aclnnSearchSorteds

📄 View Source Code

Applicable Products

Function

• Operator function: Finds the position where a given scalar value (self) should be inserted in a sorted one-dimensional tensor (sortedSequence). Returns a tensor of shape [1], indicating the position where the given scalar value should be inserted in the original tensor. If self is of the Tensor type, refer to aclnnSearchSorted.

• Formula: Let the length of the sequence to be searched be $N$, and the scalar input be $x=self$:

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

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

Function Prototype

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

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

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

aclnnSearchSortedsGetWorkspaceSize

• Parameters

  Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Shape	Non-contiguous Tensor
  sortedSequence (aclTensor*)	Input	A sorted one-dimensional input tensor.	Must be a one-dimensional tensor. The data type must satisfy the deduction relationship with self. In the formula, N is the length of sortedSequence, and sortedSequencej is the j-th element.	DOUBLE, FLOAT, FLOAT16, UINT8, INT8, INT16, INT32, INT64	ND	1D	√
  self (aclScalar*)	Input.	The scalar value whose insertion position is to be found.	Host-side scalar. Its data type must satisfy the deduction relationship with that of sortedSequence. In the formula, x=self.	DOUBLE, FLOAT, FLOAT16, UINT8, INT8, INT16, INT32, INT64	-	Scalar.	-
  outInt32 (bool)	Input	Whether to output INT32 results.	Specifies the data type of the output index.	BOOL	-	-	-
  right (bool)	Input.	Returns the left/right insertion point when an equal value is matched.	false corresponds to sortedSequencej ≥ x in the formula. true corresponds to sortedSequencej > x.	BOOL	-	-	-
  sorter (aclTensor*)	Input.	Specifies the order of elements in sortedSequence.	The data type must be INT64. The shape must be consistent with sortedSequence. When sorter is passed in, 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 output is a position index. The out in the formula corresponds to the output parameter out.	INT32, INT64	ND	Usually [1].	-
  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 process.	-	-	-	-	-

• 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	A null pointer is passed 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 sorter differs from that of sortedSequence.
  		sortedSequence is not a one-dimensional tensor.

aclnnSearchSorteds

• Parameters

  Parameter	Input/Output	Description
  workspace	Input	Address of the workspace memory applied on the device side.
  workspaceSize	Input	Workspace size obtained by the first-phase API aclnnSearchSortedsGetWorkspaceSize.
  executor	Input	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 description: aclnnSearchSorteds 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 memory on the device side.
  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 the host side 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);

  // Calculate 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 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.
  // Fill in the 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 the input and output. Customize the construction based on the API interface.
  std::vector<int64_t> sortedSequenceShape = {4};
  std::vector<int64_t> sorterShape = {4};
  std::vector<int64_t> outShape = {1};
  void* sortedSequenceDeviceAddr = nullptr;
  void* sorterDeviceAddr = nullptr;
  void* outDeviceAddr = nullptr;
  aclTensor* sortedSequence = nullptr;
  aclTensor* sorter = nullptr;
  aclTensor* out = nullptr;
  std::vector<float> sortedSequenceHostData = {1,3,6,8};
  std::vector<int64_t> sorterHostData = {0,1,2,3};
  std::vector<int64_t> outHostData = {0};

  // 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);
  float value = 5;
  auto self = aclCreateScalar(&value, aclDataType::ACL_FLOAT);
  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 aclnnSearchSorteds.
  ret = aclnnSearchSortedsGetWorkspaceSize(sortedSequence, self, outInt32, right, sorter, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSearchSortedsGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
  // Apply for 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 aclnnSearchSorteds.
  ret = aclnnSearchSorteds(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSearchSorteds 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 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 the aclTensor. Modify based on the specific API definition.
  aclDestroyScalar(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;
}