aclnnAddRelu&aclnnInplaceAddRelu

Supported Products

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

Function

Description: Performs the addition operation and activates the result.
Formula: $out_i = self_i+alpha \times other_i$ $relu(self) = \begin{cases} self, & self\gt 0 \\ 0, & self\le 0 \end{cases}$

Prototype

aclnnAddRelu and aclnnInplaceAddRelu implement the same function. The differences are as follows. Select a proper operator based on the actual scenario.
- aclnnAddRelu: You need to create an output tensor object to store the computation result.
- aclnnInplaceAddRelu: You do not need to create an output tensor object. The computation result is stored in the memory of the input tensor.
Each operator has two-phase API calls. First, aclnnAddReluGetWorkspaceSize or aclnnInplaceAddReluGetWorkspaceSize is called to obtain the workspace size required for computation and the executor that contains the operator computation process. Then, aclnnAddRelu or aclnnInplaceAddRelu is called to perform computation.
- aclnnStatus aclnnAddReluGetWorkspaceSize(const aclTensor* self, const aclTensor* other, aclScalar* alpha, aclTensor* out, uint64_t* workspaceSize, aclOpExecutor** executor)
- aclnnStatus aclnnAddRelu(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, aclrtStream stream)
- aclnnStatus aclnnInplaceAddReluGetWorkspaceSize(aclTensor* selfRef, const aclTensor* other, aclScalar* alpha, uint64_t* workspaceSize, aclOpExecutor** executor)
- aclnnStatus aclnnInplaceAddRelu(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor, aclrtStream stream)

aclnnAddReluGetWorkspaceSize

Parameters:
- self (aclTensor*, compute input): input self in the formula, which is an aclTensor on the device and indicates the target tensor to be converted. The data type must meet the type deduction rules with other. The shape must meet the broadcast relationship with other. Non-contiguous tensors are supported. The data format can be ND.
  - Atlas training series products: The data type can be FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training series products/Atlas A3 inference series products: The data type can be BFLOAT16, FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
- other (aclTensor*, compute input): input other in the formula. The data type must meet the type deduction rules with self. The shape must meet the broadcast relationship with self. Non-contiguous tensors are supported. The data format can be ND.
  - Atlas training series products: The data type can be FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training series products/Atlas A3 inference series products: The data type can be BFLOAT16, FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
- alpha (aclScalar*, compute input): alpha in the formula. The data type can be converted to the data type deduced from self and other.
  - Atlas training series products: The data type can be FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training series products/Atlas A3 inference series products: The data type can be BFLOAT16, FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
- out (aclTensor*, compute output): out in the formula. The data type must be convertible to the data type deduced from self and other, and the shape must be the shape after self and other are broadcast. Non-contiguous tensors are supported. The data format can be ND.
  - Atlas training series products: The data type can be FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training series products/Atlas A3 inference series products: The data type can be BFLOAT16, FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
- workspaceSize (uint64_t*, output): size of the workspace to be allocated on the device.
- executor (aclOpExecutor**, output): operator executor, containing the operator computation process.
Returns:

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

   The first-phase API implements input parameter verification. The following errors may be thrown.
 161001 (ACLNN_ERR_PARAM_NULLPTR): 1. The passed self, other, alpha, or out is a null pointer.
 161002 (ACLNN_ERR_PARAM_INVALID): 1. The data type of self or other is not supported.
                                      2. Data type deduction cannot be performed for self and other.
                                      3. The deduced data type cannot be converted to the data type of out.
                                      4. Broadcasting cannot be performed for the shapes of self and other.
                                      5. alpha cannot be converted to the data type deduced from self and other.

aclnnAddRelu

Parameters:
- workspace (void*, input): address of the workspace to be allocated on the device.
- workspaceSize (uint64_t, input): workspace size allocated on the device, which is obtained by the first-phase API aclnnAddReluGetWorkspaceSize.
- executor (aclOpExecutor*, input): operator executor, containing the operator computation process.
- stream (aclrtStream, input): stream for executing the task.
Returns:

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

aclnnInplaceAddReluGetWorkspaceSize

Parameters:
- selfRef (aclTensor*, compute input|compute output): input and output tensors, that is, self and out in the formula. They are aclTensors on the device and indicate the target tensors to be converted. Its data type and the data type of other must meet the type deduction rules (see Deduction Relationship) and must be the data type that can be converted after deduction. Non-contiguous tensors are supported. The data format can be ND.
  - Atlas training series products: The data type can be FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training series products/Atlas A3 inference series products: The data type can be BFLOAT16, FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
- other (aclTensor*, compute input): input other in the formula. The data type must meet the type deduction rules with selfRef. The shape must meet the broadcast relationship with selfRef. Non-contiguous tensors are supported. The data format can be ND.
  - Atlas training series products: The data type can be FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training series products/Atlas A3 inference series products: The data type can be BFLOAT16, FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
- alpha (aclScalar*, compute input): alpha in the formula. The data type can be converted to the data type deduced from selfRef and other.
  - Atlas training series products: The data type can be FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
  - Atlas A2 training products/Atlas A2 inference products and Atlas A3 training series products/Atlas A3 inference series products: The data type can be BFLOAT16, FLOAT16, FLOAT32, INT8, UINT8, INT16, INT32, or INT64.
- workspaceSize (uint64_t*, output): size of the workspace to be allocated on the device.
- executor (aclOpExecutor**, output): operator executor, containing the operator computation process.
Returns:

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

   The first-phase API implements input parameter verification. The following errors may be thrown.
 161001 (ACLNN_ERR_PARAM_NULLPTR): 1. The passed selfRef, other, or alpha is a null pointer.
 161002 (ACLNN_ERR_PARAM_INVALID): 1. The data types of selfRef and other are not supported.
                                   2. Data type deduction cannot be performed for selfRef and other.
                                   3. The deduced data type cannot be converted to the type of selfRef.
                                   4. Broadcasting cannot be performed for the shapes of selfRef and other.
                                   5. The shape after broadcasting is not equal to the shape of selfRef.
                                   6. alpha cannot be converted to the data type deduced from selfRef and other.

aclnnInplaceAddRelu

Parameters:
- workspace (void*, input): address of the workspace to be allocated on the device.
- workspaceSize (uint64_t, input): workspace size allocated on the device, which is obtained by the first-phase API aclnnInplaceAddReluGetWorkspaceSize.
- executor (aclOpExecutor*, input): operator executor, containing the operator computation process.
- stream (aclrtStream, input): stream for executing the task.
Returns:

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

Constraints

Deterministic compute:
- aclnnAddRelu&aclnnInplaceAddRelu defaults to a deterministic implementation.
For the scenario where the data type of selfRef is INT8 and that of other is INT32: The cast operator has a precision issue when converting the INT32 type to the INT8 type (see aclnnCast. In this scenario, the output result precision cannot be ensured.

Example

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

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_add_relu.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) {
  // (Fixed writing) Initialize resources.
  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. (Fixed writing) Initialize the device and stream. 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 input and output based on the API.
  std::vector<int64_t> selfShape = {4, 2};
  std::vector<int64_t> otherShape = {4, 2};
  std::vector<int64_t> outShape = {4, 2};
  void* selfDeviceAddr = nullptr;
  void* otherDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* other = nullptr;
  aclScalar* alpha = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> otherHostData = {1, 1, 1, 2, 2, 2, 3, 3};
  std::vector<float> outHostData(8, 0);
  float alphaValue = 1.2f;
  // Create a self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create an other aclTensor.
  ret = CreateAclTensor(otherHostData, otherShape, &otherDeviceAddr, aclDataType::ACL_FLOAT, &other);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create an alpha aclScalar.
  alpha = aclCreateScalar(&alphaValue, aclDataType::ACL_FLOAT);
  CHECK_RET(alpha != nullptr, return ret);
  // Create an out aclTensor.
  ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_FLOAT, &out);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;

  // aclnnAddRelu API call example
  // 3. Call the CANN operator library API.
  // Call the first-phase API of aclnnAddRelu.
  ret = aclnnAddReluGetWorkspaceSize(self, other, alpha, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAddReluGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate device memory based on 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 aclnnAddRelu.
  ret = aclnnAddRelu(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnAddRelu failed. ERROR: %d\n", ret); return ret);

  // 4. (Fixed writing) 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 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]);
  }


  // aclnnInplaceAddRelu API call example
  // 3. Call the CANN operator library API.
  LOG_PRINT("\ntest aclnnInplaceAddRelu\n");
  // Call the first-phase API of aclnnInplaceAddRelu.
  ret = aclnnInplaceAddReluGetWorkspaceSize(self, other, alpha, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceAddReluGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate device memory based on workspaceSize calculated by the first-phase API.
  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 aclnnInplaceAddRelu.
  ret = aclnnInplaceAddRelu(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceAddRelu failed. ERROR: %d\n", ret); return ret);

  // 4. (Fixed writing) 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.
  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. Release aclTensor and aclScalar. Modify the configuration based on the API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(other);
  aclDestroyScalar(alpha);
  aclDestroyTensor(out);

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

  return 0;
}