aclnnSearchSorted

📄 View Source Code

Applicable Products

Function

• Description: Finds the positions where given tensor values (self) should be inserted in a sorted tensor (sortedSequence). Returns a tensor of the same size as self, where each element indicates the insertion position of the given value in the original tensor. If self is of Scalar type, see aclnnSearchSorteds.

• Formula: Let the length of the innermost sequence to be searched be N. For each input element x = self_i:

  • When right=false, returns the left insertion point:
  $$out_i=\min\{j\in[0,N]\mid sortedSequence_j\ge x\}$$
  • When right=true, returns the right insertion point:
  $$out_i=\min\{j\in[0,N]\mid sortedSequence_j>x\}$$

  If no such $j$ exists, returns $N$.

Function Prototype

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

aclnnStatus aclnnSearchSortedGetWorkspaceSize(
  const aclTensor* sortedSequence,
  const aclTensor* self,
  bool             outInt32,
  bool             right,
  const aclTensor* sorter,
  aclTensor*       out,
  uint64_t*        workspaceSize,
  aclOpExecutor**  executor)

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

aclnnSearchSortedGetWorkspaceSize

• Parameters

  Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
  sortedSequence (aclTensor*)	Input	The sorted input tensor.	The data type must satisfy the deduction relationship with self. In the formula, N is the innermost length of sortedSequence, and sortedSequencej is the j-th element in the innermost dimension.	DOUBLE, FLOAT, FLOAT16, UINT8, INT8, INT16, INT32, INT64	ND	No more than 8 dimensions.	√
  self (aclTensor*)	Input	The input tensor for which to find the insertion positions.	Must satisfy the deduction relationship with the sortedSequence data type. In the formula, x=selfi, where i is the index of the element in self.	DOUBLE, FLOAT, FLOAT16, UINT8, INT8, INT16, INT32, INT64	ND	No more than 8 dimensions.	√
  outInt32 (bool)	Input	Specifies whether to output INT32 results.	Specifies the data type of the output index.	BOOL	-	-	-
  right (bool)	Input	Whether to return the left or right insertion point when a match is found.	false corresponds to sortedSequencej ≥ x in the formula. true corresponds to sortedSequencej > x.	BOOL	-	-	-
  sorter (aclTensor*)	Input	Specifies the element order of sortedSequence.	The shape must be the same as sortedSequence. The element value range is [0, innermost dimension dim-1]. When sorter is passed, sortedSequencej in the formula takes values in the order specified by sorter.	INT64	ND	Same as sortedSequence.	√
  out (aclTensor*)	Output	Outputs the insertion position result.	The shape must be consistent with self. outi in the formula corresponds to the i-th element of the output out.	INT32, INT64	ND	Same as self.	√
  workspaceSize (uint64_t*)	Output	Returns the workspace size to be applied for 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 a null pointer exists in sortedSequence, self, or out.
  ACLNN_ERR_PARAM_INVALID	161002	The data types of sortedSequence and self are not within the supported range.
  		The data type of out contradicts the meaning of the outInt32 value.
  		Data type deduction cannot be performed when sortedSequence and self have different data types.
  		The input sorter is not of the INT64 type.
  		The shape of out differs from the shape of self, or the shape of sorter differs from the shape of sortedSequence.
  		When sortedSequence has more than one dimension and the dimensions of self, except the last dimension, do not match the corresponding dimensions of sortedSequence.

aclnnSearchSorted

• Parameters

  Parameter	Input/Output	Description
  workspace	Input	Specifies the workspace memory address applied on the device.
  workspaceSize	Input	Specifies the workspace size obtained by the first-phase API aclnnSearchSortedGetWorkspaceSize.
  executor	Input	Operator executor that contains the operator computation process.
  stream	Input	Specifies the stream for task execution.

• Return Value

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

Constraints

• Deterministic computation: aclnnSearchSorted 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_searchsorted.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 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 apply for 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);

  // Calculate 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. Refer to 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 result as needed.
  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.
  std::vector<int64_t> selfShape = {4, 2};
  std::vector<int64_t> sortedSequenceShape = {4, 4};
  std::vector<int64_t> sorterShape = {4, 4};
  std::vector<int64_t> outShape = {4, 2};
  void* selfDeviceAddr = nullptr;
  void* sortedSequenceDeviceAddr = nullptr;
  void* sorterDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* self = nullptr;
  aclTensor* sortedSequence = nullptr;
  aclTensor* sorter = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> selfHostData = {1,3,6,8,10,11,14,16};
  std::vector<float> sortedSequenceHostData = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16};
  std::vector<int64_t> sorterHostData = {0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3};
  std::vector<int64_t> outHostData = {0, 0, 0, 0, 0, 0, 0, 0};

  // Create the self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the sortedSequence aclTensor.
  ret = CreateAclTensor(sortedSequenceHostData, sortedSequenceShape, &sortedSequenceDeviceAddr, aclDataType::ACL_FLOAT, &sortedSequence);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the sorter aclTensor.
  ret = CreateAclTensor(sorterHostData, sorterShape, &sorterDeviceAddr, aclDataType::ACL_INT64, &sorter);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  // Create the out aclTensor.
  ret = CreateAclTensor(outHostData, outShape, &outDeviceAddr, aclDataType::ACL_INT64, &out);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  bool outInt32 = false;
  bool right = false;
  // 3. Call the CANN operator library API. Replace with the specific API.
  uint64_t workspaceSize = 0;
  aclOpExecutor* executor;
  // Call the first-phase API of aclnnSearchSorted.
  ret = aclnnSearchSortedGetWorkspaceSize(sortedSequence, self, outInt32, right, sorter, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSearchSortedGetWorkspaceSize 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 aclnnSearchSorted.
  ret = aclnnSearchSorted(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSearchSorted 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-side memory to the host-side memory. Modify based on the specific API definition.
  auto size = GetShapeSize(outShape);
  std::vector<int64_t> resultData(size, 0);
  ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), outDeviceAddr, size * sizeof(int64_t),
                    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: %ld\n", i, resultData[i]);
  }

  // 6. Release aclTensor. Modify based on the specific API definition.
  aclDestroyTensor(self);
  aclDestroyTensor(sortedSequence);
  aclDestroyTensor(sorter);
  aclDestroyTensor(out);

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

  return 0;
}