Ignore:
Timestamp:
Aug 31, 2011, 3:13:36 PM (8 years ago)
Author:
ashriram
Message:

Minor bug fixes

File:
1 edited

Legend:

Unmodified
Added
Removed
  • docs/HPCA2012/01-intro.tex

    r1405 r1407  
    6868
    6969
     70
     71
     72\begin{figure}
     73\begin{center}
     74\includegraphics[width=85mm]{plots/performance_energy_chart.pdf}
     75\end{center}
     76\caption{XML Parser Technology Energy vs. Performance}
     77\label{perf-energy}
     78\end{figure}
     79
     80
     81
    7082Figure~\ref{perf-energy} showcases the overall efficiency of our
    7183framework. The Parabix-XML parser improves the
    72 performance %by ?$\times$ 
     84performance %by ?$\times$
    7385and energy efficiency %by ?$\times$
    7486several-fold compared
     
    105117
    106118
    107 \begin{comment}
    108 Figure~\ref{perf-energy} is an energy-performance scatter plot showing
    109 the results obtained.
    110 
    111 
    112 With all this XML processing, a substantial literature has arisen
    113 addressing XML processing performance in general and the performance
    114 of XML parsers in particular.  Nicola and John specifically identified
    115 XML parsing as a threat to database performance and outlined a number
    116 of potential directions for potential performance improvements
    117 \cite{NicolaJohn03}.  The nature of XML APIs was found to have a
    118 significant affect on performance with event-based SAX (Simple API for
    119 XML) parsers avoiding the tree construction costs of the more flexible
    120 DOM (Document Object Model) parsers \cite{Perkins05}.  The commercial
    121 importance of XML parsing spurred developments of hardware-based
    122 approaches including the development of a custom XML chip
    123 \cite{Leventhal2009} as well as FPGA-based implementations
    124 \cite{DaiNiZhu2010}.  However promising these approaches may be for
    125 particular niche applications, it is likely that the bulk of the
    126 world's XML processing workload will be carried out on commodity
    127 processors using software-based solutions.
    128 
    129 To accelerate XML parsing performance in software, most recent
    130 work has focused on parallelization.  The use of multicore
    131 parallelism for chip multiprocessors has attracted
    132 the attention of several groups \cite{ZhangPanChiu09, ParaDOM2009, LiWangLiuLi2009},
    133 while SIMD (Single Instruction Multiple Data) parallelism
    134 has been of interest to Intel in designing new SIMD instructions\cite{XMLSSE42}
    135 , as well as to the developers of parallel bit stream technology
    136 \cite{CameronHerdyLin2008,Cameron2009,Cameron2010}.
    137 Each of these approaches has shown considerable performance
    138 benefits over traditional sequential parsing techniques that follow the
    139 byte-at-a-time model.
    140 \end{comment}
    141 
    142 
    143 
    144 \begin{figure}
    145 \begin{center}
    146 \includegraphics[width=85mm]{plots/performance_energy_chart.pdf}
    147 \end{center}
    148 \caption{XML Parser Technology Energy vs. Performance}
    149 \label{perf-energy}
    150 \end{figure}
    151 
    152119The remainder of this paper is organized as follows.
    153120Section~\ref{section:background} presents background material on XML
     
    163130Section~\ref{section:scalability} compares the performance and energy
    164131efficiency of 128 bit SIMD extensions across three generations of
    165 intel processors and includes a comparison with the ARM Cortex-A8
     132Intel processors and includes a comparison with the ARM Cortex-A8
    166133processor.  Section~\ref{section:avx} examines the Intel's new 256-bit
    167134AVX technology and comments on the benefits and challenges compared to
     
    170137Parabix XML parser which seeks to exploit the SIMD units scattered
    171138across multiple cores.
     139
     140
    172141
    173142
Note: See TracChangeset for help on using the changeset viewer.