aclnnSplitTensor

📄 View Source Code

Applicable Products

Function

Splits the input self evenly along the dim axis by the splitSections size. If the dim axis cannot be evenly divided, the size of all but the last chunk equals splitSections, and the last chunk is smaller than splitSections.

Function Prototype

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

aclnnStatus aclnnSplitTensorGetWorkspaceSize(
    const aclTensor* self,
    uint64_t         splitSections,
    int64_t          dim,
    aclTensorList*   out,
    uint64_t*        workspaceSize,
    aclOpExecutor**  executor)

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

aclnnSplitTensorGetWorkspaceSize

• Parameters

  Name	Input/Output	Description	Usage	Data Type	Data Format	Dimension (Shape)	Non-contiguous Tensor
  self (aclTensor*)	Input	Input tensor to be split	-	FLOAT, FLOAT16, DOUBLE, INT32, UINT32, INT64, UINT64, INT16, UINT16, INT8, UINT8, BOOL, COMPLEX128, COMPLEX64, BFLOAT16	ND	1-8	√
  splitSections (uint64_t)	Input	Size of the block after even splitting along the dim axis. The last block can be smaller than splitSections.	-	UINT64	-	-	-
  dim (int64_t)	Input	Dimension of the input self to be split. The value range is [–self.dim(), self.dim()).	-	INT64	-	-	-
  out (aclTensorList*)	Output	List of output tensors after splitting.	The dtype of each output must be the same. The number of dimensions of each output tensor in `TensorList` is the same as that of `self`.	FLOAT, FLOAT16, DOUBLE, INT32, UINT32, INT64, UINT64, INT16, UINT16, INT8, UINT8, BOOL, COMPLEX128, COMPLEX64, BFLOAT16	ND	-	√
  workspaceSize (uint64_t*)	Output	Size of the workspace to be allocated on the device.	-	-	-	-	-
  executor (aclOpExecutor**)	Output	Operator executor, containing the operator computation process.	-	-	-	-	-

  • Atlas inference products and Atlas training products: The data type cannot be BFLOAT16.

• Returns

  aclnnStatus: status 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 or out is a null pointer.
  ACLNN_ERR_PARAM_INVALID	161002	The data types of self and out are not within the supported range.
  		The length of self is not within the supported range.
  		The length of a tensor in out is not within the supported range.
  		The value of dim is out of bounds.
  		The shape of the dimension being split is 0 and splitSections is not 0.
  		The shape of the split dimension is not 0 and splitSections is 0.

aclnnSplitTensor

• Parameters

  Parameter	Input/Output	Description
  workspace	Input	Workspace memory address applied on the Device side.
  workspaceSize	Input	Workspace size applied on the Device side, obtained from the first-phase API aclnnSplitTensorGetWorkspaceSize.
  executor	Input	Operator executor that contains the operator computation process.
  stream	Input	Specifies the stream that executes the task.

• Return Value

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

Constraints

• Deterministic computation:

  • aclnnSplitTensor 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 <chrono>
#include <algorithm>
#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_split_tensor.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 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 data from the host to the device.
  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;
}

void CheckResult(const std::vector<std::vector<int64_t>> &shapeList, const std::vector<void *> addrList) {
  for (size_t i = 0; i < shapeList.size(); i++) {
    auto size = GetShapeSize(shapeList[i]);
    std::vector<float> resultData(size, 0);
    auto ret = aclrtMemcpy(resultData.data(), resultData.size() * sizeof(resultData[0]), addrList[i],
                           size * sizeof(resultData[0]), ACL_MEMCPY_DEVICE_TO_HOST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return);
    for (int64_t j = 0; j < size; j++) {
      LOG_PRINT("result[%ld] is: %f\n", j, resultData[j]);
    }
  }
}

int main() {
  // 1. (Boilerplate) Initialize the device/stream. Refer to the ACL API manual.
  // Fill in the deviceId based on your 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. Custom construction is required based on the API.
  std::vector<int64_t> selfShape = {4, 2};
  std::vector<int64_t> shape1 = {2, 2};
  std::vector<int64_t> shape2 = {2, 2};
  uint64_t splitSections = 2;
  int64_t dim = 0;

  void* selfDeviceAddr = nullptr;
  void* shape1DeviceAddr = nullptr;
  void* shape2DeviceAddr = nullptr;

  aclTensor* self = nullptr;
  aclTensor* shape1Addr = nullptr;
  aclTensor* shape2Addr = nullptr;

  std::vector<float> selfHostData = {0, 1, 2, 3, 4, 5, 6, 7};
  std::vector<float> shape1HostData = {0, 1, 4, 5};
  std::vector<float> shape2HostData = {2, 3, 6, 7};

  // Create the self aclTensor.
  ret = CreateAclTensor(selfHostData, selfShape, &selfDeviceAddr, aclDataType::ACL_FLOAT, &self);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  ret = CreateAclTensor(shape1HostData, shape1, &shape1DeviceAddr, aclDataType::ACL_FLOAT, &shape1Addr);
  CHECK_RET(ret == ACL_SUCCESS, return ret);
  ret = CreateAclTensor(shape2HostData, shape2, &shape2DeviceAddr, aclDataType::ACL_FLOAT, &shape2Addr);
  CHECK_RET(ret == ACL_SUCCESS, return ret);

  // Create the out aclTensorList.
  std::vector<aclTensor*> tmp = {shape1Addr, shape2Addr};
  aclTensorList* out = aclCreateTensorList(tmp.data(), tmp.size());
  CHECK_RET(out != nullptr, return ret);

  // Call the CANN operator library API. Replace with the actual API name.
  uint64_t workspaceSize = 0;
  aclOpExecutor *executor;
  // Call the first-phase API of aclnnSplitTensor.
  ret = aclnnSplitTensorGetWorkspaceSize(self, splitSections, dim, out, &workspaceSize, &executor);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSplitTensorGetWorkspaceSize 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) {
    auto 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 aclnnSplitTensor.
  ret = aclnnSplitTensor(workspaceAddr, workspaceSize, executor, stream);
  CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnSplitTensor failed. ERROR: %d\n", ret); return ret);

  ret = aclrtSynchronizeStream(stream);
  CheckResult({shape1, shape2}, {shape1DeviceAddr, shape2DeviceAddr});

  // 6. Release aclTensor and aclScalar. Modify based on the specific API definition.
  aclDestroyTensor(self);
  aclDestroyTensorList(out);
  aclDestroyTensor(shape1Addr);
  aclDestroyTensor(shape2Addr);

  // 7. Release device resources.
  aclrtFree(selfDeviceAddr);
  aclrtFree(shape1DeviceAddr);
  aclrtFree(shape2DeviceAddr);
  if (workspaceSize > 0) {
    aclrtFree(workspaceAddr);
  }
  aclrtDestroyStream(stream);
  aclrtResetDevice(deviceId);
  aclFinalize();

  return 0;
}