source: docs/Working/icGrep/fig-compiler.tex @ 4461

Last change on this file since 4461 was 4461, checked in by lindanl, 4 years ago

Add description for exection diagram.

File size: 10.3 KB
Line 
1\def\RegularExpression{RegEx}
2\def\Pablo{Pablo}
3\def\CodeUnit{Code Unit}
4\def\REParser{\RegularExpression{} Parser}
5\def\CodeUnitCompiler{\CodeUnit{} Compiler}
6\def\RegularExpressionCompiler{\RegularExpression{} Compiler}
7\def\PabloCompiler{\Pablo{} Compiler}
8
9\begin{figure}[h] \label{fig:compiler}
10\begin{center}
11% Define block styles
12%\tikzstyle{decision} = [diamond, shape aspect=2, rotate=30, draw, text width=4.5em, text badly centered, inner sep=0pt, thick]
13\tikzstyle{block} = [rectangle, draw, text width=15em, text centered, minimum height=1.75em, thick, font=\ttfamily\bfseries, node distance=3.5em]
14\tikzstyle{line} = [draw, ->, line width=1.4pt]
15\tikzstyle{seperator} = [draw, line width=0.125em, dashed]
16\tikzset{block/.append style={execute at begin node=\footnotesize}}   
17\begin{tikzpicture}[node distance=3cm, auto, >=stealth]
18
19    % Place nodes
20    \node [draw=none] (RE) {\RegularExpression{}};
21    \node [block, right of=RE, node distance=12em] (PropertyExtraction) {Property Extraction};
22    \node [block, below of=RE] (REParser) {\REParser{}};
23    \node [block, below of=REParser] (RETransform) {\RegularExpression{} Transformations};   
24    \coordinate[below of=RETransform, node distance=3em] (Point);   
25    \node [block, left of=Point, node distance=10em] (CUCompiler) {\CodeUnitCompiler{}};
26    \node [block, right of=Point, node distance=10em] (RECompiler) {\RegularExpressionCompiler{}};   
27    \node [block, below of=Point, node distance=3em] (PabloTransform) {\Pablo{} Transformations};   
28    \node [block, below of=PabloTransform] (PabloCompiler) {\PabloCompiler{}};
29    \node [block, below of=PabloCompiler] (LLVMCompiler) {LLVM Compiler};
30    \node [draw=none, below of=LLVMCompiler, node distance=3.5em] (Matcher) {Dynamically-Generated Match Function};
31   
32    % Draw edges
33    \path [line] (RE) -- (REParser);
34    \path [line] (RE) -- (PropertyExtraction);
35    \path [line] (REParser) -- (RETransform);
36    \path [line] (RETransform) -| (CUCompiler);
37    \path [line] (RETransform) -| (RECompiler);
38    \path [line] (CUCompiler) |- (PabloTransform);
39    \path [line] (RECompiler) |- (PabloTransform);
40    \path [line] (PabloTransform) -- (PabloCompiler);
41    \path [line] (PabloCompiler) -- (LLVMCompiler);
42    \path [line] (LLVMCompiler) -- (Matcher);
43   
44    % Draw seperators
45    \coordinate[right of=REParser, node distance=20em] (SR);
46    \coordinate[left of=REParser, node distance=20em] (SL);
47    \path [seperator] (SL) -- (REParser);
48    \path [seperator] (REParser) -- (SR);
49   
50    \coordinate[left of=Point, node distance=20em] (PL);
51    \coordinate[right of=Point, node distance=20em] (PR);
52    \path [seperator] (PL) -- (CUCompiler);
53    \path [seperator] (CUCompiler) -- (RECompiler);
54    \path [seperator] (RECompiler) -- (PR);
55
56    \coordinate[right of=PabloCompiler, node distance=20em] (LR);
57    \coordinate[left of=PabloCompiler, node distance=20em] (LL);
58    \path [seperator] (LL) -- (PabloCompiler);
59    \path [seperator] (PabloCompiler) -- (LR);   
60   
61    \coordinate[right of=LLVMCompiler, node distance=20em] (OR);
62    \coordinate[left of=LLVMCompiler, node distance=20em] (OL);
63    \path [seperator] (OL) -- (LLVMCompiler);
64    \path [seperator] (LLVMCompiler) -- (OR);       
65   
66    % Seperator text
67    \node [draw=none,anchor=west] at ($(SL)!0.5!(PL)$) {1)~\RegularExpression{}};
68    \node [draw=none,anchor=west] at ($(PL)!0.5!(LL)$) {2)~\Pablo{}};
69    \node [draw=none,anchor=west] at ($(LL)!0.5!(OL)$) {3)~LLVM};
70   
71   
72\end{tikzpicture}
73
74\end{center}
75\caption{icGrep Architectural Diagram.}
76\end{figure} 
77
78As show in Figure \ref{fig:compiler},
79icGrep is composed of three logical layers: \RegularExpression{}, \Pablo{} and the LLVM layer, each with their own intermediate representation
80(IR), transformation and compilation modules.
81%
82As we traverse the layers, the IR becomes significantly more complex as it begins to mirror the final machine code.
83%
84The \REParser{} validates and transforms the input \RegularExpression{} into an abstract syntax tree (AST).
85%
86The AST is a minimalistic representation that, unlike traditional \RegularExpression{}, is not converted into a NFA or DFA for further processing.
87%
88Instead, icGrep passes the AST into the transformation module, which includes a set of \RegularExpression{} specific optimization passes.
89%
90The initial \emph{Nullable} pass, determines whether the \RegularExpression{} contains any prefixes or suffixes that may be removed or
91modified whilst still providing the same number of matches as the original expression.
92%
93For example, ``\verb|a*bc+|'' is equivalent to ``\verb|bc|'' because the Kleene Star (Plus) operator matches zero (one) or more instances of a
94specific character.
95%
96The \emph{toUTF8} pass converts the characters in the input \RegularExpression{} into the equivalent expression(s) that represent the sequences
97of 8-bit code units necessary to identify the presence of a particular character.
98%
99Since some characters have multiple logically equivalent representations, such as \textcolor{red}{\textbf{????}}, this may produce nested sequences or alternations.
100%
101This is described in more detail in \S\ref{sec:Unicode:toUTF8}.
102%
103To alleviate this, the final \emph{Simplification} pass flattens nested sequences and alternations into their simplest legal form.
104%
105For example, ``\verb`a(b((c|d)|e))`'' would become ``\verb`ab(c|d|e)`'' and ``\verb`([0-9]{3,5}){3,5}`'', ``\verb`[0-9]{9,25}`''.
106%
107
108%% DISCUSS ANALYSIS MODULE?
109
110
111The \RegularExpression{} layer has two compilers: the \CodeUnit{} and \RegularExpressionCompiler{}, both of which produce \Pablo{} IR.
112%
113Recall that the \Pablo{} layer assumes a transposed view of the input data.
114%
115The \emph{\CodeUnitCompiler{}} transforms the input code unit classes, either extracted from the \RegularExpression{} or produced by the
116\emph{toUTF8} transformation, into a series of bit stream equations.
117%
118The \emph{\RegularExpressionCompiler{}} assumes that these have been calculated and transforms the \RegularExpression{} AST into
119a sequence of instructions.
120%
121For instance, it would convert any alternations into a sequence of calculations that are merged with \verb|OR|s.
122%
123The results of these passes are combined and transformed through a series of typical optimization passes, including dead code elimination
124(DCE), common subexpression elimination (CSE), and constant folding.
125%
126These are necessary at this stage because the \RegularExpression{} AST may include common subsequences that are costly to recognize in
127that form.
128%
129Similarly, to keep the \CodeUnitCompiler{} a linear time function, it may introduce redundant IR instructions as it applies traditional Boolean
130algebra transformations, such as de Morgan's law, to the computed streams.
131%
132An intended side-effect of these passes is that they eliminate the need to analyze the data-dependencies inherent in the carry-bit logic,
133which is necessary for some \Pablo{} instructions but problematic for optimizers to reason about non-conservatively.
134%
135The \PabloCompiler{} then converts the \Pablo{} IR into LLVM IR.
136%
137This is a relatively straightforward conversion:
138%
139the only complexities it introduces is the generation of Phi nodes, linking of statically-compiled functions, and assignment of carry variables.
140%
141It produces the dynamically-generated match function used by the icGrep.
142
143\begin{figure}[h] \label{fig:execution}
144\begin{center}
145\tikzstyle{block} = [rectangle, draw, text width=15em, text centered, minimum height=1.75em, thick, font=\ttfamily\bfseries, node distance=3.5em]
146\tikzstyle{line} = [draw, ->, line width=1.4pt]
147\tikzstyle{seperator} = [draw, line width=0.125em, dashed]
148\tikzset{block/.append style={execute at begin node=\footnotesize}}   
149\begin{tikzpicture}[node distance=3cm, auto, >=stealth]
150
151    % Place nodes
152    \node [draw=none] (InputData) {Input Data};
153    \node [block, below of=InputData] (S2P) {S2P};
154    \node [block, below of=S2P] (RequiredStreamsGenerator) {Required Streams Generator};
155    \node [block, below of=RequiredStreamsGenerator] (JITFunction) {JIT Function};
156    \node [block, right of=JITFunction, node distance=20em] (NamedPropertyLibaray) {Named Property Library};
157    \node [block, below of=JITFunction] (MatchScanner) {Match Scanner};
158    \node [draw=none, below of=MatchScanner, node distance=3.5em] (OutputResult) {Output Result};
159   
160    % Draw edges
161    \path [line] (InputData) -- (S2P);
162    \path [line] (S2P) -- (RequiredStreamsGenerator);
163    \path [line] (RequiredStreamsGenerator) -- (JITFunction);
164    \path [line] (NamedPropertyLibaray) -- (JITFunction);
165    \path [line] (JITFunction) -- (MatchScanner);
166    \path [line] (MatchScanner) -- (OutputResult);
167   
168\end{tikzpicture}
169
170\end{center}
171\caption{icGrep Execution Diagram.}
172\end{figure} 
173
174As shown in Figure \ref{fig:execution}, icGrep takes the input data and transposed it into 8 parallel bit streams through S2P module.
175The required streams, e.g. line break stream, can then be generated using the 8 basis bits streams.
176The JIT function retrieves the 8 basis bits and the required streams from their memory addresses and starts the matching process.
177Named Property Library that includes all the predefined Unicode categories is installed into JIT function and can be called during the matching process.
178JIT function returns one bitstream that marks all the matching positions.
179A match scanner will scan through this bitstream and calculate the total counts or write the context of each match position.
180
181We can also apply a pipeline parallelism strategy to further speed up the process of icGrep.
182S2P and Required Streams Generator can be process in a separate thread and start even before the dynamic compilation starts.
183The output of S2P and Required Streams Generator, that is the 8 basis bits streams and the required streams,
184needs to be stored in a shared memory space so that the JIT function can read from it.
185To be more efficient of memory space usage, we only allocate limit amount of space for the shared data.
186When each chunk of the shared space is filled up with the bitstream data,
187the thread will start writing to the first chunk if it is released by JIT function.
188Otherwise, it will wait for JIT function until it finishes processing that chunk.
189Therefore, the performance is depended on the slowest thread.
190In the case that the cost of transposition and required stream generation is more than the matching process,
191we can further divide up the work and assign two threads for S2P and Required Streams Generator.
192
193
194
195
196
Note: See TracBrowser for help on using the repository browser.