Resistive-RAM-based (ReRAM-based) computing shows great potential on accelerating DNN inference by its highly parallel structure. Regrettably, computing accuracy in practical is much lower than expected due to the non-ideal ReRAM device. Conventional computing flow with fixed wordline activation scheme can effectively protect computing accuracy, but at the cost of significant performance and energy savings reduction. For such embarrassment of accuracy, performance and energy, this paper proposes a new Adaptive-Wordline-Activation control scheme (AWA-control) and combines it with a theoretical Output-Significance-Aligned computing flow (OSA-flow) to enable fine-grained control on output significance with distinct impact on final result. We demonstrate AWA-control-supported OSA-flow architecture with maximal compatibility to conventional crossbar by input retiming and weight remapping using shifting registers to enable the new flow. On the other hand, in contrast to the conventional computing architecture, the OSA-flow architecture shows the better capability to exploit data sparsity commonly seen in DNN models. So we also design a sparsity-aware OSA-flow architecture for further DNN speedup. Evaluation results show that OSA-flow architecture can provide significant performance improvement of 21.6X, and energy savings of 96.2% over conventional computing architecture with similar DNN accuracy.

基于电阻式 RAM（基于 ReRAM）的计算通过其高度并行的结构在加速 DNN 推理方面显示出巨大的潜力。遗憾的是，由于 ReRAM 设备不理想，实际计算精度远低于预期。具有固定字线激活方案的常规计算流程可以有效地保护计算精度，但代价是性能和节能显着降低。针对这种精度、性能和能量的尴尬，本文提出了一种新的 Adaptive-Wordline-Activation 控制方案（AWA-control），并将其与理论上的 Output-Significance-Aligned 计算流程（OSA-flow）相结合，以实现细粒度控制输出显着性，对最终结果有明显影响。我们演示AWA 控制- 支持OSA 流架构，通过输入重定时和权重重映射使用移位寄存器实现与传统交叉开关的最大兼容性，以启用新流。另一方面，与传统的计算架构相比，OSA-flow架构显示出更好的利用 DNN 模型中常见的数据稀疏性的能力。因此，我们还设计了一个稀疏感知OSA 流架构，以进一步加速 DNN。评估结果表明，与具有相似 DNN 精度的传统计算架构相比， OSA-flow架构可以提供 21.6 倍的显着性能提升和 96.2% 的能耗。