Abstract:Multiplications have been commonly conducted in quantized CNNs, filters, and reconfigurable cores, and so on, which are widely deployed in mobile and embedded applications. Most multipliers are designed to perform multiplications with symmetric bit-widths, i.e., n - by n -bit multiplication. Such features would cause extra area overhead and performance loss when m - by n -bit multiplications ( m > n ) are deployed in the same hardware design, resulting in inefficient multiplication operations. It is highly desired and challenging to propose a reconfigurable multiplier design to accommodate operands with both symmetric and asymmetric bit-widths. In this work, we propose a reconfigurable approximate multiplier to support multiplications at various precisions, i.e., bit-widths. Unlike prior works of approximate adders assuming a uniform weight distribution with bit-wise independence, scenarios like a quantized CNN may have a centralized weight distribution and hence follow a Gaussian-like distribution with correlated adjacent bits. Thus, a new block-based approximate adder is also proposed as part of the multiplier to ensure energy-efficient operation with an awareness of the bit-wise correlation. Our experimental results show that the proposed approximate adder significantly reduces the error rate by 76% to 98% over a state-of-the-art approximate adder for Gaussian-like distribution scenarios. Evaluation results show that the proposed multiplier is 19% faster and 22% more power saving than a Xilinx multiplier IP at the same bit precision and achieves a 23.94-dB peak signal-to-noise ratio, which is comparable to the accurate one of 24.10 dB when deployed in a Gaussian filter for image processing tasks.

Compressors Evolution Based High Speed and Energy Efficient Approximate Signed Multiplier

High Quality Multiplier Design Based on Improved Circuits

A Reconfigurable Multiplier for Signed Multiplications with Asymmetric Bit-Widths.

An Optimized Compression Strategy for Compressor-Based Approximate Multiplier

A Reconfigurable Approximate Multiplier for Quantized CNN Applications.

Efficient Approximate Multiplier Based on a New 1-Gate Approximate Compressor

Optimally Approximated and Unbiased Floating-Point Multiplier with Runtime Configurability

An Efficient Approximate Multiplier with Encoded Partial Products and Inexact Counter for Joint Photographic Experts Group Compression

High-speed Area-Efficient and Power-Aware Multiplier Design Using Approximate Compressors along with Bottom-Up Tree Topology

Permutation-Based Approximate Multiplier with High Accuracy.

A Novel Current Mode Approximate Multiplier Scheme Based on 4:2 and 5:2 Compressors with Low Power Consumption and High Speed in CNTFET Technology

Hardware-accuracy trade-offs for error-resilient applications using an ultra-efficient hybrid approximate multiplier

Ultra-Fast, High-Performance 8x8 Approximate Multipliers by a New Multicolumn 3,3:2 Inexact Compressor and its Derivatives

A Hardware- and Accuracy-Efficient Approximate Multiplier with Error Compensation for Neural Network and Image Processing Applications

Image Multiplication Using Novel 4:1 Approximate Compressor*

OPACT: Optimization of Approximate Compressor Tree for Approximate Multiplier

Design and Analysis of Approximate 4–2 Compressors for High-Accuracy Multipliers

Design and implementation of hybrid (radix-8 Booth and TRAM) approximate multiplier using 15-4 approximate compressors for image processing application

Implementation of High-speed Parallel Multiplier Based on Optimized Partial Product

Comparison and design of energy-efficient approximate multiplier schemes for image processing by CNTFET

Adaptable Approximate Multiplier Design Based on Input Distribution and Polarity