• DocumentCode
    1783285
  • Title

    Mitigating the Mismatch between the Coherence Protocol and Conflict Detection in Hardware Transactional Memory

  • Author

    Lihang Zhao ; Lizhong Chen ; Draper, J.

  • fYear
    2014
  • fDate
    19-23 May 2014
  • Firstpage
    605
  • Lastpage
    614
  • Abstract
    Hardware Transactional Memory (HTM) usually piggybacks onto the cache coherence protocol to detect data access conflicts between transactions. We identify an intrinsic mismatch between the typical coherence scheme and transaction execution, which causes a sizable amount of unnecessary transaction aborts. This pathological behavior is called false aborting and increases the amount of wasted computation and on-chip communication. For the TM applications we studied, 41% of the transactional write requests incur false aborting. To combat false aborting, we propose Predictive Unicast and Notification (PUNO), a novel hardware mechanism to 1) replace the inefficient coherence multicast with a unicast scheme to prevent transactions from being disrupted unnecessarily and 2) restrain transaction polling through proactive notification. PUNO reduces transaction aborts by 61% and network traffic by 32% in workloads representative of future TM applications with a VLSI implementation area overhead of 0.41%.
  • Keywords
    cache storage; protocols; shared memory systems; HTM; PUNO; VLSI; cache coherence protocol; coherence scheme; data access conflict detection; false aborting; hardware transactional memory; inefficient coherence multicast; intrinsic mismatch identification; network traffic; on-chip communication; pathological behavior; predictive unicast and notification scheme; restrain transaction polling; transaction execution; transactional write requests; Benchmark testing; Coherence; Hardware; Protocols; Radiation detectors; System-on-chip; Unicast; Chip Multiprocessor; HTM; PUNO; Transactional Memory;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Parallel and Distributed Processing Symposium, 2014 IEEE 28th International
  • Conference_Location
    Phoenix, AZ
  • ISSN
    1530-2075
  • Print_ISBN
    978-1-4799-3799-8
  • Type

    conf

  • DOI
    10.1109/IPDPS.2014.69
  • Filename
    6877293