Runtime parallelization of legacy code on a transactional memory system

Matthew DeVuyst, Dean M. Tullsen, Seon Wook Kim

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    16 Citations (Scopus)

    Abstract

    This paper proposes a new runtime parallelization technique, based on a dynamic optimization framework, to automatically parallelize single-threaded legacy programs. It heavily leverages the optimistic concurrency of transactional memory. This work addresses a number of challenges posed by this type of parallelization and quantifies the trade-offs of some of the design decisions, such as how to select good loops for parallelization, how to partition the iteration space among parallel threads, how to handle loop-carried dependencies, and how to transition from serial to parallel execution and back. The simulated implementation of runtime parallelization shows a potential speedup of 1.36 for the NAS benchmarks and a 1.34 speedup for the SPEC 2000 CPU floating point benchmarks when using two cores for parallel execution.

    Original languageEnglish
    Title of host publicationHiPEAC'11 - Proceedings of the 6th International Conference on High Performance and Embedded Architectures and Compilers
    Pages127-136
    Number of pages10
    DOIs
    Publication statusPublished - 2011
    Event6th International Conference on High Performance and Embedded Architectures and Compilers, HiPEAC'11 - Heraklion, Crete, Greece
    Duration: 2011 Jan 242011 Jan 26

    Publication series

    NameHiPEAC'11 - Proceedings of the 6th International Conference on High Performance and Embedded Architectures and Compilers

    Other

    Other6th International Conference on High Performance and Embedded Architectures and Compilers, HiPEAC'11
    Country/TerritoryGreece
    CityHeraklion, Crete
    Period11/1/2411/1/26

    Keywords

    • Dynamic optimization
    • Parallelization
    • Transactional memory

    ASJC Scopus subject areas

    • Hardware and Architecture
    • Software
    • Electrical and Electronic Engineering

    Fingerprint

    Dive into the research topics of 'Runtime parallelization of legacy code on a transactional memory system'. Together they form a unique fingerprint.

    Cite this