aclnnUnique2

Applicable Products

Function

Performs deduplication on the input tensor self and returns the unique elements in self. This is an enhanced version of the unique function, with a new return value countsOut that indicates the number of occurrences of each element in valueOut within the input self, controlled by the returnCounts parameter.

Function Prototype

Each operator is divided into a two-phase API. The "aclnnUnique2GetWorkspaceSize" API must be called first to obtain the workspace size required for computation and the executor that encapsulates the operator computation flow. Then, the "aclnnUnique2" API is called to perform the computation.

aclnnStatus aclnnUnique2GetWorkspaceSize(
  const aclTensor* self,
  bool             sorted,
  bool             returnInverse,
  bool             returnCounts,
  aclTensor*       valueOut,
  aclTensor*       inverseOut,
  aclTensor*       countsOut,
  uint64_t*        workspaceSize,
  aclOpExecutor**  executor)

aclnnStatus aclnnUnique2(
  void*          workspace,
  uint64_t       workspaceSize,
  aclOpExecutor* executor,
  aclrtStream    stream)

aclnnUnique2GetWorkspaceSize

• Parameters

  Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Shape	Non-contiguous Tensor
  self	Input	Represents the target tensor to be deduplicated.	-	BOOL, FLOAT, FLOAT16, DOUBLE, UINT8, INT8, UINT16, INT16, INT32, UINT32, UINT64, INT64, BFLOAT16	ND	1~8	√
  sorted	Input	Indicates whether to sort valueOut in ascending order.	-	-	-	-	-
  returnInverse	Input	Indicates whether to return the index of each element in the input data within valueOut.	-	-	-	-	-
  returnCounts	Input	Indicates whether to return the count of each unique element in valueOut within the original input tensor.	-	-	-	-	-
  valueOut	Output	The first output tensor, which stores the unique elements from the input tensor. Only 1D shape is supported, and the number of elements is the same as that of self.	-	BOOL, FLOAT, FLOAT16, DOUBLE, UINT8, INT8, UINT16, INT16, INT32, UINT32, UINT64, INT64, BFLOAT16	-	-	-
  inverseOut	Output	The second output tensor. It is meaningful when returnInverse is True or returnCounts is True. Returns the position indices of each element in self within valueOut, with the same shape as self.	-	INT64	ND	-	√
  countsOut	Output	The third output tensor, meaningful only when returnCounts is True. Indicates the occurrence count of each element in valueOut within self, with the same shape as valueOut.	-	INT64	ND	-	√
  workspaceSize	Output	Returns the workspace size to be requested on the Device side.	-	-	-	-	-
  executor	Output	Returns the operator executor, which contains the operator computation flow.	-	-	-	-	-

• Return Value

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

  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	The passed self, valueOut, inverseOut, or countsOut is a null pointer.
  ACLNN_ERR_PARAM_INVALID	161002	The data type of self or valueOut is not within the supported range.
  		self is a non-contiguous tensor.
  		returnInverse is True, and the shape of inverseOut is inconsistent with that of self.
  		returnCounts is True, and the shape of countsOut is inconsistent with that of valueOut.

aclnnUnique2

• Parameters

  Parameter	Input/Output	Description
  workspace	Input	Address of the workspace memory applied on the Device side.
  workspaceSize	Input	Size of the workspace memory applied on the Device side, obtained from the first-phase API aclnnUnique2GetWorkspaceSize.
  executor	Input	Operator executor, which contains the operator computation flow.
  stream	Input	Specifies the stream to execute the task.

• Return Value

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

Constraints

• Deterministic computation:

  • aclnnUnique2 defaults to a deterministic implementation.

  • Ascend 950PR/Ascend 950DT:

    • Due to differences in the deduplication algorithm implementation, when all of the following conditions are met, the operator ignores the value of the sorted input parameter and always sorts the output results in ascending order:

      • self is a 1D input.

      • The data type of self is one of the following: FLOAT, FLOAT16, UINT8, INT8, UINT16, INT16, INT32, UINT32, UINT64, INT64, BFLOAT16.

    • Due to differences in the deduplication algorithm implementation, the inverseOut output of the operator is meaningless when all of the following conditions are met:

      • returnInverse is set to false.

      • self is a 1D input.

      • The data type of self is one of the following: FLOAT, FLOAT16, UINT8, INT8, UINT16, INT16, INT32, UINT32, UINT64, INT64, BFLOAT16.

  • Atlas A2 training products/Atlas A2 inference products, Atlas A3 training products/Atlas A3 inference products: When the input self contains 0, the operator output may contain both positive zero and negative zero, rather than only one zero.

  • Atlas A2 training products/Atlas A2 inference products, Atlas A3 training products/Atlas A3 inference products: When the data volume of self exceeds 200 million, the execution time is long and may cause a timeout.

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_unique2.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: 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 data from host to 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 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 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 device/stream. See the acl API reference.
  // Set deviceId based on your actual device.
  int32_t deviceId = 0;
  aclrtStream stream;
  auto ret = Init(deviceId, &stream);
  // Handle the check based on your needs.
  CHECK_RET(ret == 0, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);
  // 2. Construct inputs and outputs. Customize the construction based on the API definition.
  std::vector<int64_t> selfShape = {8};
  std::vector<int64_t> valueShape = {8};
  std::vector<int64_t> inverseShape = {8};
  std::vector<int64_t> countsShape = {8};
  void* selfDeviceAddr = nullptr;
  void* valueDeviceAddr = nullptr;
  void* inverseDeviceAddr = nullptr;
  void* countsDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* valueOut = nullptr;
  aclTensor* inverseOut = nullptr;
  aclTensor* countsOut = nullptr;
  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> valueHostData = {0, 0, 0, 0, 0, 0, 0, 0};
  std::vector<int64_t> inverseHostData = {0, 0, 0, 0, 0, 0, 0, 0};
  std::vector<int64_t> countsHostData = {0, 0, 0, 0, 0, 0, 0, 0};
  bool sorted = false;
  bool returnInverse = false;
  bool returnCounts = false;

  // Create self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create valueOut aclTensor.
  ret = CreateAclTensor(valueHostData, valueShape, &valueDeviceAddr, aclDataType::ACL_FLOAT, &valueOut);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create inverseOut aclTensor.
  ret = CreateAclTensor(inverseHostData, inverseShape, &inverseDeviceAddr, aclDataType::ACL_INT64, &inverseOut);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create countsOut aclTensor.
  ret = CreateAclTensor(countsHostData, countsShape, &countsDeviceAddr, aclDataType::ACL_INT64, &countsOut);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  // 3. Call the CANN operator library API.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnUnique2.
  ret = aclnnUnique2GetWorkspaceSize(self, sorted, returnInverse, returnCounts, valueOut, inverseOut, countsOut, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnUnique2GetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Allocate device memory based on the 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 aclnnUnique2.
  ret = aclnnUnique2(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnUnique2 failed. ERROR: %d\n", ret); return ret);
  // 4. (Boilerplate) Synchronize and wait for task completion.
  ret = aclrtSynchronizeStream(stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);
  // 5. Obtain the output value. Copy the result from device memory to host memory. Modify based on the specific API definition.
  auto size = GetShapeSize(valueShape);
  std::vector<float> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), valueDeviceAddr, 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("result[%ld] is: %f\n", i, resultData[i]);
  }

  // 6. Release aclTensor and aclScalar. Modify based on the specific API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(valueOut);
  aclDestroyTensor(inverseOut);
  aclDestroyTensor(countsOut);

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