White Rose University Consortium logo
University of Leeds logo University of Sheffield logo York University logo

Hardware-accelerated Evolutionary Hard Real-Time Task Mapping for Wormhole Network-on-Chip with Priority-Preemptive Arbitration

Ma, Yunfeng (2016) Hardware-accelerated Evolutionary Hard Real-Time Task Mapping for Wormhole Network-on-Chip with Priority-Preemptive Arbitration. PhD thesis, University of York.

[img]
Preview
Text
yunfeng Ma PhD Thesis.pdf - Examined Thesis (PDF)
Available under License Creative Commons Attribution-Noncommercial-No Derivative Works 2.0 UK: England & Wales.

Download (11Mb) | Preview

Abstract

Network-on-Chip (NoC) is an alternative on-chip interconnection paradigm to replace existing ones such as Point-to-Point and shared bus. NoCs designed for hard real-time systems need to guarantee the system timing performance, even in the worst-case scenario. A carefully planned task mapping which indicates how tasks are distributed on a NoC platform can improve or guarantee their timing performance. While existing offline mapping optimisations can satisfy timing requirements, this is obtained by sacrificing the flexibility of the system. In addition, the design exploration process will be prolonged with the continuous enlargement of the design space. Online mapping optimisations, by contrast, are affected by low success rates for remapping or a lack of guarantee of systems timing performance after remapping, especially in hard real-time systems. The existing limitations therefore motivate this research to concentrate on the mapping optimisation of real-time NoCs, and specifically dynamic task allocation in hard real-time systems. Four techniques and implementations are proposed to address this issue. The first enhances the evaluation efficiency of a hard real-time evaluation method from a theoretical point of view. The second technique addresses the evaluation efficiency from a practical point of view to enable online hard real-time timing analysis. The third technique advocates a dynamic mapper to enhance the remapping success rate with the accelerated model and architecture. The final technique yields a dynamic mapping algorithm that can search schedulable task allocation for hard real-time NoCs at run time, while simultaneously reducing the task migration cost after remapping.

Item Type: Thesis (PhD)
Academic Units: The University of York > Computer Science (York)
Identification Number/EthosID: uk.bl.ethos.745711
Depositing User: Mr Yunfeng Ma
Date Deposited: 11 Jun 2018 09:48
Last Modified: 24 Jul 2018 15:24
URI: http://etheses.whiterose.ac.uk/id/eprint/20446

You can contact us about this item. If you need to make a general enquiry, please see the Contact us page.

Actions (repository staff only: login required)