基于新型压缩树的近似Booth乘法器

随着大数据、云计算、物联网等技术的兴起,终端设备在硬件开销和供电方面面临巨大挑战,对于新型高效低功耗运算单元的需求日益迫切。针对运算单元功耗高的问题,提出了一种新型高效低功耗的近似Booth乘法器,可应用于图像处理、多媒体处理、模式识别等可容错应用领域。实验结果表明,与已有乘法器相比,所提出的近似Booth乘法器在功耗和延时方面分别降低了19.3%和28.6%,在面积方面节省了29.0%。同时,所提出的近似Booth乘法器的运算精度也具备一定的优势。最后,在高斯滤波的应用中验证了所提出的近似Booth乘法器的实用性。

An Approximate Booth Multiplier Based on Novel Wallace Tree

With the rise of big data, cloud computing and the internet of things, terminal devices face great challenges in hardware and power. The demand of new high efficiency low power operation unit is becoming more and more urgent. A novel Booth multiplier with high efficiency and low power consumption was proposed to solve the problem of high power consumption of operational units. It could be used in image processing, multimedia processing, pattern recognition and other fault-tolerant applications. Experimental results showed that compared with the existing Booth multipliers, the proposed approximate Booth multiplier reduced the power consumption and delay by 19.3% and 28.6% respectively, and saved area by 29.0%. Meanwhile the approximate Booth multiplier also had some advantages in operation accuracy. Finally, the validity of the proposed approximate Booth multiplier was verified in the application of Gaussian filtering.