Realizing Parsimonious Systems through Inexact/Approximate Computing: A Survey and Future Directions

摘要

The notion of exact computation, where outputs of the computational element (circuit) have precise deterministic values, has been pervasive in the computing domain for many decades owing to the overwhelming success of the integrated circuit design using reliable transistors, particularly in CMOS technology. However, it is facing serious challenges today as diminishing transistor sizes driven by Moore's law are leading to increasing process variations arising as lithographic scaling lags behind device scaling, and due to increasing parameter variations owning to perturbations such as (thermal) noise. While one obvious way to counter the antagonistic effects of this scaling induced "inexactness" is through error-correction mechanisms, a radically different approach was developed by Palem which we refer to now as inexact/approximate circuit design1. Inexact circuit design refers to an approach to realizing information processing frameworks-transistors, gates, data-path elements or more macroscopic engines-which are deliberately designed to be erroneous and used as such without adding any error-correction or compensatory mechanisms, in return for savings in energy, performance and/or area. This approach has been receiving increasing prominence of late as a consequence of the variations/perturbations stated above and significantly, the quest for ultra-low energy systems emanating from the desire for a longer battery life and a general need for mitigating the rising Information Technology (IT) sector's carbon footprint.