aclCreateTensorList

Function Usage

Creates an aclTensorList object as the input parameter for single-operator API execution.

aclTensorList is a framework-defined list structure used to manage and store multiple tensors. You can directly use it without paying attention to its internal implementation.

Prototype

aclTensorList *aclCreateTensorList(const aclTensor *const *value, uint64_t size)

Parameters

Returns

Created aclTensorList on success; else, nullptr.

Constraints

• Before calling this API, call aclCreateTensor to create an aclTensor.
• This API must be used together with aclDestroyTensorList. They are used to create and destroy the aclTensorList, respectively.
• The size of the aclTensorList can be obtained by calling aclGetTensorListSize.
• The following APIs can be called to update the device memory addresses recorded in the input/output aclTensorList:
  • aclSetDynamicInputTensorAddr
  • aclSetDynamicOutputTensorAddr
  • aclSetDynamicTensorAddr

Examples

The following code examples are for reference only and are not intended for direct copying and execution:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15

//Create aclTensors input1 and input2.
std::vector<int64_t> shape = {1, 2, 3};
aclTensor *input1 = aclCreateTensor(shape.data(), shape.size(), aclDataType::ACL_FLOAT,
nullptr, 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(), nullptr);
aclTensor *input2 = aclCreateTensor(shape.data(), shape.size(), aclDataType::ACL_FLOAT,
nullptr, 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(), nullptr);
// Create an aclTensorList.
std::vector<aclTensor *> tmp{input1, input2};
aclTensorList* tensorList = aclCreateTensorList(tmp.data(), tmp.size());
// Use the aclTensorList as the input parameter for single-operator API execution.
auto ret = aclxxXxxGetWorkspaceSize(srcTensor, tensorList, ..., outTensor, ..., &workspaceSize, &executor);
ret = aclxxXxx(...);
...
// Destroy the aclTensorList.
ret = aclDestroyTensorList(tensorList);