access icon free Hardware thread-context switching

Numerous proposals have advanced fine-grained thread migration as a mechanism to address power, performance, reliability and memory coherence problems. However, exploiting conventional context switch mechanisms carries significant overhead, limiting the granularity of thread movement. Proposed is a novel hardware context switching circuit that enables low-overhead hardware thread migration between cores in a single-chip multiprocessor. This switching circuit supports multiple simultaneous thread switches and can store the context of both currently running and time-multiplexed threads. The circuit drastically reduces the direct cost of context switches.

Inspec keywords: switching circuits; reliability; microprocessor chips; switching

Other keywords: fine-grained thread migration; multiple simultaneous thread switches; low-overhead hardware thread migration; hardware thread-context switching circuit; single-chip multiprocessor; cost reduction; time-multiplexed threads

Subjects: Microprocessors and microcomputers; Microprocessor chips; Logic and switching circuits

References

    1. 1)
      • 11. Sawalha, L., et al: ‘Phase-guided scheduling on single-ISA heterogeneous multicore processors,Proc. of 14th Euromicro Conf. on Dig. Sys. Design, Oulu, Finland, August-September 2011, pp. 736745.
    2. 2)
      • 6. Khan, O., et al: ‘DCC: a dependable cache coherence multicore architecture,Comput. Archit. Lett., 2011, 10, 1, pp. 1215 (doi: 10.1109/L-CA.2011.3).
    3. 3)
      • 5. Coşkun, A.K., et al: ‘Static and dynamic temperature-aware scheduling for multiprocessor SoCs,IEEE Trans. on Very Large Scale Integr. (VLSI) Syst., 2008, 16, 9, pp. 11271140 (doi: 10.1109/TVLSI.2008.2000726).
    4. 4)
      • 1. Ronen, R.: ‘The practical challenges of heterogeneous architectures,Invited Talk at SHAW-3, New Orleans, LA, USA, February 2012.
    5. 5)
      • 4. Sanchez, D., et al: ‘Flexible architectural support for fine-grain scheduling’, AspLos, Pittsburg, PA, Pennsylvania, pp. 311322, March 2010.
    6. 6)
      • 10. Muralimanohar, N., et al: ‘Optimizing NUCA organizations and wiring alternatives for large caches with CACTI 6.0,Proc. of 40th IEEE/ACM Symp. on Microarchitecture, Chicago, IL, USA, December 2007, pp. 314.
    7. 7)
      • 2. Kumar, R., et al: ‘Heterogeneous chip multiprocessors,Computer, 2005, 38, 11, pp. 3238 (doi: 10.1109/MC.2005.379).
    8. 8)
      • 9. Hwu, W.W., Patt, Y., ‘Checkpoint repair for high-performance out-of-order execution,IEEE Trans. Comput., 1987, C-36, 12, pp. 14961514 (doi: 10.1109/TC.1987.5009500).
    9. 9)
      • 7. Powell, M.D., et al: ‘Architectural core salvaging in a multi-core processor for hard-error tolerance,ACM SIGARCH News, 2009, 37, 3, pp. 93104 (doi: 10.1145/1555815.1555769).
    10. 10)
      • 3. Sawalha, L., et al: ‘Thread scheduling for heterogeneous multicore processors using phase identification,SIGMETRICS PER, 2011, 39, 3, pp. 125127 (doi: 10.1145/2160803.2160879).
    11. 11)
      • 8. Eggers, S., et al: ‘Simultaneous multithreading: a platform for next-generation processors,IEEE Micro, 1997, 17, 5, pp. 1219 (doi: 10.1109/40.621209).
    12. 12)
      • Muralimanohar, N.: `Optimizing NUCA organizations and wiring alternatives for large caches with CACTI 6.0', Proc. of 40th IEEE/ACM Symp. on Microarchitecture, December 2007, Chicago, IL, USA, p. 3–14.
    13. 13)
    14. 14)
    15. 15)
    16. 16)
    17. 17)
    18. 18)
    19. 19)
      • Sawalha, L.: `Phase-guided scheduling on single-ISA heterogeneous multicore processors', Proc. of 14th Euromicro Conf. on Dig. Sys. Design, August-September 2011, Oulu, Finland, p. 736–745.
    20. 20)
      • D. Sanchez . (2010) Flexible architectural support for fine-grain scheduling.
    21. 21)
    22. 22)
      • Ronen, R.: `The practical challenges of heterogeneous architectures', Invited Talk at SHAW-3, February 2012, New Orleans, LA, USA.
http://iet.metastore.ingenta.com/content/journals/10.1049/el.2012.2887
Loading

Related content

content/journals/10.1049/el.2012.2887
pub_keyword,iet_inspecKeyword,pub_concept
6
6
Loading