aclnnTriu&aclnnInplaceTriu

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: Sets the elements of the input tensor self (last two dimensions) to 0 in the lower triangular part below the main diagonal. The diagonal parameter can be set to a positive or negative value. The default value is 0. A positive value indicates diagonals above the main diagonal. A negative value indicates diagonals below the main diagonal.
Formula: In the following formula, i indicates the sequence number of the element in the second-to-last dimension (i is the row index), j indicates the sequence number of the element in the last dimension (j is the column index), and d indicates diagonal. In the two-dimensional coordinate diagram corresponding to (i, j), i+d==j indicates that the element is on the diagonal.
$For the diagonal and its upper triangular part, that is, i + d < = j, the original value is retained: out_{i, j} = self_{i, j}\\. For the lower triangular part, that is, i + d > j, the value is set to zero (excluding the diagonal): out_{i, j} = 0.$
Example:

$self = \begin{bmatrix} [9&6&3] \\ [1&2&3] \\ [3&4&1] \end{bmatrix}$ ,

The result of triu(self, diagonal=0) is as follows:

$\begin{bmatrix} [9&6&3] \\ [0&2&3] \\ [0&0&1] \end{bmatrix}$ ;

Adjust the value of diagonal. The result of triu(self, diagonal=1) is as follows:

$\begin{bmatrix} [0&6&3] \\ [0&0&3] \\ [0&0&0] \end{bmatrix}$ ;

Adjust the value of diagonal to -1. The result of triu(self, diagonal=-1) is as follows:

$\begin{bmatrix} [9&6&3] \\ [1&2&3] \\ [0&4&1] \end{bmatrix}$ .

Prototype

aclnnTriu and aclnnInplaceTriu implement the same function in different ways. Select a proper operator based on your requirements.
- aclnnTriu: An output tensor object needs to be created to store the computation result.
- aclnnInplaceTriu: No output tensor object needs to be created, and the computation result is stored in the memory of the input tensor.
Each operator has two-phase API calls. First, aclnnTriuGetWorkspaceSize or aclnnInplaceTriuGetWorkspaceSize is called to obtain input parameters and calculate the required workspace size based on the computation process. Then, aclnnTriu or aclnnInplaceTriu is called to perform computation.

aclnnStatus aclnnTriuGetWorkspaceSize(
  const aclTensor* self, 
  int64_t          diagonal, 
  aclTensor*       out, 
  uint64_t*        workspaceSize, 
  aclOpExecutor**  executor)

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

aclnnStatus aclnnInplaceTriuGetWorkspaceSize(
  aclTensor*      selfRef, 
  int64_t         diagonal, 
  uint64_t*       workspaceSize, 
  aclOpExecutor** executor)

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

aclnnTriuGetWorkspaceSize

Parameters

Name	Input/Output	Description	Precaution	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
self	Input	Target tensor to be converted, corresponding to `self` in the formula.	Empty tensors are supported.	UINT64, INT64, UINT32, INT32, UINT16, INT16, UINT8, INT8, FLOAT16, FLOAT32, DOUBLE, BOOL, BFLOAT16, COMPLEX32, COMPLEX64	ND, NCHW, NHWC, HWCN, NDHWC, NCDHW	2–8 dimensions	-
diagonal	Input	Diagonal offset.	-	int64_t	-	-	-
out	Output	Indicates the result tensor after conversion, that is, `out` in the formula.	The data type and shape must consistent with those of `self`. Empty tensors are supported.	UINT64, INT64, UINT32, INT32, UINT16, INT16, UINT8, INT8, FLOAT16, FLOAT32, DOUBLE, BOOL, BFLOAT16, COMPLEX32, COMPLEX64	ND, NCHW, NHWC, HWCN, NDHWC, NCDHW	2–8 dimensions	-
workspaceSize	Output	Size of the workspace required to be allocated on the device.	-	-	-	-	-
executor	Output	Operator executor, containing the operator computation process.	-	-	-	-	-

Returns

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

The first-phase API performs input parameter validation. The following errors may be returned:

Return	Error Code	Description
ACLNN_ERR_PARAM_NULLPTR	161001	`self` or `out` is a null pointer.
ACLNN_ERR_PARAM_INVALID	161002	The data type of `self` or `out` is not supported.
		The data format of `self` or `out` is private.
		The data types of `self` and `out` are different.
		The shapes of `self` and `out` are different.
		The data formats of `self` and `out` are different.
		The shape of `self` is greater than 8D or less than 2D.

aclnnTriu

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, which is obtained by calling the first-phase API `aclnnTriuGetWorkspaceSize`.
executor	Input	Operator executor, containing the operator computation process.
stream	Input	Stream for executing the task.

Returns

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

aclnnInplaceTriuGetWorkspaceSize

Parameters

Name	Input/Output	Description	Precaution	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
selfRef	Input	Indicates input and output tensors.	-	UINT64, INT64, UINT32, INT32, UINT16, INT16, UINT8, INT8, FLOAT16, FLOAT32, DOUBLE, BOOL, BFLOAT16	ND, NCHW, NHWC, HWCN, NDHWC, NCDHW	2–8 dimensions or empty tensors.	-
diagonal	Input	Diagonal offset.	-	int64_t	-	-	-
workspaceSize	Output	Size of the workspace required to be allocated on the device.	-	-	-	-	-
executor	Output	Operator executor, containing the operator computation process.	-	-	-	-	-

Returns

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

The first-phase API performs input parameter validation. The following errors may be returned:

Return	Error Code	Description
ACLNN_ERR_PARAM_NULLPTR	161001	`selfRef` is a null pointer.
ACLNN_ERR_PARAM_INVALID	161002	The data type of `selfRef` is not supported.
		The data format of `selfRef` is private.
		The shape of `selfRef` is greater than 8D or less than 2D

aclnnInplaceTriu

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, which is obtained by calling the first-phase API `aclnnInplaceTriuGetWorkspaceSize`.
executor	Input	Operator executor, containing the operator computation process.
stream	Input	Stream for executing the task.

Returns

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

Constraints

Deterministic computation:
- aclnnTriu&aclnnInplaceTriu defaults to a deterministic implementation.

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_triu.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) 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 device 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 data to the device 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 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) Initialize the device and stream. For details, see the ACL API manual.
  // Set device ID based on the actual device.
  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, 4};
  std::vector<int64_t> outShape = {4, 4};
  int64_t diagonal = 1;
  void* selfDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> selfHostData = {0, 1.123, -2.001, 303.45, 40009, -50.1234, 60.666, -7.6543,
                                     8000, -9.009, 1024, -11.23345, 12, 1356, -14.99, -15.34023};
  std::vector<float> outHostData = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};

  // Create a self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, 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. Change the API name to the actual one.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnTriu.
  ret = aclnnTriuGetWorkspaceSize(self, diagonal, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnTriuGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate device memory based on workspaceSize computed 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 aclnnTriu.
  ret = aclnnTriu(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnTriu failed. ERROR: %d\n", ret); return ret);

  uint64_t inplaceWorkspaceSize = 0;
  aclOpExecutor* inplaceExecutor;
  // Call the first-phase API of aclnnInplaceTriu.
  ret = aclnnInplaceTriuGetWorkspaceSize(self, diagonal, &inplaceWorkspaceSize, &inplaceExecutor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceTriuGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate device memory based on workspaceSize computed by the first-phase API.
  void* inplaceWorkspaceAddr = nullptr;
  if (inplaceWorkspaceSize > 0) {
    ret = aclrtMalloc(&inplaceWorkspaceAddr, inplaceWorkspaceSize, 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 aclnnInplaceTriu.
  ret = aclnnInplaceTriu(inplaceWorkspaceAddr, inplaceWorkspaceSize, inplaceExecutor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnInplaceTriu failed. ERROR: %d\n", ret); return ret);

  // 4. (Boilerplate) 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 memory 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]);
  }

  auto inplaceSize = GetShapeSize(selfShape);
  std::vector<float> inplaceResultData(inplaceSize, 0);
  ret = aclrtMemcpy(inplaceResultData.data(), inplaceResultData.size() * sizeof(inplaceResultData[0]), selfDeviceAddr,
                    inplaceSize * 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 < inplaceSize; i++) {
    LOG_PRINT("inplaceResult[%ld] is: %f\n", i, inplaceResultData[i]);
  }

  // 6. Release aclTensors and aclScalars. Modify the code based on the API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(out);

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