AscendCL-based pyACL

AscendCL (Ascend Computing Language) provides a collection of C language APIs for the development of deep neural network (DNN) apps. These APIs are designed for facilitating deep learning inference computing, graphics and image preprocessing, and single-operator accelerated computing on the Ascend CANN platform by leveraging Ascend hardware compute resources. In a nutshell, AscendCL is a unified API framework used to invoke every resource. The computing resource layer is the hardware computing power basis of Ascend AI Processor. It performs matrix-related computing of neural networks, general computing and execution control functions such as control operators, scalars, and vectors, and image and video data preprocessing, ensuring execution of deep neural network computing.

Python Ascend Computing Language (pyACL) is a Python API library encapsulated using CPython based on AscendCL. It provides APIs for runtime management, model inference, and media data processing, and supports single-operator execution, enabling you to use Python to manage the running and resources of Ascend AI Processors.

Figure 1 Logical architecture