shapeSize *= i;
  }
  return shapeSize;
}

int Init(int32_t deviceId, aclrtStream* stream) {
  // Boilerplate code for resource initialization.
  auto ret = aclInit(nullptr);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
  ret = aclrtSetDevice(deviceId);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return ret);
  ret = aclrtCreateStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return ret);
  return 0;
}

template <typename T>
int CreateAclTensor(const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr,
                    aclDataType dataType, aclTensor** tensor) {
  auto size = GetShapeSize(shape) * sizeof(T);
  // Call aclrtMalloc to allocate memory on the device.
  auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return ret);
  // Call aclrtMemcpy to copy the data from the host to the device.
  ret = aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return ret);

  // Compute the strides of the contiguous tensor.
  std::vector<int64_t> strides(shape.size(), 1);
  for (int64_t i = shape.size() - 2; i >= 0; i--) {
    strides[i] = shape[i + 1] * strides[i + 1];
  }

  // Call aclCreateTensor to create an aclTensor.
  *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND,
                            shape.data(), shape.size(), *deviceAddr);
  return 0;
}

int main() {
  // 1. Boilerplate code for device/stream initialization. For details, see the ACL API manual.
  // Set the device ID in use.
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);

  // 2. Construct the inputs and outputs based on the API definition.
  std::vector<int64_t> selfShape = {4, 2};
  std::vector<int64_t> tensor1Shape = {4, 2};
  std::vector<int64_t> tensor2Shape = {4, 2};
  std::vector<int64_t> outShape = {4, 2};
  void* selfDeviceAddr = nullptr;
  void* tensor1DeviceAddr = nullptr;
  void* tensor2DeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* tensor1 = nullptr;
  aclTensor* tensor2 = nullptr;
  aclScalar* value = nullptr;
  aclTensor* out = nullptr;

  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> tensor1HostData = {2, 2, 2, 2, 2, 2, 2, 2};
  std::vector<float> tensor2HostData = {2, 2, 2, 2, 2, 2, 2, 2};
  std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};
  float scalarValue = 1.2f;

  // Create a self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create a tensor1 aclTensor.
  ret = CreateAclTensor(tensor1HostData, tensor1Shape, &tensor1DeviceAddr, aclDataType::ACL_FLOAT, &tensor1);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create a tensor2 aclTensor.
  ret = CreateAclTensor(tensor2HostData, tensor2Shape, &tensor2DeviceAddr, aclDataType::ACL_FLOAT, &tensor2);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create a value aclScalar.
  value = aclCreateScalar(&scalarValue, aclDataType::ACL_FLOAT);
  CHECK_RET(value != nullptr, return ret);
  // Create an out aclTensor.
  ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  // 3. Call the CANN operator library API, which needs to be replaced with the actual one.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnAddcmul.
  ret = aclnnAddcmulGetWorkspaceSize(self, tensor1, tensor2, value, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAddcmulGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate device memory based on the workspaceSize calculated by the first-phase API.
  void* workspaceAddr = nullptr;
  if (workspaceSize > 0) {
    ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret);
  }
  // Call the second-phase API of aclnnAddcmul.
  ret = aclnnAddcmul(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAddcmul failed. ERROR: %d\n", ret); return ret);

  // 4. (Boilerplate code) Wait until the task execution is complete.
  ret = aclrtSynchronizeStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);

  // 5. Obtain the output value and copy the result from the device to the host. Modify the code based on the API definition.
  auto size = GetShapeSize(outShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]),
                    outDeviceAddr, size * sizeof(resultData[0]), ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy resultData from device to host failed. ERROR: %d\n", ret);
            return ret);
  for (int64_t i = 0; i < size; i++) {
    LOG_PRINT("resultData[%ld] is: %f\n", i, resultData[i]);
  }

  // 6. Destroy aclTensor and aclScalar. Modify the code based on the API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(tensor1);
  aclDestroyTensor(tensor2);
  aclDestroyTensor(out);
  aclDestroyScalar(value);

  // 7. Free device resources. Modify the code based on the API definition.
  aclrtFree(selfDeviceAddr);
  aclrtFree(tensor1DeviceAddr);
  aclrtFree(tensor2DeviceAddr);
  aclrtFree(outDeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();

  return 0;
}

aclnnInplaceAddcmul

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_addcmul.h"

#define CHECK_RET(cond, return_expr) \
  do {                               \
    if (!(cond)) {                   \
      return_expr;                   \
    }                                \
  } while (0)

#define LOG_PRINT(message, ...)     \
  do {                              \
    printf(message, ##__VA_ARGS__); \
  } while (0)

int64_t GetShapeSize(const std::vector<int64_t>& shape) {
  int64_t shapeSize = 1;
  for (auto i : shape) {
    shapeSize *= i;
  }
  return shapeSize;
}

int Init(int32_t deviceId, aclrtStream* stream) {
  // Boilerplate code for resource initialization.
  auto ret = aclInit(nullptr);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
  ret = aclrtSetDevice(deviceId);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return ret);
  ret = aclrtCreateStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return ret);
  return 0;
}

template <typename T>
int CreateAclTensor(const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr,
                    aclDataType dataType, aclTensor** tensor) {
  auto size = GetShapeSize(shape) * sizeof(T);
  // Call aclrtMalloc to allocate memory on the device.
  auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return ret);
  // Call aclrtMemcpy to copy the data from the host to the device.
  ret = aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return ret);

  // Compute the strides of the contiguous tensor.
  std::vector<int64_t> strides(shape.size(), 1);
  for (int64_t i = shape.size() - 2; i >= 0; i--) {
    strides[i] = shape[i + 1] * strides[i + 1];
  }

  // Call aclCreateTensor to create an aclTensor.
  *tensor = aclCreateTensor(shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND,
                            shape.data(), shape.size(), *deviceAddr);
  return 0;
}

int main() {
  // 1. Boilerplate code for device/stream initialization. For details, see the ACL API manual.
  // Set the device ID in use.
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);

  // 2. Construct the inputs and outputs based on the API definition.
  std::vector<int64_t> selfShape = {4, 2};
  std::vector<int64_t> tensor1Shape = {4, 2};
  std::vector<int64_t> tensor2Shape = {4, 2};
  void* selfDeviceAddr = nullptr;
  void* tensor1DeviceAddr = nullptr;
  void* tensor2DeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* tensor1 = nullptr;
  aclTensor* tensor2 = nullptr;
  aclScalar* value = nullptr;

  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> tensor1HostData = {2, 2, 2, 2, 2, 2, 2, 2};
  std::vector<float> tensor2HostData = {2, 2, 2, 2, 2, 2, 2, 2};
  float scalarValue = 1.2f;

  // Create a self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create a tensor1 aclTensor.
  ret = CreateAclTensor(tensor1HostData, tensor1Shape, &tensor1DeviceAddr, aclDataType::ACL_FLOAT, &tensor1);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create a tensor2 aclTensor.
  ret = CreateAclTensor(tensor2HostData, tensor2Shape, &tensor2DeviceAddr, aclDataType::ACL_FLOAT, &tensor2);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create a value aclScalar.
  value = aclCreateScalar(&scalarValue, aclDataType::ACL_FLOAT);
  CHECK_RET(value != nullptr, return ret);

  // 3. Call the CANN operator library API, which needs to be replaced with the actual one.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnInplaceAddcmul.
  ret = aclnnInplaceAddcmulGetWorkspaceSize(self, tensor1, tensor2, value, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceAddcmulGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate device memory based on the workspaceSize calculated by the first-phase API.
  void* workspaceAddr = nullptr;
  if (workspaceSize > 0) {
    ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret);
  }
  // Call the second-phase API of aclnnInplaceAddcmul.
  ret = aclnnInplaceAddcmul(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceAddcmul failed. ERROR: %d\n", ret); return ret);

  // 4. (Boilerplate code) Wait until the task execution is complete.
  ret = aclrtSynchronizeStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);

  // 5. Obtain the output value and copy the result from the device to the host. Modify the code based on the API definition.
  auto size = GetShapeSize(selfShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), selfDeviceAddr,
                    size * sizeof(resultData[0]), ACL_MEMCPY_DEVICE_TO_HOST);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy result from device to host failed. ERROR: %d\n", ret); return ret);
  for (int64_t i = 0; i < size; i++) {
    LOG_PRINT("result[%ld] is: %f\n", i, resultData[i]);
  }

  // 6. Destroy aclTensor and aclScalar. Modify the code based on the API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(tensor1);
  aclDestroyTensor(tensor2);
  aclDestroyScalar(value);

  // 7. Free device resources. Modify the code based on the API definition.
  aclrtFree(selfDeviceAddr);
  aclrtFree(tensor1DeviceAddr);
  aclrtFree(tensor2DeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();

  return 0;
}