Summary

We observe from Figure 6.3 that our energy-credit based scheduler consumes 5.5% and 3.4% less energy than CFS with the on-demand governor and CFS with the conservative governor respectively under the light-weight workload. CFS consumes more energy be-cause CFS assigns light-weight tasks to the big core cluster. Also our scheduler consumes

29.8% and 6.3% less energy than CFS with the on-demand governor and CFS with the con-servative governor respectively under the median-weight workload. ECS consumes less energy in this case because it can classify the task and accurately adjust CPU frequency according to the total workload for each core.

VLC can smoothly play the video with ECS under heavy-weight workload because ECS assigns all of heavy-weight tasks to the big core cluster. VLC cannot smoothly play the video on CFS with either on-demand or conservative governor because CFS assigns some of heavy tasks to the little core cluster.

Chapter 7 Conclusion

In this paper, we design an energy-credit based scheduler for throughput guaranteed tasks on an asymmetric multi-core platform. The proposed scheduler consists of four key components - task classification, frequency selection, time assignment, and task schedul-ing. The system classifies tasks suitable for big and little cores respectively, determines the frequency of each core, assigns the percentage of time each task should run on each core, and ensures that tasks will only run on cores they are assigned, and tasks will receive the percentage of CPU time they are granted on the cores they are assigned. We imple-ment our energy credit-based scheduler by adding a throughput guaranteed task scheduling class within Linux. The experiment results indicate that our proposed scheduler consumes 29.8% and 6.3% less energy than the Completely Fair Scheduler with on-demand and con-servative frequency governors respectively.

Bibliography

[1] A. Das A. K. Singh and A. Kumar. Energy optimization by exploiting execution slacks in streaming applications on multiprocessor systems. Design Automation Conference, 2013.

[2] Akash Kumar Anup Das and Bharadwaj Veeravalli. Reliability and energy-aware mapping and scheduling of multimedia applications on multiprocessor systems.

IEEE Transactions on Parallel and Distributed Systems, 2016.

[3] ARM. Streamline performance analyzer, ds-5 development studio. https:

//developer.arm.com/products/software-development-tools/ds-5-development-studio/streamline/overview.

[4] Robert D. Blumofe and Dionisios Papadopoulos. The performance of work steal-ing in multiprogrammed environments (extended abstract. In In Proceedsteal-ings of the 1998 ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems, Poster Session, pages 266–267. ACM Press, 1998.

[5] P. Bogdan, S. Garg, and U.Y. Ogras. Energy-efficient computing from systems-on-chip to micro-server and data centers. In Green Computing Conference and Sus-tainable Computing Conference (IGSC), 2015 Sixth International, pages 1–6, Dec 2015.

[6] Quan Chen and Minyi Guo. Adaptive workload-aware task scheduling for single-isa asymmetric multicore architectures. ACM Trans. Archit. Code Optim., 11(1):

[7] FFmpeg. https://ffmpeg.org/.

[8] Andrei Frumusanu. The samsung exynos 7420 deep dive - inside a modern 14nm soc, 2015.

[9] Peter Greenhalgh. Big. little processing with arm cortex-a15 & cortex-a7. ARM White Paper, 2011.

[10] Vishal Gupta, Min Lee, and Karsten Schwan. Heterovisor: Exploiting resource het-erogeneity to enhance the elasticity of cloud platforms. In Proceedings of the 11th ACM SIGPLAN/SIGOPS International Conference on Virtual Execution Environ-ments, VEE ’15, pages 79–92, New York, NY, USA, 2015. ACM.

[11] JUAN HAMERS and LIEVEN EECKHOUT. Exploiting media stream similarity for energy-efficient decoding and resource prediction. ACM Transactions on Embedded Computing Systems, 2012.

[12] Brian Jeff. big.little technology moves towards fully heterogeneous global task scheduling. November 2013.

[13] David Atienza Karim Kanoun, Nicholas Mastronarde and Mihaela van der Schaar.

Online energy-efficient task-graph scheduling for multicore platforms. IEEE Trans-actions on Computer-Aided Design of Integrated Circuits and Systems, 2014.

[14] Myungsun Kim, Soonhyun Noh, Sungju Huh, and Seongsoo Hong. Fair-share scheduling for performance-asymmetric multicore architecture via scaled virtual runtime. In Embedded and Real-Time Computing Systems and Applications (RTCSA), 2015 IEEE 21st International Conference on, pages 60–69, Aug 2015.

[15] Youngjin Kwon, Changdae Kim, Seungryoul Maeng, and Jaehyuk Huh. Virtualizing performance asymmetric multi-core systems. In Proceedings of the 38th Annual International Symposium on Computer Architecture, ISCA ’11, pages 45–56, New York, NY, USA, 2011. ACM.

[16] VLC media player. http://www.videolan.org/vlc/.

[17] Chandandeep Singh Pabla. Completely fair scheduler. Linux J., 2009(184), August 2009.

[18] Padmanabhan Pillai and Kang G Shin. Real-time dynamic voltage scaling for low-power embedded operating systems. In ACM SIGOPS Operating Systems Review, volume 35, pages 89–102. ACM, 2001.

[19] Nicolas Pitre. Linux support for arm big.little. http://lwn.net/Articles/

481055/, 2012.

[20] Chin-Chiang Pan Po-Hsien Tseng, Pi-Cheng Hsiu and Tei-Wei Kuo. User-centric energy-efficient scheduling on multi-core mobile devices. Design Automation Con-ference, 2014.

[21] Rafael Rodríguez-Sánchez and Enrique S Quintana-Ortí. Architecture-aware opti-mization of an hevc decoder on asymmetric multicore processors. arXiv preprint arXiv:1601.05313, 2016.

[22] Elsayed Saad, Medhat Awadalla, Mohamed Shalan, and Abdullah Elewi. Energy-aware task partitioning on heterogeneous multiprocessor platforms. arXiv preprint arXiv:1206.0396, 2012.

[23] Wonik Seo, Daegil Im, Jeongim Choi, and Jaehyuk Huh. Big or little: A study of mobile interactive applications on an asymmetric multi-core platform. In Workload Characterization (IISWC), 2015 IEEE International Symposium on, pages 1–11, Oct 2015.

[24] William Thies, Michal Karczmarek, and Saman Amarasinghe. Compiler Construc-tion, chapter StreamIt: A Language for Streaming Applications, pages 179–196.

Springer Berlin Heidelberg, Berlin, Heidelberg, 2002.

[25] Frances Yao, Alan Demers, and Scott Shenker. A scheduling model for reduced cpu energy. In Foundations of Computer Science, 1995. Proceedings., 36th Annual Symposium on, pages 374–382. IEEE, 1995.

[26] Yuan-Hao Chang Yu-Ming Chang, Pi-Cheng Hsiu and Che-Wei Chang. A resource-driven dvfs scheme for smart handheld devices. ACM Transactions on Embedded Computing Systems, 2013.