DRUM: A Dynamic Range Unbiased Multiplier for approximate applications

Hashemi, Soheil; Bahar, R. Iris; Reda, Sherief

doi:10.1109/iccad.2015.7372600

Cited by 265 publications

(205 citation statements)

References 15 publications

Supporting

Mentioning

164

Contrasting

Order By: Relevance

“…Input Normalization and Batch Normalization Figure 2 illustrates a histogram of a sample error matrix which is used to simulate an MRE of ~3.6% and an SD of ~4.5%. Many of the reported approximate multipliers MREs have a near zeromean Gaussian distribution, this can be seen in the approximate multipliers [3] and [4]. Therefore, to provide a generic simulation that can be applicable to many approximate multiplier models, all the simulated MRE values in this paper were based on a near zero-mean Gaussian distribution.…”

Section: Simulation Environmentmentioning

confidence: 99%

“…To clearly demonstrate the benefit of this simulation, a mapping can be done between the simulated test cases and the reported performances of approximate multipliers in the literature. For example, DRUM [3] reported performance enhancements of 47%, 50% and 59% in the speed, area, and power, respectively with a cost of a near zero-mean Gaussian distribution with MRE=1.47% and SD=1.803%. This is very close to test case 2 in Table II with MRE=~1.4% and SD=~1.8% which also has a zero-mean Gaussian distribution.…”

Section: Training With Simulated Approximate Multiplier Errormentioning

confidence: 99%

“…However, these enhancements have a cost of inaccuracy in the output which is usually defined by the Mean Relative Error (MRE). As an example, the multiplier in [3] achieves performance enhancements of 47% in speed, 50% in area, and 59% in power. However, these enhancements have a cost of an inaccuracy in the output defined by a Gaussian MRE of 1.47% with Standard Deviation (SD) of 1.803%.…”

Section: Introductionmentioning

confidence: 99%

“…In a previous work [7], we have studied the impact of using approximate multipliers on the inference stage of a pre-trained CNN network. The simulated MRE and SD in [7] were selected to approximately simulate the reported inaccuracies by various approximate multipliers in the literature such as [3][4][5][6]. The work in [7] has demonstrated that with minimal cost of added inaccuracy, approximate multipliers can be used to enhance to significantly boost the inference performance of a pre-trained deep convolutional neural network (CNN) in terms of speed, power, and area.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Deep Learning Training with Simulated Approximate Multipliers

Hammad

El‐Sankary

2019

2019 IEEE International Conference on Robotics and Biomimetics (ROBIO)

View full text Add to dashboard Cite

This paper presents by simulation how approximate multipliers can be utilized to enhance the training performance of convolutional neural networks (CNNs). Approximate multipliers have significantly better performance in terms of speed, power, and area compared to exact multipliers. However, approximate multipliers have an inaccuracy which is defined in terms of the Mean Relative Error (MRE). To assess the applicability of approximate multipliers in enhancing CNN training performance, a simulation for the impact of approximate multipliers error on CNN training is presented. The paper demonstrates that using approximate multipliers for CNN training can significantly enhance the performance in terms of speed, power, and area at the cost of a small negative impact on the achieved accuracy. Additionally, the paper proposes a hybrid training method which mitigates this negative impact on the accuracy. Using the proposed hybrid method, the training can start using approximate multipliers then switches to exact multipliers for the last few epochs. Using this method, the performance benefits of approximate multipliers in terms of speed, power, and area can be attained for a large portion of the training stage. On the other hand, the negative impact on the accuracy is diminished by using the exact multipliers for the last epochs of training.

show abstract

Section: Simulation Environmentmentioning

confidence: 99%

Section: Training With Simulated Approximate Multiplier Errormentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep Learning Training with Simulated Approximate Multipliers

Hammad

El‐Sankary

2019

2019 IEEE International Conference on Robotics and Biomimetics (ROBIO)

View full text Add to dashboard Cite

show abstract

“…Another method consists in relaxing the timing constraints on the critical path of the adder, by splitting the carry chain like in speculative adders [15]- [17] or by transforming it into a false path as in the Carry Cut-Back (CCB) adder [18]. The Dynamic Range Unbiased Mulitiplier (DRUM) [19] features a dynamic-range selection scheme, which is essential for general purpose circuits.…”

mentioning

confidence: 99%

Design and Applications of Approximate Circuits by Gate-Level Pruning

Schlachter

Camus

Palem

et al. 2017

IEEE Trans. VLSI Syst.

View full text Add to dashboard Cite

Abstract-Energy-efficiency is a critical concern for many systems, ranging from IoT objects and mobile devices to highperformance computers. Moreover, after 40 years of prosperity, Moore's law is starting to show its economic and technical limits. Noticing that many circuits are over-engineered and that many applications are error-resilient or require less precision than offered by the existing hardware, approximate computing has emerged as a potential solution to pursue improvements of digital circuits. In this regard, a technique to systematically trade off accuracy in exchange for area, power and delay savings in digital circuits is proposed: Gate-Level Pruning. A CAD tool is build and integrated into a standard digital flow to offer a wide range of costs-accuracy tradeoffs for any conventional design. The methodology is first demonstrated on adders, achieving up to 78 % energy-delay-area reduction for 10 % mean relative error. It is then detailed how this methodology can be applied on a more complex system composed of a multitude of arithmetic blocks and memory: the Discrete Cosine Transform (DCT), which is a key building block for image and video processing applications. Even though arithmetic circuits represent less than 4 % of the entire DCT area, it is shown that the Gate-Level Pruning technique can lead to 21 % energy-delay-area savings over the entire system for a reasonable image quality loss of 24 dB. This significant saving is achieved thanks to the pruned arithmetic circuits which sets some nodes at constant values, enabling the synthesis tool to further simplify the circuit and memory.

show abstract

On Accelerating Data Analytics: An Introduction to the Approximate Computing Technique

Zervakis

2018

IoT for Smart Grids

View full text Add to dashboard Cite

DRUM: A Dynamic Range Unbiased Multiplier for approximate applications

Cited by 265 publications

References 15 publications

Deep Learning Training with Simulated Approximate Multipliers

Deep Learning Training with Simulated Approximate Multipliers

Design and Applications of Approximate Circuits by Gate-Level Pruning

On Accelerating Data Analytics: An Introduction to the Approximate Computing Technique

Contact Info

Product

Resources

About