An OpenCL(TM) Deep Learning Accelerator on Arria 10

no code implementations • 13 Jan 2017 • Utku Aydonat, Shane O'Connell, Davor Capalija, Andrew C. Ling, Gordon R. Chiu

We show a novel architecture written in OpenCL(TM), which we refer to as a Deep Learning Accelerator (DLA), that maximizes data reuse and minimizes external memory bandwidth.

Image Classification