arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2409.14017cs.AR
收藏

SPEED:一种可实现高效多精度DNN推理的可扩展RISC-V向量处理器

SPEED: A Scalable RISC-V Vector Processor Enabling Efficient Multi-Precision DNN Inference

Chuanning Wang, Chao Fang, Xiao Wu, Zhongfeng Wang, Jun Lin

AI总结:

针对边缘平台部署多精度DNN的难题,提出可扩展RVV处理器SPEED,通过定制指令、参数化张量单元与混合数据流提升效率,28nm工艺下在吞吐量、能效与面积效率上均优于现有同类处理器。

摘要:

深度神经网络(DNN)巨大的计算与存储需求,阻碍了其在资源受限的边缘平台上的部署。量化多精度DNN(记为MP-DNN)为解决这些局限提供了可行方案,但由于指令复杂、并行处理效果不佳、数据流映射效率低,现有RISC-V处理器难以适配这类网络。针对上述问题,本文提出了可扩展RISC-V向量(RVV)处理器SPEED,可实现高效的MP-DNN推理,在定制指令、硬件架构和数据流映射三方面均有创新。首先,基于RVV扩展引入了专用定制RISC-V指令,降低了指令复杂度,使SPEED能够支持4位到16位的处理精度,且硬件开销极小。其次,设计了参数化多精度张量单元,集成在可扩展模块中,通过提供可重构并行度匹配不同MP-DNN的计算模式,提升了并行处理能力。最后,采用灵活的混合数据流方法,根据不同DNN算子的计算模式提升计算效率与能效。SPEED基于台积电28nm工艺完成综合,实验结果显示,其处理4位算子时峰值吞吐量达737.9 GOPS,能效达1383.4 GOPS/W。此外,SPEED的面积效率优于现有RVV处理器,8位算子性能与最佳整数性能分别提升5.9~26.9倍与8.2~18.5倍,凸显了其在高效MP-DNN推理方面的巨大潜力。

英文摘要:

Deploying deep neural networks (DNNs) on those resource-constrained edge platforms is hindered by their substantial computation and storage demands. Quantized multi-precision DNNs, denoted as MP-DNNs, offer a promising solution for these limitations but pose challenges for existing RISC-V processors due to complex instructions, suboptimal parallel processing, and inefficient dataflow mapping. To tackle the challenges mentioned above, SPEED, a scalable RISC-V vector (RVV) processor, is proposed to enable efficient MP-DNN inference, incorporating innovations in customized instructions, hardware architecture, and dataflow mapping. Firstly, some dedicated customized RISC-V instructions are introduced based on RVV extensions to reduce the instruction complexity, allowing SPEED to support processing precision ranging from 4-bit to 16-bit with minimized hardware overhead. Secondly, a parameterized multi-precision tensor unit is developed and integrated within the scalable module to enhance parallel processing capability by providing reconfigurable parallelism that matches the computation patterns of diverse MP-DNNs. Finally, a flexible mixed dataflow method is adopted to improve computational and energy efficiency according to the computing patterns of different DNN operators. The synthesis of SPEED is conducted on TSMC 28nm technology. Experimental results show that SPEED achieves a peak throughput of 737.9 GOPS and an energy efficiency of 1383.4 GOPS/W for 4-bit operators. Furthermore, SPEED exhibits superior area efficiency compared to prior RVV processors, with enhancements of 5.9$\sim$26.9$\times$ and 8.2$\sim$18.5$\times$ for 8-bit operator and best integer performance, respectively, which highlights SPEED's significant potential for efficient MP-DNN inference.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑