30 November 2021 in IEEE Transactions on Circuits and Systems I: Regular Papers
In-memory computing using emerging technologies such as resistive random-access memory (ReRAM) addresses the ‘von Neumann bottleneck’ and strengthens the present research impetus to overcome the memory wall. While many methods have been recently proposed to implement Boolean logic in memory, the latency of arithmetic circuits (adders and consequently multipliers) implemented as a sequence of such Boolean operations increases greatly with bit-width. Existing in-memory multipliers require $O(n^{2})$ cycles which is inefficient both in terms of latency and energy. In this work, we tackle this exorbitant latency by adopting Wallace Tree multiplier architecture and optimizing the addition operation in each phase of the Wallace Tree. Majority logic primitive was used for addition since it is better than NAND/NOR/IMPLY primitives. Furthermore, high degree of gate-level parallelism is employed at the array level by executing multiple majority gates in the columns of the array. In this manner, an in-memory multiplier of $O(n.log(n))$ latency is achieved which outperforms all reported in-memory multipliers. Furthermore, the proposed multiplier can be implemented in a regular transistor-accessed memory array without any major modifications to its peripheral circuitry and is also energy-efficient.
30 October 2018 in IEEE Transactions on Circuits and Systems II: Express Briefs
In this paper, we first theoretically re-defined output decimal carry in terms of majority gates and proposed a carry lookahead structure for calculating all the intermediate output carries. We have used this method for designing the multi-digit decimal adders. Theoretically, our best n-digit decimal adder design reduces the delay and area-delay product (ADP) by 50% compared with previous designs. We have implemented our designs using QCADesigner tool. The proposed QCADesigner based 8-digit PBA-BCD adder achieves over 38% less delay compared with the best existing designs.
14 June 2017 in IEEE Transactions on Circuits and Systems II: Express Briefs
We are proposing a lightweight and secure compressive sensing (CS) method for compression of data generated by image sensors. This method simultaneously compresses and encrypts image and video files, thereby reducing communication cost, storage space, runtime, and power/energy overhead. Our proposed method combines the CS technique with stream cipher to implement the secure CS. Furthermore, the usage of stream cipher to generate the measurement matrix results in minimal overhead to achieve the security objective. Our experimental studies using the field-programmable gate array prototype show that the security enhancements to the circuit comes, with no timing overhead, reduces 41%-84% of memory requirements and there is only a 8%-33% increase in the LUTs depending on the compression ratio.
29 December 2023 in IEEE Transactions on Circuits and Systems II: Express Briefs
Smart devices for image/video sensing are needed to work within the constraints of limited bandwidth and low computing capabilities. In this context, Block based Compressive Sensing (BCS) emerged as the most viable method for balancing image/video quality and transmission bandwidth computing overheads. However, in comparison with conventional image and video acquisition systems, BCS cannot reduce the bitrate due to its straightforward nature of system of linear equations, which still incurs high transmission and storage overhead. To address this shortcoming, in this paper we propose a novel Near Lossless Predictive Coding (NLPC) approach to compress BCS measurements. The NLPC method encodes the prediction error measurement between the target and current measurement, resulting in lower data size. We designed and implemented a complete BCS integrated with NLPC with scalar quantization (BCS-NLPC-SQ) and studied the image quality at different compression ratios with varying block sizes. The BCS-NLPC-SQ method can improve roughly on an average PSNR of +3.06 dB and the average SSIM gain of +0.11 with respect to the existing works. The synthesis results shows that, BCS-NLPC-SQ requires 83.01%, 69.03%, 53.26%, and 14.45% less area, power, ADP and PDP over JPEG compression and we have achieved an additional compression of up to 56.25% in the best case. Our proposed BCS-NLPC-SQ method outperformed the existing methods in terms of PSNR, SSIM, and bpp.
1 August 2022 in IEEE Transactions on Circuits and Systems I: Regular Papers
In-memory computing using emerging technologies such as Resistive Random-Access Memory (ReRAM) has been proposed as a promising substitute for future computing applications to address the ‘von Neumann bottleneck’. Multiplication is the key component for inner product computation in every digital signal processing (DSP) application and the complexity of multipliers increases greatly with bit-width. Distributed arithmetic (DA) using look-up tables and adder-shifter module has been proposed for inner product computation to achieve multiplier-less efficient DSP architectures, particularly when one of the vectors is a constant and known in advance. Due to the memory wall, DA can be made furthermore latency and energy-efficient when implemented ‘in memory’. In this work, for the first time, we propose two design techniques to compute inner product completely in memory using DA. This is accomplished by storing the precomputed look-up table contents in a ReRAM array and implementing adder-shifter module also in the same array. The adder-shifter is implemented in memory using majority gates which are in turn realized as READ operations in the memory array. Two methods of mapping: latency-optimized and area-optimized and their comparison in terms of latency and area are presented. The proposed method-1 achieves $\approx$ 60% energy savings compared to CMOS and the proposed method-2 achieves 10.59 times higher throughput compared to CMOS.
1 August 2015 in IET Computer Vision
Autonomous vehicles engaged in terrain exploration are typically equipped with a camera. The camera is subjected to vibration as the vehicle moves so that the videos captured require stabilisation to facilitate accurate interpretation by remote operators. Dedicated architectures for video stabilisation that offer high performance while consuming low area and power are desirable for this application. This study presents a pipelined very large‐scale integration architecture. It is based on exploiting the separability property of the two‐dimensional (2‐D) Sobel matrix and the 2‐D Gaussian filtering matrix to obtain an efficient corner point detection architecture. It also employs the coordinate rotation digital computer architecture for global motion vector calculation. The proposed architecture has been coded in Verilog and synthesised for a field programmable gate array (FPGA), which offers massive parallelism at fairly low power. The proposed architecture is shown to be highly area efficient. An FPGA‐based autonomous vehicle has been fabricated, and experiments with a camera mounted on the vehicle are presented and analysed.