ABSTRACT
Run-time resource management of heterogeneous multi-core systems is challenging due to i) dynamic workloads, that often result in ii) conflicting knob actuation decisions, which potentially iii) compromise on performance for thermal safety. We present a runtime resource management strategy for performance guarantees under power constraints using functionally approximate kernels that exploit accuracy-performance trade-offs within error resilient applications. Our controller integrates approximation with power knobs - DVFS, CPU quota, task migration - in coordinated manner to make performance-aware decisions on power management under variable workloads. Experimental results on Odroid XU3 show the effectiveness of this strategy in meeting performance requirements without power violations compared to existing solutions.
- W. Baek et al. Green: A Framework for Supporting Energy-Conscious Programming using Controlled Approximation. In PLDI, 2010. Google ScholarDigital Library
- R. Cochran et al. Pack & cap: adaptive dvfs and thread packing under power caps. In MICRO, 2011. Google ScholarDigital Library
- B. Donyanavard et al. SPARTA: Runtime Task Allocation for Energy Efficient Heterogeneous Many-cores. In Proc. of CODES+ISSS, pages 27:1--27:10, 2016. Google ScholarDigital Library
- H. Esmaeilzadeh et al. Neural Acceleration for General-Purpose Approximate Programs. In MICRO, 2012. Google ScholarDigital Library
- S. Pagani et al. TSP: Thermal Safe Power: Efficient Power Budgeting for many-core systems in dark silicon era. In CODES+ISSS, 2014. Google ScholarDigital Library
- F. Gaspar et al. Performance-Aware Task Management and Frequency Scaling in Embedded Systems. In In Proc. of SBAC-PAD, pages 65--72, 2014. Google ScholarDigital Library
- F Gaspar et al. A framework for application-guided task management on heterogeneous embedded systems. ACM TACO, 12(4):42, 2016. Google ScholarDigital Library
- H. Hoffmann et al. Application heartbeats: a generic interface for specifying program performance and goals in autonomous computing environments. In Proc. Int. Conf. on Autonomic computing, pages 79--88. ACM, 2010. Google ScholarDigital Library
- H. Hoffmann et al. Dynamic knobs for responsive power-aware computing. ACM SIGPLAN Notices, 2012. Google ScholarDigital Library
- S. Holmback et al. Performance Monitor Based Power Management for big. LITTLE Platforms. In Proc. HIPEAC Workshop on Energy Efficiency with Heterogeneous Computing, pages 1--6, 2015.Google Scholar
- C. Imes and H. Hoffmann. Minimizing Energy Under Performance Constraints on Embedded Platforms: Resource Allocation Heuristics for Homogeneous and single-ISA Heterogeneous Multi-cores. SIGBED Rev., 11(4):49--54, January 2015. Google ScholarDigital Library
- Norman P Jouppi et al. In-datacenter performance analysis of a tensor processing unit. SIGARCH Comput. Archit. News, 45(2):1--12, June 2017. Google ScholarDigital Library
- A. Kanduri et al. Approximation knob: Power capping meets energy efficiency. In In Proc. of ICCAD, pages 1--8. IEEE, 2016. Google ScholarDigital Library
- K. Ma and X. Wang. PGCapping: Exploiting power gating for power capping and core lifetime balancing in CMPs. PACT, 2012. Google ScholarDigital Library
- T.S. Muthukaruppan et al. Hierarchical power management for asymmetric multi-core in dark silicon era. In Proc. of DAC, pages 1--9, 2013. Google ScholarDigital Library
- D. Palomino et al. Thermal optimization using adaptive approximate computing for video coding. In In Proc. of DATE, pages 1207--1212, 2016. Google ScholarDigital Library
- A. Pathania et al. Integrated CPU-GPU power management for 3D mobile games. In Proc. of DAC, pages 1--6, 2014. Google ScholarDigital Library
- A. Rahmani et al. Dynamic power management for many-core platforms in the dark silicon era: A multi-objective control approach. In ISLPED, 2015.Google ScholarCross Ref
- A.M. Rahmani, P. Liljeberg, A. Hemani, A. Jantsch, and H. Tenhunen. The Dark Side of Silicon. Springer, 1st edition edition, 2016.Google Scholar
- H. Rexha et al. Core Level Utilization for Achieving Energy Efficiency in Heterogeneous Systems. In Proc. of PDP, pages 401--407, 2017.Google ScholarCross Ref
- P. Schulz et al. Latency critical iot applications in 5g: Perspective on the design of radio interface and network architecture. IEEE Communications Magazine, 55(2):70--78, 2017. Google ScholarDigital Library
- S. Sidiroglou et al. Managing performance vs. accuracy trade-offs with loop perforation. In FSE, 2011.Google ScholarDigital Library
- X. Sui et al. Proactive control of approximate programs. In Proc. of ASPLOS, pages 607--621. ACM, 2016. Google ScholarDigital Library
- C. Tan et al. Approximation-aware scheduling on heterogeneous multi-core architectures. In Proc. of ASP-DAC, pages 618--623, 2015.Google Scholar
- A. Vega et al. Crank it up or dial it down: Coordinated multiprocessor frequency and folding control. In MICRO, pages 210--221, 2013. Google ScholarDigital Library
- K. Yu et al. Power-aware task scheduling for big. LITTLE mobile processor. In Proc. Int. SoC Design Conf., pages 208--212, 2013.Google Scholar
Index Terms
- Approximation-aware coordinated power/performance management for heterogeneous multi-cores
Recommendations
QoS-aware stochastic power management for many-cores
DAC '18: Proceedings of the 55th Annual Design Automation ConferenceA many-core processor can execute hundreds of multi-threaded tasks in parallel on its 100s - 1000s of processing cores. When deployed in a Quality of Service (QoS)-based system, the many-core must execute a task at a target QoS. The amount of processing ...
Approximation-Aware Coordinated Power/Performance Management for Heterogeneous Multi-cores
2018 55th ACM/ESDA/IEEE Design Automation Conference (DAC)Run-time resource management of heterogeneous multi-core systems is challenging due to i) dynamic workloads, that often result in ii) conflicting knob actuation decisions, which potentially iii) compromise on performance for thermal safety. We present a ...
Coordinated energy management in heterogeneous processors
SC13 --The International Conference for High Performance Computing, Networking, Storage and AnalysisThis paper examines energy management in a heterogeneous processor consisting of an integrated CPU--GPU for high-performance computing HPC applications. Energy management for HPC applications is challenged by their uncompromising performance ...
Comments