This section describes the basic data structures required for calling CANN operator APIs. You may directly use these structures without delving into their internal implementations.
Note that the basic data structures can be created as required by using the public APIs in the , for example, aclCreateTensor.
aclTensor
A framework-defined structure used to manage and store tensor data (such as vectors and matrices). You can create this object by using the aclCreateTensor API.
[object Object]aclScalar
A framework-defined structure used to manage and store scalar data (a single value). You can create this object by using the aclCreateScalar API.
[object Object]aclIntArray
A framework-defined structure used to manage and store an array of integer data. You can create this object by using the aclCreateIntArray API.
[object Object]aclFloatArray
A framework-defined structure used to manage and store an array of float32 data. You can create this object by using the aclCreateFloatArray API.
[object Object]aclBoolArray
A framework-defined structure used to manage and store an array of Boolean data. You can create this object by using the aclCreateBoolArray API.
[object Object]aclTensorList
A framework-defined structure used to manage and store an array of multiple tensors. You can create this object by using the aclCreateTensorList API.
[object Object]aclScalarList
A framework-defined structure used to manage and store an array of scalar data. You can create this object by using the aclCreateScalarList API.
[object Object]aclOpExecutor
A framework-defined executor data structure, which is a container for operator execution.
Generally, when the first-phase API aclxxXxxGetWorkspaceSize is called, the framework automatically creates an aclOpExecutor object. When the second-phase API aclxxXxx is called, the object is automatically released.
[object Object]aclrtStream
A framework-defined stream processing data structure, which is used to manage and maintain the execution sequence of asynchronous operations.
[object Object]