List of Common APIs
This section describes the common meta APIs required for calling the APIs of the NN, fusion, and DVPP operators for creating objects such as aclTensor, aclScalar, and aclIntArray. For details about the dependent header files and library files, see Table 1.
API |
Description |
|---|---|
Creates an aclBoolArray. |
|
Creates an aclFloatArray. |
|
Creates an aclIntArray. |
|
Creates an aclScalar. |
|
Creates an aclScalarList. |
|
Creates an aclTensor. |
|
Creates an aclTensorList. |
|
Destroys a reusable aclOpExecutor. |
|
Destroys the created aclBoolArray. |
|
Destroys the created aclFloatArray. |
|
Destroys the created aclIntArray. |
|
Destroys the created aclScalar. |
|
Destroys the created aclScalarList. The scalars in the aclScalarList do not need to be destroyed again. |
|
Destroys the created aclTensor. |
|
Destroys the created aclTensor list. The tensors in the aclTensor list do not need to be destroyed again. |
|
Dumps the input and output tensor data of the operator. |
|
Obtains the size of an aclBoolArray. |
|
Obtains the data type of an aclTensor. |
|
Obtains the size of an aclFloatArray. |
|
Obtains the format of an aclTensor. |
|
Obtains the size of an aclIntArray. |
|
Obtains the device memory address originally recorded in an aclTensor. |
|
Obtains the size of an aclScalarList. |
|
Obtains the StorageShape of an aclTensor. |
|
Obtains the size of the aclTensorList. |
|
Obtains the ViewOffset of an aclTensor, that is, the offset corresponding to ViewShape. |
|
Obtains the ViewShape of an aclTensor. |
|
Obtains the ViewStrides of an aclTensor, that is, the stride corresponding to ViewShape. |
|
Initializes the parameters of a given tensor. |
|
Enables aclOpExecutor to be reusable. |
|
After aclOpExecutor reuse is enabled, if the input device memory address changes, the device memory address recorded in the input aclTensorList needs to be updated. |
|
After aclOpExecutor reuse is enabled, if the output device memory address changes, the device memory address recorded in the output aclTensorList needs to be updated. |
|
After aclOpExecutor reuse is enabled, if the input or output device memory address changes, the device memory address recorded in the corresponding aclTensorList needs to be updated. |
|
After aclOpExecutor reuse is enabled, if the input device memory address changes, the device memory address recorded in the input aclTensor needs to be updated. |
|
After aclOpExecutor reuse is enabled, if the output device memory address changes, the device memory address recorded in the output aclTensor needs to be updated. |
|
Updates the device memory address originally recorded in the aclTensor. |
|
After aclOpExecutor reuse is enabled, if the input or output device memory address changes, the device memory address recorded in the corresponding aclTensor needs to be updated. |
|
AclSetInputTensorAddr |
Reserved. No setup required. |
AclSetOutputTensorAddr |
Reserved. No setup required. |
AclSetDynamicInputTensorAddr |
Reserved. No setup required. |
AclSetDynamicOutputTensorAddr |
Reserved. No setup required. |
AclSetTensorAddr |
Reserved. No setup required. |
AclSetDynamicTensorAddr |
Reserved. No setup required. |