/[pcre]/code/trunk/doc/pcrejit.3
ViewVC logotype

Contents of /code/trunk/doc/pcrejit.3

Parent Directory Parent Directory | Revision Log Revision Log


Revision 693 - (hide annotations) (download)
Mon Sep 12 15:59:57 2011 UTC (3 years, 2 months ago) by ph10
File size: 9826 byte(s)
Fix typo.

1 ph10 678 .TH PCREJIT 3
2     .SH NAME
3     PCRE - Perl-compatible regular expressions
4     .SH "PCRE JUST-IN-TIME COMPILER SUPPORT"
5     .rs
6     .sp
7     Just-in-time compiling is a heavyweight optimization that can greatly speed up
8 ph10 683 pattern matching. However, it comes at the cost of extra processing before the
9     match is performed. Therefore, it is of most benefit when the same pattern is
10     going to be matched many times. This does not necessarily mean many calls of
11 ph10 678 \fPpcre_exec()\fP; if the pattern is not anchored, matching attempts may take
12     place many times at various positions in the subject, even for a single call to
13 ph10 683 \fBpcre_exec()\fP. If the subject string is very long, it may still pay to use
14 ph10 678 JIT for one-off matches.
15     .P
16 ph10 683 JIT support applies only to the traditional matching function,
17 ph10 678 \fBpcre_exec()\fP. It does not apply when \fBpcre_dfa_exec()\fP is being used.
18     The code for this support was written by Zoltan Herczeg.
19     .
20     .
21     .SH "AVAILABILITY OF JIT SUPPORT"
22     .rs
23     .sp
24     JIT support is an optional feature of PCRE. The "configure" option --enable-jit
25     (or equivalent CMake option) must be set when PCRE is built if you want to use
26     JIT. The support is limited to the following hardware platforms:
27     .sp
28     ARM v5, v7, and Thumb2
29 ph10 683 Intel x86 32-bit and 64-bit
30 ph10 678 MIPS 32-bit
31     Power PC 32-bit and 64-bit
32 ph10 683 .sp
33 ph10 678 If --enable-jit is set on an unsupported platform, compilation fails.
34     .P
35 ph10 683 A program can tell if JIT support is available by calling \fBpcre_config()\fP
36     with the PCRE_CONFIG_JIT option. The result is 1 when JIT is available, and 0
37 ph10 678 otherwise. However, a simple program does not need to check this in order to
38     use JIT. The API is implemented in a way that falls back to the ordinary PCRE
39     code if JIT is not available.
40     .
41     .
42     .SH "SIMPLE USE OF JIT"
43     .rs
44     .sp
45     You have to do two things to make use of the JIT support in the simplest way:
46     .sp
47     (1) Call \fBpcre_study()\fP with the PCRE_STUDY_JIT_COMPILE option for
48     each compiled pattern, and pass the resulting \fBpcre_extra\fP block to
49     \fBpcre_exec()\fP.
50 ph10 691 .sp
51 ph10 678 (2) Use \fBpcre_free_study()\fP to free the \fBpcre_extra\fP block when it is
52 ph10 691 no longer needed instead of just freeing it yourself. This
53 ph10 683 ensures that any JIT data is also freed.
54 ph10 678 .sp
55 ph10 683 In some circumstances you may need to call additional functions. These are
56 ph10 678 described in the section entitled
57     .\" HTML <a href="#stackcontrol">
58     .\" </a>
59     "Controlling the JIT stack"
60     .\"
61     below.
62     .P
63 ph10 683 If JIT support is not available, PCRE_STUDY_JIT_COMPILE is ignored, and no JIT
64     data is set up. Otherwise, the compiled pattern is passed to the JIT compiler,
65     which turns it into machine code that executes much faster than the normal
66     interpretive code. When \fBpcre_exec()\fP is passed a \fBpcre_extra\fP block
67     containing a pointer to JIT code, it obeys that instead of the normal code. The
68     result is identical, but the code runs much faster.
69 ph10 678 .P
70     There are some \fBpcre_exec()\fP options that are not supported for JIT
71 ph10 683 execution. There are also some pattern items that JIT cannot handle. Details
72     are given below. In both cases, execution automatically falls back to the
73 ph10 678 interpretive code.
74     .P
75     If the JIT compiler finds an unsupported item, no JIT data is generated. You
76     can find out if JIT execution is available after studying a pattern by calling
77     \fBpcre_fullinfo()\fP with the PCRE_INFO_JIT option. A result of 1 means that
78 ph10 693 JIT compilation was successful. A result of 0 means that JIT support is not
79 ph10 678 available, or the pattern was not studied with PCRE_STUDY_JIT_COMPILE, or the
80     JIT compiler was not able to handle the pattern.
81     .
82     .
83     .SH "UNSUPPORTED OPTIONS AND PATTERN ITEMS"
84     .rs
85     .sp
86     The only \fBpcre_exec()\fP options that are supported for JIT execution are
87 ph10 683 PCRE_NO_UTF8_CHECK, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, and
88     PCRE_NOTEMPTY_ATSTART. Note in particular that partial matching is not
89 ph10 678 supported.
90     .P
91     The unsupported pattern items are:
92     .sp
93     \eC match a single byte, even in UTF-8 mode
94     (?Cn) callouts
95     (?(<name>)... conditional test on setting of a named subpattern
96 ph10 683 (?(R)... conditional test on whole pattern recursion
97 ph10 678 (?(Rn)... conditional test on recursion, by number
98     (?(R&name)... conditional test on recursion, by name
99     (*COMMIT) )
100     (*MARK) )
101     (*PRUNE) ) the backtracking control verbs
102     (*SKIP) )
103     (*THEN) )
104 ph10 683 .sp
105 ph10 678 Support for some of these may be added in future.
106     .
107     .
108     .SH "RETURN VALUES FROM JIT EXECUTION"
109     .rs
110     .sp
111 ph10 683 When a pattern is matched using JIT execution, the return values are the same
112     as those given by the interpretive \fBpcre_exec()\fP code, with the addition of
113     one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means that the memory used
114 ph10 678 for the JIT stack was insufficient. See
115     .\" HTML <a href="#stackcontrol">
116     .\" </a>
117     "Controlling the JIT stack"
118     .\"
119 ph10 683 below for a discussion of JIT stack usage. For compatibility with the
120     interpretive \fBpcre_exec()\fP code, no more than two-thirds of the
121     \fIovector\fP argument is used for passing back captured substrings.
122 ph10 678 .P
123     The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if searching a
124     very large pattern tree goes on for too long, as it is in the same circumstance
125     when JIT is not used, but the details of exactly what is counted are not the
126     same. The PCRE_ERROR_RECURSIONLIMIT error code is never returned by JIT
127     execution.
128     .
129     .
130     .SH "SAVING AND RESTORING COMPILED PATTERNS"
131     .rs
132     .sp
133 ph10 683 The code that is generated by the JIT compiler is architecture-specific, and is
134     also position dependent. For those reasons it cannot be saved and restored like
135 ph10 678 the bytecode and other data of a compiled pattern. You should be able run
136     \fBpcre_study()\fP on a saved and restored pattern, and thereby recreate the
137     JIT data, but because JIT compilation uses significant resources, it is
138 ph10 683 probably not worth doing this.
139 ph10 678 .
140     .
141     .\" HTML <a name="stackcontrol"></a>
142     .SH "CONTROLLING THE JIT STACK"
143     .rs
144     .sp
145 ph10 683 When the compiled JIT code runs, it needs a block of memory to use as a stack.
146     By default, it uses 32K on the machine stack. However, some large or
147     complicated patterns need more than this. The error PCRE_ERROR_JIT_STACKLIMIT
148     is given when there is not enough stack. Three functions are provided for
149     managing blocks of memory for use as JIT stacks.
150 ph10 678 .P
151 ph10 683 The \fBpcre_jit_stack_alloc()\fP function creates a JIT stack. Its arguments
152 ph10 691 are a starting size and a maximum size, and it returns a pointer to an opaque
153 ph10 686 structure of type \fBpcre_jit_stack\fP, or NULL if there is an error. The
154     \fBpcre_jit_stack_free()\fP function can be used to free a stack that is no
155     longer needed. (For the technically minded: the address space is allocated by
156     mmap or VirtualAlloc.)
157 ph10 678 .P
158 ph10 691 JIT uses far less memory for recursion than the interpretive code,
159 ph10 683 and a maximum stack size of 512K to 1M should be more than enough for any
160     pattern.
161     .P
162     The \fBpcre_assign_jit_stack()\fP function specifies which stack JIT code
163 ph10 678 should use. Its arguments are as follows:
164     .sp
165     pcre_extra *extra
166     pcre_jit_callback callback
167     void *data
168 ph10 683 .sp
169     The \fIextra\fP argument must be the result of studying a pattern with
170     PCRE_STUDY_JIT_COMPILE. There are three cases for the values of the other two
171 ph10 678 options:
172     .sp
173     (1) If \fIcallback\fP is NULL and \fIdata\fP is NULL, an internal 32K block
174     on the machine stack is used.
175     .sp
176     (2) If \fIcallback\fP is NULL and \fIdata\fP is not NULL, \fIdata\fP must be
177     a valid JIT stack, the result of calling \fBpcre_jit_stack_alloc()\fP.
178     .sp
179     (3) If \fIcallback\fP not NULL, it must point to a function that is called
180 ph10 683 with \fIdata\fP as an argument at the start of matching, in order to
181     set up a JIT stack. If the result is NULL, the internal 32K stack
182     is used; otherwise the return value must be a valid JIT stack,
183 ph10 678 the result of calling \fBpcre_jit_stack_alloc()\fP.
184     .sp
185     You may safely assign the same JIT stack to more than one pattern, as long as
186     they are all matched sequentially in the same thread. In a multithread
187     application, each thread must use its own JIT stack.
188     .P
189 ph10 683 Strictly speaking, even more is allowed. You can assign the same stack to any
190     number of patterns as long as they are not used for matching by multiple
191     threads at the same time. For example, you can assign the same stack to all
192     compiled patterns, and use a global mutex in the callback to wait until the
193     stack is available for use. However, this is an inefficient solution, and
194     not recommended.
195     .P
196     This is a suggestion for how a typical multithreaded program might operate:
197     .sp
198     During thread initalization
199     thread_local_var = pcre_jit_stack_alloc(...)
200 ph10 691 .sp
201 ph10 683 During thread exit
202     pcre_jit_stack_free(thread_local_var)
203 ph10 691 .sp
204 ph10 683 Use a one-line callback function
205     return thread_local_var
206     .sp
207 ph10 678 All the functions described in this section do nothing if JIT is not available,
208 ph10 683 and \fBpcre_assign_jit_stack()\fP does nothing unless the \fBextra\fP argument
209     is non-NULL and points to a \fBpcre_extra\fP block that is the result of a
210 ph10 678 successful study with PCRE_STUDY_JIT_COMPILE.
211     .
212     .
213     .SH "EXAMPLE CODE"
214     .rs
215     .sp
216 ph10 683 This is a single-threaded example that specifies a JIT stack without using a
217 ph10 691 callback.
218 ph10 678 .sp
219     int rc;
220 ph10 683 int ovector[30];
221 ph10 678 pcre *re;
222 ph10 683 pcre_extra *extra;
223     pcre_jit_stack *jit_stack;
224     .sp
225 ph10 678 re = pcre_compile(pattern, 0, &error, &erroffset, NULL);
226     /* Check for errors */
227     extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error);
228 ph10 683 jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024);
229 ph10 678 /* Check for error (NULL) */
230     pcre_assign_jit_stack(extra, NULL, jit_stack);
231 ph10 683 rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30);
232 ph10 678 /* Check results */
233     pcre_free(re);
234 ph10 683 pcre_free_study(extra);
235 ph10 691 pcre_jit_stack_free(jit_stack);
236 ph10 678 .sp
237     .
238     .
239     .SH "SEE ALSO"
240     .rs
241     .sp
242     \fBpcreapi\fP(3)
243     .
244     .
245     .SH AUTHOR
246     .rs
247     .sp
248     .nf
249     Philip Hazel
250     University Computing Service
251     Cambridge CB2 3QH, England.
252     .fi
253     .
254     .
255     .SH REVISION
256     .rs
257     .sp
258     .nf
259 ph10 683 Last updated: 06 September 2011
260 ph10 678 Copyright (c) 1997-2011 University of Cambridge.
261     .fi

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12