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.

Novel Bit-Parallel Multiplier for Gf(2(M)) Defined by All-One Polynomial Using Generalized Karatsuba Algorithm

Low complexity bit parallel multiplier for GF(2m) generated by equally-spaced trinomials

Research of Fast Modular Multiplier for a Class of Finite Fields

Implementation of Bit-Parallel Multiplier over Finite Field

Efficient bit-parallel multiplier for finite field of trinomial using SPB and its weakly dual basis.

A Parallel Architecture for Computing Multiplication in GF (2M)

Speedup of bit-parallel Karatsuba multiplier in GF(2m) generated by trinomials

Scalable Karatsuba multiplier over finite field GF (2m )

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

A Chinese Remainder Theorem Approach to Bit-Parallel GF(2^n) Polynomial Basis Multipliers for Irreducible Trinomials

A Survey of Some Recent Bit-Parallel GF(2n) Multipliers

Fast Bit Parallel-Shifted Polynomial Basis Multipliers in GF(2n)

Low-space complexity digit-serial multiplier based on modified polynomial basis over GF (2m)

Overlap-free Karatsuba-Ofman Polynomial Multiplication Algorithms

A New Approach to Subquadratic Space Complexity Parallel Multipliers for Extended Binary Fields

Fast Bit-Parallel GF(2^n) Multiplier for All Trinomials

Low-Complexity Digit-Serial and Scalable SPB/GPB Multipliers over Large Binary Extension Fields Using (B,2)-Way Karatsuba Decomposition

Low Complexity Bit-Parallel Normal Bases Multipliers for Gf(2(N))

Optimized Interpolation of Four-Term Karatsuba Multiplication and a Method of Avoiding Negative Multiplicands

Efficient Modulo <formula formulatype="inline"><tex Notation="TeX">$2^{n}+1$</tex></formula> Multipliers

Low-Latency Scalable Modular Multiplier Without Montgomery Algorithm