source: docs/Working/icXML/arch-overview.tex @ 2496

Last change on this file since 2496 was 2496, checked in by nmedfort, 7 years ago

temp checkin

File size: 8.3 KB
Line 
1\subsection{Overview}
2
3\icXML{} is more than an optimized version of Xerces. Many components were grouped, restructured and
4rearchitected with pipeline parallelism in mind.
5In this section, we highlight the core differences between the two systems.
6As shown in Figure \ref{fig:xerces-arch}, Xerces
7is comprised of five main modules: the transcoder, reader, scanner, namespace binder, and validator.
8The {\it Transcoder} converts source data into UTF-16 before Xerces parses it as XML;
9the majority of the character set encoding validation is performed as a byproduct of this process.
10The {\it Reader} is responsible for the streaming and buffering of all raw and transcoded (UTF-16) text;
11it keeps track of the current line/column of the cursor (which is reported to the end user in
12the unlikely event that the input contains an error), performs all line-break normalization
13and validates context-specific character set issues, such as tokenization of qualified-names and
14ensures each character is legal w.r.t. the XML specification.
15The {\it Scanner} pulls data through the reader and constructs the intermediate (and near-final)
16representation of the document; it deals with all issues related to entity expansion, validates
17the XML well-formedness constraints and any character set encoding issues that cannot
18be completely handled by the reader or transcoder (e.g., surrogate characters, validation
19and normalization of character references, etc.)
20The {\it Namespace Binder}, which is a core piece of their element stack, is primarily tasked
21with handling namespace scoping issues between different XML vocabularies and faciliates
22the scanner with the construction and utilization of Schema grammar structures.
23The {\it Validator} takes the intermediate representation produced by the Scanner (and
24potentially annotated by the Namespace Binder) and assesses whether the final output matches
25the user-defined DTD and Schema grammar(s) before passing the information to the end-user.
26
27\begin{figure}
28\begin{center}
29\includegraphics[width=0.15\textwidth]{plots/xerces.pdf}
30\caption{Xerces Architecture} 
31\label{fig:xerces-arch}
32\end{center}
33\end{figure}
34
35In \icXML{} functions are grouped into logical components.
36As shown in Figure \ref{fig:icxml-arch}, two major categories exist: (1) the \PS{} and (2) the \MP{}.
37All tasks in (1) use the Parabix Framework \cite{HPCA2012}, which represents data as a set of parallel bit streams.
38The {\it Character Set Adapter}, discussed in Section \ref{arch:character-set-adapter},
39mirrors Xerces's Transcoder duties; however instead of producing UTF-16 it produces a
40set of lexical bit streams, similar to those shown in Figure \ref{fig:parabix1}.
41These lexical bit streams are later transformed into UTF-16 in the Content Buffer Generator, after additional processing is performed.
42The first precursor to producing UTF-16 is the {\it Parallel Markup Parser} phase.
43It takes the lexical streams and produces a set of marker bit streams in which a 1-bit identifies
44significant positions within the input data. One bit stream for each of the critical piece of information is created, such as
45the beginning and ending of start tags, end tags, element names, attribute names, attribute values and content.
46Intra-element well-formedness validation is performed as an artifact of this process.
47Like Xerces, \icXML{} must provide the Line and Column position of each error.
48The {\it Line-Column Tracker} uses the lexical information to keep track of the cursor position(s) through the use of an
49optimized population count algorithm; this is described in Section \ref{section:arch:errorhandling}.
50From here, two major data-independent branches remain: the {\bf symbol resolver} and the {\bf content stream generator}.
51% The output of both are required by the \MP{}.
52Apart from the Parabix framework, another core difference between Xerces and \icXML{} is the use of symbols.
53A typical XML document will contain relatively few unique element and attribute names---but each of them will occur frequently throughout the document.
54In \icXML{}, names are represented by distinct symbol structures and global identifiers (GIDs).
55Using the information produced by the parallel markup parser, the {\it Symbol Resolver} uses a bitscan intrinsic to
56iterate through a symbol bit stream (64-bits at a time) to generate a set of GIDs.
57% It keys each symbol on its raw data representation, which means it can potentially be run in parallel with the content stream generator.
58One of the main advantages of using GIDs is that grammar information can be associated with the symbol itself and help bypass
59the lookup cost in the validation process.
60The final component of the \PS{} is the {\it Content Stream Generator}. This component has a multitude of
61responsibilities, which will be discussed in Section \ref{sec:parfilter}, but its primary function is to produce
62near-final UTF-16 content.
63
64The {\it \MP{}} parses a compressed representation of the XML document, generated by the
65symbol resolver and content stream generator, to validate and produce the final (sequential) output for the end user.
66The {\it WF checker} performs all remaining inter-element wellformedness validation that would be too costly
67to perform in bitspace, such as ensuring every start tag has a matching end tag.
68The {\it Namespace Processor} replaces Xerces's namespace binding functionality. Unlike Xerces,
69this is performed as a discrete phase and simply produces a set of URI identifiers (URI IDs), to
70be associated with each occurrence of a symbol.
71This is discussed in Section \ref{section:arch:namespacehandling}.
72The final {\it Validation} process is responsible for the same tasks as Xerces's validator, however,
73the majority of the grammar look up operations are performed beforehand and stored within the symbols themselves.
74
75\begin{figure}
76\includegraphics[width=0.50\textwidth]{plots/icxml.pdf}
77\caption{\icXML{} Architecture}
78\label{fig:icxml-arch}
79\end{figure}
80
81% Probably not the right area but should we discuss issues with Xerces design that we tried to correct?
82% - over-reliance on hash tables when domain knowledge dictated none would be needed
83% - constant buffering of text to ensure that every QName/NCName and content was contained within a single string
84% - abundant use of heap allocated memory
85% - text conversions done in multiple areas
86% - poor cache utilization; attempted to improve by using smaller layers of tasks in bulk
87
88% As the previous section aluded, the greatest difference between sequential parsing methods
89% and the Parabix parsing model is how data is processed.
90% Consider Figure \ref{fig:parabix1} again. In it, the start tags are located independent of the end
91% tags. In order to produce Xerces-equivalent output, icXML must emit the start and end tag
92% events in sequential order, with all attribute data associated with the correct tag.
93%
94%
95
96% The Parabix framework, however, does not allow for this (and would be hindered performance wise if
97% forced to.)
98% Thus our first question was, ``How can we how can we take full advantage
99% of Parabix whilst producing Xerces-equivalent output?'' Our answer came by analyzing what Xerces produced
100% when given an input text.
101%
102% By analyzing Xerces internal data structures and its produced output, two major observations were obvious:
103% (1) input data is transcoded into UTF-16 to ensure that there is a single standard character type, both
104% internally (within the grammar structures and hash tables) and externally (for the end user).
105% (2) all elements and attributes (both qualified and unqualified) are associated with a unique element
106% declaration or attribute definition within a specific grammar structure. Xerces emits the appropriate
107% grammar reference in place of the element or attribute string.
108
109
110
111
112
113%   From Xerces to icXML
114%
115%   - Philosophy:  Maximizing Bit Stream Processing
116%
117%   - Character Set Adapters vs. Transcoding
118%   - Bitstreams 1: Charset Validation and Transcoding equations
119%   - Bitstreams 2: Parabix style parsing and validation
120%
121%   - Bitstreams 3: Parallel filtering and normalization
122%           - LB normalization
123%           - reference compression -> single code unit speculation
124%           - parallel string termination
125%
126%   - Bitstreams 4: Symbol processing
127%
128%   - From bit streams to doublebyte streams: the content buffer
129%     
130%   - Namespace Processing: A Bitset approach.
Note: See TracBrowser for help on using the repository browser.