aclnnStdMeanCorrection

Applicable Products

Product	Supported
Ascend 950PR/Ascend 950DT	√
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: Computes the sample standard deviation and mean.
Formula:

Assume dim is $i$ , then the computation is performed along this dimension. $N$ is the shape of this dimension. Take $self_{i}$ and compute the mean $\bar{x_{i}}$ along this dimension.
$\left\{ \begin{array} {rcl} meanOut& &= \bar{x_{i}}\\ stdOut& &= \sqrt{\frac{1}{max(0, N - \delta N)}\sum_{j=0}^{N-1}(self_{ij}-\bar{x_{i}})^2} \end{array} \right.$
When keepdim = true, the dimension is retained after reduction, and its value in the output shape is 1. When keepdim = false, the dimension is not retained.

Function Prototype

Each operator uses a two-phase API. You must first call the "aclnnStdMeanCorrectionGetWorkspaceSize" API to obtain the input parameters and calculate the required workspace size based on the computation process, and then call the "aclnnStdMeanCorrection" API to perform the computation.

aclnnStatus aclnnStdMeanCorrectionGetWorkspaceSize(
  const aclTensor*   self,
  const aclIntArray* dim,
  int64_t            correction,
  bool               keepdim,
  aclTensor*         stdOut,
  aclTensor*         meanOut,
  uint64_t*          workspaceSize,
  aclOpExecutor**    executor)

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

aclnnStdMeanCorrectionGetWorkspaceSize

Parameters

Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
self (aclTensor*)	Input	self in the formula.	-	FLOAT, BFLOAT16, FLOAT16	ND	-	√
dim (aclIntArray*)	Input.	dim in the formula, indicating the dimensions involved in the computation.	The value range is [-self.dim(), self.dim()-1], and the values in it cannot be the same. When dim is nullptr or [], it is treated as computing all dimensions.	INT64	-	-	-
correction (int64_t)	Input	The δN value in the formula, which serves as the correction value.	-	-	-	-	-
keepdim (bool)	Input.	keepdim in the formula. Specifies whether to retain the input tensor's dimensions in the output tensor.	-	-	-	-	-
stdOut (aclTensor*)	Output	Represents stdOut in the formula.	-	FLOAT, BFLOAT16, FLOAT16	ND	-	√
meanOut (aclTensor*)	Output	meanOut in the formula.	-	FLOAT, BFLOAT16, FLOAT16	ND	-	√
workspaceSize (uint64_t*)	Output	Returns the workspace size to be requested on the Device side.	-	-	-	-	-
executor (aclOpExecutor**)	Output	Returns the op executor, which contains the operator computation flow.	-	-	-	-	-

Return value

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

The first-phase API performs input parameter validation and returns an error in the following scenarios:

Return Value	Error Code	Description
ACLNN_ERR_PARAM_NULLPTR	161001	Indicates that self, stdOut, or meanOut is a null pointer.
ACLNN_ERR_PARAM_INVALID	161002	The data types of self, stdOut, and meanOut are not within the supported range.
		A dimension in the dim array exceeds the dimension range of self.
		Elements in the dim array are duplicated.
		An error occurs when the shape of stdOut meets the following conditions: When keepdim is true, stdOut.shape != self.shape (with the specified dimension dim set to 1). When keepdim is false, stdOut.shape != self.shape (with the specified dimension dim removed).
		An error occurs if the shape of meanOut meets the following conditions: When keepdim is true, meanOut.shape != self.shape (the shape with the specified dimension dim set to 1); When keepdim is false, meanOut.shape != self.shape (the shape with the specified dimension dim removed).

aclnnStdMeanCorrection

Parameters

Parameter	Input/Output	Description
workspace	Input	Specifies the workspace memory address applied for on the Device side.
workspaceSize	Input	Specifies the workspace size applied for on the Device side, obtained from the first-phase API aclnnStdMeanCorrectionGetWorkspaceSize.
executor	Input.	Specifies the operator executor, which contains the operator computation flow.
stream	Input.	Specifies the stream for task execution.

Return value

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

Constraints

Deterministic computation:
- aclnnStdMeanCorrection defaults to a deterministic implementation.

Example

The following provides example code for reference only. For details about compilation and running, see Compile and Run Samples.

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_std_mean_correction.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 shape_size = 1;
  for (auto i : shape) {
    shape_size *= i;
  }
  return shape_size;
}

int Init(int32_t deviceId, aclrtStream* stream) {
  // Boilerplate, 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 device-side memory.
  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 host-side data to device-side memory.
  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 a 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 the aclCreateTensor API 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) Initialize the device/stream. See the acl API manual.
  // Set deviceId based on the actual device.
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  // Handle the check as needed.
  CHECK_RET(ret == 0, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);
  // 2. Construct the input and output. Custom construction is required based on the API interface.
  std::vector<int64_t> selfShape = {2, 3, 4};
  std::vector<int64_t> stdOutShape = {2, 4};
  std::vector<int64_t> meanOutShape = {2, 4};
  void* selfDeviceAddr = nullptr;
  void* stdOutDeviceAddr = nullptr;
  void* meanOutDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* stdOut = nullptr;
  aclTensor* meanOut = nullptr;
  std::vector<float> selfHostData = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18,
                                     19, 20, 21, 22, 23, 24};
  std::vector<float> stdOutHostData = {1, 2, 3, 4, 5, 6, 7, 8.0};
  std::vector<float> meanOutHostData = {1, 2, 3, 4, 5, 6, 7, 8.0};
  std::vector<int64_t> dimData = {1};
  int64_t correction = 1;
  bool keepdim = false;
  // Create the self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the stdOut aclTensor.
  ret = CreateAclTensor(stdOutHostData, stdOutShape, &stdOutDeviceAddr, aclDataType::ACL_FLOAT, &stdOut);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the meanOut aclTensor.
  ret = CreateAclTensor(meanOutHostData, meanOutShape, &meanOutDeviceAddr, aclDataType::ACL_FLOAT, &meanOut);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  const aclIntArray *dim = aclCreateIntArray(dimData.data(), dimData.size());
  CHECK_RET(dim != nullptr, return ACL_ERROR_INTERNAL_ERROR);

  // 3. Call the CANN operator library API. Modify it to the specific API.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnStdMeanCorrection.
  ret = aclnnStdMeanCorrectionGetWorkspaceSize(self, dim, correction, keepdim, stdOut, meanOut, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnStdMeanCorrectionGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Apply for 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 aclnnStdMeanCorrection.
  ret = aclnnStdMeanCorrection(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnStdMeanCorrection failed. ERROR: %d\n", ret); return ret);
  // 4. (Boilerplate) Synchronize and wait for task execution to 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 memory to the host memory. Modify this step based on the specific API definition.
  auto size = GetShapeSize(stdOutShape);
  std::vector<float> stdResultData(size, 0);
  ret = aclrtMemcpy(stdResultData.data(), stdResultData.size() * sizeof(stdResultData[0]), stdOutDeviceAddr, size * sizeof(float),
                    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("stdResultData[%ld] is: %f\n", i, stdResultData[i]);
  }

  std::vector<float> meanResultData(size, 0);
  ret = aclrtMemcpy(meanResultData.data(), meanResultData.size() * sizeof(meanResultData[0]), meanOutDeviceAddr, size * sizeof(float),
                    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("meanResultData[%ld] is: %f\n", i, meanResultData[i]);
  }

  // 6. Release aclTensor and aclScalar. Modify based on the specific API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(stdOut);
  aclDestroyTensor(meanOut);

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