Integer-Value Encoding for Approximate On-Chip Communication

Amir Najafi,Ardalan Najafi,Lennart Bamberg,Alberto Garcia-Ortiz

doi:10.1109/access.2019.2959446

Amir Najafi, Ardalan Najafi + Show 2 more

Open Access

https://doi.org/10.1109/access.2019.2959446

Copy DOI

Abstract

Approximate computing has been broadly investigated in the computing domain to maintain the benefits of technology scaling. Recently, it has been extended to the communication domain. The idea is to trade transmission quality for energy or/and performance efficiency. Exploiting the integer-value representation of the transmitted signals in parallel buses, we propose two memoryless encoding approaches for approximate communications in order to reduce the integer value deviation. First, restricted to area-constrained applications, we propose a Combined Integer-Value (CIV) coding technique based on the swap and inversion of the input signals. Second, a Crosstalk-Avoidance-based Integer-Value (CAIV) coding technique for applications with a more relaxed area constraint is presented. The optimal mapping of the data words and codewords combined with the selective inversion of the input signals minimize the error magnitude in this coding technique. A comprehensive experimental result using a 65 nm commercial technology evaluates the proposed coding approaches. For example, our proposed CIV coding scheme improves the quality of the received images and the sampled radio communication signals with different modulation up to factor of 2 8 and 2 7 , respectively, as it compares with conventional transmission of signals with no coding. CAIV coder can improve the data transmission accuracy by about 2 18 as it compares with the conventional transmission. Furthermore, to assess the applicability of the proposed encoders, we carried out two case studies employing the scale-invariant feature transform algorithm and SOBEL edge detector.

Highlights

On-chip interconnects do not benefit as much from technology scaling as computation units do
Fault tolerance of the transmission is ensured by incorporating error control schemes such as forward error correction (FEC) and automatic repeat request (ARQ) [4]
We propose a combination of Swap- and InversionCodings (we refer to it as Combined Integer-Value (CIV) encoding) to address the real scenarios

Summary

INTRODUCTION

On-chip interconnects do not benefit as much from technology scaling as computation units do. For a more relaxed area-constrained applications, we propose a Crosstalk-Avoidance-based Integer-Value (CAIV) coding technique This coding technique is capable of error magnitude reduction in higher frequencies at the expense of a small area overhead. We refer to the conventional data transmission without coding as (8, 4, 2, 1), and the nomenclature for the combined coding of Fig. 2 will be (8 ̄, 2 ̄, 1, 4) The numbers in this nomenclature represent the weights associated to each signal transmitted through the bus, i.e., (wx , wx2 , wx , wx0 ). increase in area exceeds the gains of this shielding topology. The aim of this paper is to minimize the integer-value error in the bus structure defined in Fig. 1 using the dedicated coding techniques.

SIGNALS ALIGNMENT IN NARROW BUSES

SWAP-CODING: A CODING SCHEME CONSIDERING UNIFORMLY DISTRIBUTED SIGNALS

INVERSION-CODING: A CODING SCHEME FOR

EXPERIMENTAL VALIDATION

Findings

CONCLUSION