Changeset 1290


Ignore:
Timestamp:
Aug 8, 2011, 1:53:24 PM (8 years ago)
Author:
ksherdy
Message:

Minor edits to improve readability and flow.

File:
1 edited

Legend:

Unmodified
Added
Removed
  • docs/PACT2011/01-intro.tex

    r1095 r1290  
    33Extensible Markup Language (XML) is a core technology standard
    44of the World Wide Web Consortium (W3C) that provides a common
    5 framework for encoding and communicating structured information
    6 of all kinds.  In applications ranging from Office Open XML in
     5framework for encoding and communicating structured information
     6In applications ranging from Office Open XML in
    77Microsoft Office to NDFD XML of the NOAA National Weather
    88Service, from KML in Google Earth to Castor XML in the Martian Rovers,
     
    2727including the development of a custom XML chip \cite{Leventhal2009}
    2828as well as FPGA-based implementations \cite{DaiNiZhu2010}.
    29 As promising as these approaches may be for particular niche applications,
    30 however, it is still likely that the bulk of the world's XML
     29However promising these approaches may be for particular niche applications,
     30it is likely that the bulk of the world's XML
    3131processing workload will be carried out on commodity processors
    3232using software-based solutions.
     
    3636parallelism for chip multiprocessors has attracted
    3737the attention of several groups \cite{ZhangPanChiu09, ParaDOM2009, LiWangLiuLi2009},
    38 while SIMD (single-instruction multiple data) parallelism
     38while SIMD (Single Instruction Multiple Data) parallelism
    3939has been of interest to Intel in designing new SIMD instructions\cite{XMLSSE42}
    40 as well as to the developers of parallel bit stream technology
     40, as well as to the developers of parallel bit stream technology
    4141\cite{CameronHerdyLin2008,Cameron2009,Cameron2010}.
    4242Each of these approaches has shown considerable performance
    43 benefits over traditional sequential parsing following the
     43benefits over traditional sequential parsing techniques that follow the
    4444byte-at-a-time model.
    4545
    46 With a focus on performance, however, relatively less attention
    47 has been paid to reducing energy consumption.  For example, in addressing
    48 performance through multicore parallelism, one generally has to
     46With this focus on performance however, relatively little attention
     47has been paid on reducing energy consumption in XML processing.  For example, in addressing
     48performance through multicore parallelism, one generally must
    4949pay an energy price for performance gains because of the
    5050increased processing required for synchronization.   
    51 A focus on reduction of energy consumption is a key topic in this
    52 paper, in which we study the energy and performance
     51This focus on reduction of energy consumption is a key topic in this
     52paper. We study the energy and performance
    5353characteristics of several XML parsers across three generations
    5454of x86-64 processor technology.  The parsers we consider are
    55 the widely used byte-at-a-time parsers Expat and Xerces as well the
     55the widely used byte-at-a-time parsers Expat and Xerces, as well the
    5656Parabix1 and Parabix2 parsers based on parallel bit stream technology. 
    57 A compelling result is that
    58 the performance benefits of parallel bit stream technology
    59 translate directly and proportionately to substantial energy savings.
     57A compelling result is that the performance benefits of parallel bit stream technology
     58translate directly and proportionally to substantial energy savings.
    6059Figure \ref{perf-energy} is an energy-performance scatter plot
    61 showing the results we obtain for the four parsers.
     60showing the results obtained.
    6261
    6362\begin{figure}
     
    6968\end{figure}
    7069
    71 
    72 
    7370The remainder of this paper is organized as follows.
    7471Section 2 presents background material on XML parsing
    75 and traditional parsing methods.  Section 3 then reviews
     72and traditional parsing methods.  Section 3 reviews
    7673parallel bit stream technology as applied to
    7774XML parsing in the Parabix1 and Parabix2 parsers.
    78 Section 4 then introduces our methodology and approach
     75Section 4 introduces our methodology and approach
    7976for the performance and energy study tackled in the
    8077remainder of the paper.  Section 5 presents a
     
    8380number of microarchitectural issues including cache
    8481misses, branch mispredictions, SIMD instruction counts
    85 and so on.  Section 6 then looks at scalability and
     82and so forth.  Section 6 examines scalability and
    8683performance gains through three generations of Intel
    87 architecture culminating with performance assessment
     84architecture culminating with a performance assessment
    8885on our two week-old \SB\ test machine.
    8986Section 7 looks specifically at issues in applying
    9087the new 256-bit AVX technology to parallel bit stream
    9188technology and notes that the major performance benefit
    92 seen so far is a result of the change to 3-operand
    93 instruction form.  Section 8 concludes the paper with
     89seen so far results from the change to the non-destructive three-operand
     90instruction format.  Section 8 concludes with
    9491a discussion of ongoing work and further research directions.
    9592
Note: See TracChangeset for help on using the changeset viewer.