• Bringing Resource Efficiency to Smartphones with Approximate Computing
The Client : Javna agencija za raziskovalno dejavnost RS
Project type: Research projects ARRS
Project duration: 2020 - 2021
  • Description

He breakdown of Moores law and Dennard scaling critically threatens the future growth of mobile computing. It indicates that further packing of computing resources cannot be sustained, should we wish to preserve the portability and energy frugality of our devices. Yet, as mobile computing affirms its central role in our everyday life, we entrust an ever increasing number of complex computational tasks to it.

In this proposal we aim to set foundations for a new mobile computing paradigm termed - approximate mobile computing (AMC). ACM is based on our insight that, depending on the context, a result need not be perfectly accurate in order to fulfil a users needs: when playing a video game outdoors a gamer tolerates imperfect 3D rendering, just as a person looking for recommendations does not mind getting a slightly shuffled order of nearby restaurants. Bringing ACM to realisation requires that the underlying foundation - the technology for enabling accuracy-adaptable computation - is implemented.

In the proposal we focus on 1) identification of approximate computing techniques (ACTs) that are suitable for mobile devices, specifically, smartphones; 2) implementation of selected ACTs on low level of the mobile computing stack, such as compiler and supporting library level, and 3) exposing control knobs for the dynamic tuning of ACTs, so that the result accuracy can be traded for resource usage at runtime.

Results of our work will provide a basis for future experimentation with AMC, in particular with respect to its context-aware adaptation and further experimentation with resource savings that AMC brings.

In this project we concentrate on the subset of work defined in the CARMA ERC StG proposal, namely, the original Work Package 1 - Mobile-Ready Approximate Computing Techniques (ACTs). This work package aims to realise the foundation needed for further adaptation of approximate computing on mobile devices according to the context of use, thus, enable resources to be saved when approximate computation satisfies a users needs for result accuracy.

In the project we will, just like in the original proposal, identify the most suitable ACTs for the mobile domain and implement them on a mobile platform. Due to budget and time constraints, we will focus on Android devices only. We will also enable result accuracy adaptation via approximation knobs, as stated in the original proposal. Other work packages (WP2-WP5) of the original proposal will not be realised, however, we will develop a prototype application of video encoding/decoding to demonstrate and quantify the benefits that AMC may bring.