An Approximate Communication Framework for Network-on-Chips

Chen, Yuechen; Louri, Ahmed

doi:10.1109/tpds.2020.2968068

Cited by 23 publications

(12 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We evaluate the proposed technique by comparing it with approximate communication framework (ACF) [26,38], Approx-NoC [32], AxBA [31], and the baseline (i.e., NoC with no approximation) from the communication efficiency perspective, which includes network latency and dynamic power consumption. The ACF includes a network interface packet approximation method using data truncation and a frequent pattern compression technique [38] with a learning-based quality management system [26]. Approx-NoC [27] includes an approximated frequent pattern compression technique and a quality control method that requires a designer to manually annotate the approximable values and their error tolerance.…”

Section: Discussionmentioning

confidence: 99%

“…Novel approximate network interfaces for the inference core, pre-processing core, memory controller, and shared cache have been proposed to implement ACT in NoCs. Compared to existing approximate communication techniques [27,31,38], ACT significantly reduces the transmitted data by efficiently approximating image classification applications. Compared to existing CNN approximation techniques [33,34,35,36,37], ACT eliminates the retraining process, which is time and energyconsuming.…”

Section: Discussionmentioning

confidence: 99%

“…Various data approximation methods have been proposed to reduce the packet size according to the approximation information. [27] and [38] have proposed lossy data compression techniques to further reduce the size of the error-resilient data based on the relative error. [28,39] have proposed techniques to reduce network congestion by dropping data in the packet prior to injecting it into the network.…”

Section: Approximate Communication Techniquesmentioning

confidence: 99%

“…Fig. 3 shows the network latency reduction normalized to the network with no approximation (baseline) for the existing approximate communication framework (ACF) [26,38]. This figure indicates that ACF only reduces network latency by less than 5% when applied to various state-of-the-art image classification applications [3,4,5,6,40,41,42,43,44,45,46,47] as executed on heterogeneous multi-core architectures with an NoC.…”

Section: Approximate Communication Techniquesmentioning

confidence: 99%

“…The proposed approximate network interface incorporates the approximation and recovery logic for the shared cache. As the shared cache handles requests from pre-processing cores and model inference cores [7], the approximation and recovery logics are used to approximate and recover image pixels, Approximate Communication Framework(ACF) [26,38]; Approx-NoC [27]; AxBA [31]; Proposed Technique model parameters, and inputs. Thus, the approximate network interface for shared cache uses the same design as the memory controller (Fig.…”

Section: Approximate Network Interface (Shared Cache)mentioning

confidence: 99%

See 4 more Smart Citations

A Technique for Approximate Communication in Network-on-Chips for Image Classification

Chen¹,

Liu²,

Lombardi³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Approximation is an effective technique for reducing power consumption and latency of on-chip communication in many computing applications. However, existing approximation techniques either achieve modest improvements in these metrics or require retraining after approximation, such when convolutional neural networks (CNNs) are employed. Since classifying many images introduces intensive on-chip communication, reductions in both network latency and power consumption are highly desired. In this paper, we propose an approximate communication technique (ACT) to improve the efficiency of on-chip communications for image classification applications. The proposed technique exploits the error-tolerance of the image classification process to reduce power consumption and latency of on-chip communications, resulting in better overall performance for image classification computation. This is achieved by incorporating novel quality control and data approximation mechanisms that reduce the packet size. In particular, the proposed quality control mechanisms identify the error-resilient variables and automatically adjust the error thresholds of the variables based on the image classification accuracy. The proposed data approximation mechanisms significantly reduce packet size when the variables are transmitted. The proposed technique reduces the number of flits in each data packet as well as the on-chip communication, while maintaining an excellent image classification accuracy. The cycle-accurate simulation results show that ACT achieves 23% in network latency reduction and 24% in dynamic power reduction compared to the existing approximate communication technique with less than 0.99% classification accuracy loss.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Approximate Communication Techniquesmentioning

confidence: 99%

Section: Approximate Communication Techniquesmentioning

confidence: 99%

Section: Approximate Network Interface (Shared Cache)mentioning

confidence: 99%

See 3 more Smart Citations

A Technique for Approximate Communication in Network-on-Chips for Image Classification

Chen¹,

Liu²,

Lombardi³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

RETRACTED: Design and implementation of network‐on‐chip router using multi‐priority based iterative round‐robin matching with slip

Singh

Ravindra²,

Naik

2022

Trans Emerging Tel Tech

View full text Add to dashboard Cite

Nowadays, network‐on‐chip (NoC) routers become important in several applications including mobile technology and digital communication. Multiple cores/processors are typically coupled using NoC routers, which considerably improve integrated circuit performance in terms of power, latency, throughput, and area. So, a novel 4 × 4 topology‐based NoC router using multi‐priority based iterative round‐robin matching with slip (MPiSLIP) routing algorithm is implemented in this work. The proposed work contains different operational blocks classification such as route computation using static straight allocator, route controlling using MPiSLIP arbiter, a crossbar switch, combined parallel virtual channel and switch allocation unit, look ahead bypass route computation (LBRC), and input‐output ports. The MPiSLIP routing mechanism also determines the priorities necessary for route development. Finally, by bypassing the switches and fulfilling all of the priories in a look‐ahead manner, LBRC is employed to establish the effective path between input ports and output ports. The simulation results show that the proposed MPiSLIP‐based NoC routing method resulted in a reduced area, power, and latency as compared to the open‐source NoCs, multi‐processor NoCs, multi‐core NoCs, and conventional routing algorithms.

show abstract

Diamont: Dynamic Monitoring of Uncertainty for Distributed Asynchronous Programs

Fernando

Joshi

Laurel

et al. 2021

Runtime Verification

View full text Add to dashboard Cite

An Approximate Communication Framework for Network-on-Chips

Cited by 23 publications

References 53 publications

A Technique for Approximate Communication in Network-on-Chips for Image Classification

A Technique for Approximate Communication in Network-on-Chips for Image Classification

RETRACTED: Design and implementation of network‐on‐chip router using multi‐priority based iterative round‐robin matching with slip

Diamont: Dynamic Monitoring of Uncertainty for Distributed Asynchronous Programs

Contact Info

Product

Resources

About