Ignore:
Timestamp:
Aug 31, 2011, 7:33:53 PM (8 years ago)
Author:
ashriram
Message:

Final Final

File:
1 edited

Legend:

Unmodified
Added
Removed
  • docs/HPCA2012/09-pipeline.tex

    r1412 r1419  
    8484
    8585Figure~\ref{multithread_perf} demonstrates the performance improvement
    86 achieved by pipelined Parabix-XML in comparison with the single-threaded
    87 version.  The 4-threaded version is $\simeq2\times$ faster compared to
    88 the single threaded version and achieves $\simeq2.7$ cycles per input
    89 byte by exploiting SIMD units of all \SB{}'s cores.  This performance approaches the
    90 1 cycle per byte
    91 performance of custom hardware solutions \ref{DaiNiZhu2010}. Parabix demonstrates the
    92 potential to enable an entire new class of applications, text
    93 processing, to exploit multicores.
     86achieved by pipelined Parabix-XML in comparison with the
     87single-threaded version.  The 4-threaded version is $\simeq2\times$
     88faster compared to the single threaded version and achieves
     89$\simeq2.7$ cycles per input byte by exploiting SIMD units of all
     90\SB{}'s cores.  This performance approaches the 1 cycle per byte
     91performance of custom hardware solutions~\cite{DaiNiZhu2010}. Parabix
     92demonstrates the potential to enable an entire new class of
     93applications, text processing, to exploit multicores.
    9494
    9595
Note: See TracChangeset for help on using the changeset viewer.