Bachelor and Master Theses

Title: Real-TIme Systems on Multicore Platforms
Subject: Computer Science
Level: Advanced
Description:

Parallel HW is an emerging current trend embraced by the industry. Multiple processing cores, multi-cores, provide real concurrency, enabling multiple threads to run simultaneously on a single processor. However, at this point the software is not able to fully exploit the benefits from simultaneously operating hardware resources. Existing concurrent programming methodology (most of them originating from decades ago) has been developed with a single core in mind, to support operating systems where multiple threads compete for the same CPU. Some efforts have been done to apply these uniprocessor techniques to multiprocessors (i.e., one processor per chip, not same as multicore which has several processor cores per chip, MPSoC), but this research is far from to be mature. Besides, multiprocessor research deals mostly with large, distributed systems, while multicores are primarily targeted to small embedded devices with scarce resources, where different applications are co-located on a single chip (e.g., CE devices, medical equipment, media processing, etc).

Real-time systems are used to execute parallel tasks on a single processor by sharing it between tasks, i.e., by using simulated parallelism. However, now there is a possibility to exploit true parallelism of mulicores, by letting parallel tasks to run on different cores. The question is if and how we can use exiting real-time methods for scheduling and resource reservation on multicore processors. What are the limitations with existing methods? Do we need to modify them? Do we need ned methods? Suggestions to new solutions? Thos are some of the questions to be addressed in this thesis.

More specifically, you will:

  • study multi-core (hardware) architecture to understand how it works

  • study existing state-of-the-art software methodologies for multi-core

  • study exisitng application areas for multicore (e.g., consumer devices)

  • identify the limitations with existing SW approaches to use multicore

  • identify the possibilities and the limitations of using real-time systems for multicores

  • study existing RT scheduling methods for multicores

  • give suggestion on novel scheduling methods for multicore


Prel. end date: 2007-06-20
Presentation date: 2007-06-20
Student: Apicha Suksompong asg06006@student.mdh.se
IDT supervisor: Damir Isovic
damir.isovic@mdh.se, +46-21-103173

  • Mälardalen University |
  • Box 883 |
  • 721 23 Västerås/Eskilstuna |
  • 021-101300, 016-153600 |
  • webmaster |
  • Latest update: 2017.10.14