Ignore:
Timestamp:
Dec 13, 2011, 10:26:14 PM (8 years ago)
Author:
cameron
Message:

Related work

Location:
docs/HPCA2012/final_ieee
Files:
2 edited

Legend:

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

    r1774 r1779  
    4545byte-oriented character data into parallel bit streams (each with one
    4646bit per input byte) which permits text processing to exploit SIMD
    47 operations on modern processors. Our earlier work on inductive doubling
    48 instructions~\cite{CameronLin2009} discusses effective techniques to
    49 transform the text into the Parabix representation.  We
    50 have used Parabix to accelerate UTF-8 to UTF-16 transcoding
    51 \cite{Cameron2008}, string matching in protein identification
    52 \cite{JMBE:31@99}, and specific parts of a traditional
    53 recursive-descent XML parser~\cite{cameron-EuroPar2011}.
    54 
    55 
    56 
    57 
     47operations on modern processors \cite{Cameron2008, CameronHerdyLin2008}.
     48Parabix also incorporates a systematic and portable SIMD programming
     49framework based on our model inductive doubling instruction set
     50architecture~\cite{CameronLin2009}.    Most recently, a new
     51parallel scanning primitive \cite{cameron-EuroPar2011} has
     52been incorporated into the technology base to form the
     53basis of the Pablo compiler documented in this paper.
    5854
    5955
     
    149145
    150146
    151 
    152 
    153 
    154 %One area in which both servers and mobile devices devote considerable
    155 %computational effort into is in the processing of Extensible Markup
    156 %Language (XML) documents.  It was predicted that corporate servers
    157 %would see a ``growth in XML traffic\ldots from 15\% [of overall
    158 %network traffic] in 2004 to just under 48\% by 2008''
    159 %\cite{coyle2005}.  Further, ``from the point of view of server
    160 %efficiency[,] XML\ldots is the closest thing there is to a ubiquitous
    161 %computing workload'' \cite{leventhal2009}.  In other words, XML is the
    162 %quickly becoming the backbone of most server/server and client/server
    163 %%information exchanges.  Similarly, there is growing interest in the
    164 %use of mobile web services for personalization, context-awareness, and
    165 %content-adaptation of mobile web sites---most of which rely on XML
    166 %\cite{canali2009}.  Whether the end user realizes it or not, XML is
    167 %part of their daily life.
    168 
    169 %Why are XML parsers important ?
    170 %Talk about XML parsers and what they do in general.
    171 %Brief few lines about byte-at-time ?
    172 %What's new with Parabix style approach ?
    173 %Introduce Parabix1 and Parabix2 ?
    174 %Present overall quantiative improvements compared to other parsers.
    175 
    176 
    177 
    178 
  • docs/HPCA2012/final_ieee/10-related.tex

    r1774 r1779  
    2525instructions~\cite{sse4}.
    2626
    27 Parallel bitstreams were introduced by Cameron et
    28 al.~\cite{CameronHerdyLin2008} and used it to implement an efficient
    29 UTF-8 to 16 parser. Subsequent work ~\cite{cameron-EuroPar2011}
    30 accelerated specific phases in an XML parser using widely available
    31 SSE2 instructions and proposed an inductive doubling instruction set
    32 ~\cite{CameronLin2009}. In this paper, we have developed a generalized
     27Parallel bitstream methods were first introduced in application
     28to the problem of UTF-8 to UTF-16 transcoding \cite{Cameron2008}
     29and in the construction of a recursive descent parser based on
     30SSE2 instructions for bitstream transposition and character
     31class formation together with processor bit scan operations
     32to accelerate sequential scanning \cite{CameronHerdyLin2008}.
     33The technique has also been employed to accelerate string
     34matching operations in protein identification \cite{JMBE:31@99}.
     35The work on an inductive doubling instruction set
     36~\cite{CameronLin2009} established the basis for our
     37portable SIMD run-time environments, while the introduction
     38of parallel scanning primitives ~\cite{cameron-EuroPar2011}
     39has provided the basis for our Pablo compiler technology on
     40unbounded bitstreams.
     41
     42 In this paper, we have developed a generalized
    3343Parabix architecture and have described the software tool chain that
    3444programmers can use to build scalable text processing applications on
Note: See TracChangeset for help on using the changeset viewer.