Abstract: To enable energy-efficient embedded execution of Deep Neural Networks (DNNs), the critical sections of these workloads, their multiply-accumulate (MAC) operations, need to be carefully ...