Software-defined Significance-Driven Computing

Approximate computing has been an emerging programming and system design paradigm that has been proposed as a way to overcome the <br/>power-wall problem that hinders the scaling of the next generation of both high-end and mobile computing systems. Towards this<br/>end, a lot of researchers have been studying the effects of approximation to applications and those hardware modifications that<br/>allow increased power benefits for reduced reliability. In this work, we focus on runtime system modifications and task-based<br/>programming models that enable software-controlled, user-driven approximate computing.<br/><br/>We employ a systematic methodology that allows us to evaluate the potential energy and performance benefits of approximate<br/>computing using as building blocks unreliable hardware components. We present a set of extensions to OpenMP 4.0 that enable the <br/>programmer to define computations suitable for approximation. We introduce task-significance, a novel concept that describes the <br/>contribution of a task to the quality of the result. We use significance as a channel of communication from domain specific<br/>knowledge about applications towards the runtime-system, where we can optimise approximate execution depending on user<br/>constraints.<br/><br/>Finally, we show extensions to the Linux kernel that enable it to operate seamlessly on top of unreliable memory and provide a<br/>user-space interface for memory allocation from the unreliable portion of the physical memory. Having this framework in place<br/>allowed us to identify what we call the refresh-by-access property of applications that use dynamic random-access memory (DRAM).<br/>We use this property to implement techniques for task-based applications that minimise the probability of errors when using<br/>unreliable memory enabling increased quality and power efficiency when using unreliable DRAM.

Name dc:type.qualificationname

Doctor of Philosophy

Level dc:type.qualificationlevel

Doctoral Thesis

Grantor dc:publisher.institution

Queen's University Belfast

Year dc:date.issued

2017

Author dc:creator

• Chalios, Charalambos

Advisor dc:contributor.advisor

• Vandierendonck, Hans

• approximate computing
• significance-based computing
• parallel computing
• reliability
• task-based programming
• DRAM
• scheduling
• linux

Language dc:language

eng

Identifier

oai:pure.qub.ac.uk/portal:studenttheses/22a3cdcb-3773-4117-a06d-23f031539a36

OAI identifier oai:identifier

oai:pure.qub.ac.uk/portal:studenttheses/22a3cdcb-3773-4117-a06d-23f031539a36