Edge AI GPU Processor Based on OpenCL/TensorFlow APIs
符合OpenCL/TensorFlow API 規範的終端AI 處理器

National Cheng Kung University Prof. Chung-Ho Chen

計畫主持人:國立成功大學 電機工程學系陳中和教授

計畫首要目標在於Virtual Prototyping 符合OpenCL 與TensorFlow API 規範的SIMT GPU運算處理器,CASLab GPU支援OpenCL1.2/2.0 與TensorFlow Framework 的軟體系統堆疊功能。本處理器的性能功耗設定在Edge AI Computing之用,配備GPU MMU 及多種效能提升架構。

The project aims at the development of an SIMT GPU that supports OpenCL and TensorFlow frameworks. This power -efficient open ISA GPU is intended for edge AI training as well as inference applications. The GPU is developed in the full system ESL methodology along with an inference unit for acceleration.