Survey on Approximate Computing and Its Intrinsic Fault Tolerance

Abstract

This work is a survey on approximate computing and its impact on fault tolerance, especially for safety-critical applications. It presents a multitude of approximation methodologies, which are typically applied at software, architecture, and circuit level. Those methodologies are discussed and compared on all their possible levels of implementations (some techniques are applied at more than one level). Approximation is also presented as a means to provide fault tolerance and high reliability: Traditional error masking techniques, such as triple modular redundancy, can be approximated and thus have their implementation and execution time costs reduced compared to the state of the art.

A Model-Based Framework to Assess the Reliability of Safety-Critical Applications

“…The memory footprint can be directly reduced by changing the data representation of the parameters (e.g., weights, activations) of an ANN implementation. Reducing memory footprint can lead to a reduction in the energy consumption of the implementation since there is a decrease in the amount of data transferred from/to the memory [9]. Methods of reducing the floating-point precision, or even the bit-width used for data representation, can significantly reduce the energy consumption with a cost of degradation in the outcomes of an application [10].…”

Investigating the Impact of Radiation-Induced Soft Errors on the Reliability of Approximate Computing Systems

“…Approximate computing techniques can be applied at different levels, where the approach can target software, hardware, or architectural level [11], [12]. Among all the possible techniques, in this work, we consider the data precision reduction, a technique that can be implemented both at the software and architectural levels.…”

An Approximate GEMM Unit for Energy-Efficient Object Detection

“…Adequately trained models are more error resilient with a lessened need for the accuracy of the results and computation, making them perfect candidates for approximate computing. Approximate computing is a new paradigm where an acceptable error is induced in the computing to achieve more energy-efficient processing [ 28 , 29 , 30 , 31 , 32 , 33 ]. It has been introduced at different system levels [ 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 ], and a large number of approximate arithmetic circuits have been designed to save chip area and energy [ 35 , 38 , 46 , 47 , 48 , 49 , 50 , 51 ].…”