Loading...

Design and Implementation of Software-Managed Energy Optimization Techniques for a Multi-Core Embedded Processor

Karimi Fatemi, Masoud | 2017

913 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 49261 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Ejlali, Alireza
  7. Abstract:
  8. Nowadays, due to the need of high performance systems, embedded systems have been redirected to multicore architectures. As energy consumption is one of the most important issues in embedded systems, careful considerations must be taken during the design of embedded systems to achieve low energy consumption. In this thesis, we propose the use of hybrid scratchpad memories to achieve low energy consumption. The hybrid memory uses both SRAM and STT-RAM memory. While STT-RAM has very low static energy consumption it suffers from high latency and energy consumption during write activities. In order to tackle this problem, we have devised a genetic algorithm that exploits architecture information and conducts application analysis to provide a near-optimal memory allocation for the input application. In this work, we assume that the system model is a frame-based real-time application where hard and soft realtime tasks are running together. Hence, the proposed algorithm is also responsible for meeting the frame deadline. Our experimental results show that compared to conventional systems, the proposed approach can provide system designs that reduce the maximum energy consumption by 37:4% and reduce average energy by 21:3
  9. Keywords:
  10. Software-based Techniques For Reliability Improvement ; Multicore Embedded System ; Nonvolatile Memory ; Real Time System ; Energy Optimization

 Digital Object List

 Bookmark