A 12.1 TOPS/W Quantized Network Acceleration Processor With Effective-Weight-Based Convolution and Error-Compensation-Based Prediction.

In this article, a quantized network acceleration processor (QNAP) is proposed to efficiently accelerate CNN processing by eliminating most unessential operations based on algorithm-hardware co-optimizations. First, an effective-weight-based convolution (EWC) is proposed to distinguish a group of effective weights (EWs) to replace the other unique weights. Therefore, the input activations corresponding to the same EW can be accumulated first and then multiplied by the EW to reduce amounts of multiplication operations, which is efficiently supported by the dedicated process elements in QNAP. The experimental results show that energy efficiency is improved by 1.59 $\times $ –3.20 $\times $ compared with different UCNN implementations. Second, an error-compensation-based prediction (ECP) method adopts trained compensated values to replace partly unimportant partial sums to further reduce potentially redundant addition operations caused by the ReLU function. Compared with SnaPEA and Pred on AlexNet, 1.23 $\times $ and 1.75 $\times $ higher energy efficiencies (TOPS/W) are achieved by ECP, respectively, with marginal accuracy loss. Third, the residual pipeline mode is proposed to efficiently implement residual blocks with a 1.5 $\times $ lower memory footprint, a 1.18 $\times $ lower power consumption, and a 13.15% higher hardware utilization on average than existing works. Finally, the QNAP processor is fabricated in the TSMC 28-nm CMOS process with a core area of 1.9 mm2. Benchmarked with AlexNet, VGGNet, GoogLeNet, and ResNet on ImageNet at 470 MHz and 0.9 V, the processor achieves 117.4 frames per second with 131.6-mW power consumption on average, which outperforms the state-of-the-art processors by 1.77 $\times $ –24.20 $\times $ in energy efficiency. [ABSTRACT FROM AUTHOR]