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

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

Parent Directory Parent Directory | Revision Log Revision Log


Revision 836 - (show annotations) (download) (as text)
Wed Dec 28 17:16:11 2011 UTC (2 years, 3 months ago) by ph10
File MIME type: text/html
File size: 15966 byte(s)
Merging all the changes from the pcre16 branch into the trunk.

1 <html>
2 <head>
3 <title>pcrejit specification</title>
4 </head>
5 <body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB">
6 <h1>pcrejit man page</h1>
7 <p>
8 Return to the <a href="index.html">PCRE index page</a>.
9 </p>
10 <p>
11 This page is part of the PCRE HTML documentation. It was generated automatically
12 from the original man page. If there is any nonsense in it, please consult the
13 man page, in case the conversion went wrong.
14 <br>
15 <ul>
16 <li><a name="TOC1" href="#SEC1">PCRE JUST-IN-TIME COMPILER SUPPORT</a>
17 <li><a name="TOC2" href="#SEC2">AVAILABILITY OF JIT SUPPORT</a>
18 <li><a name="TOC3" href="#SEC3">SIMPLE USE OF JIT</a>
19 <li><a name="TOC4" href="#SEC4">UNSUPPORTED OPTIONS AND PATTERN ITEMS</a>
20 <li><a name="TOC5" href="#SEC5">RETURN VALUES FROM JIT EXECUTION</a>
21 <li><a name="TOC6" href="#SEC6">SAVING AND RESTORING COMPILED PATTERNS</a>
22 <li><a name="TOC7" href="#SEC7">CONTROLLING THE JIT STACK</a>
23 <li><a name="TOC8" href="#SEC8">JIT STACK FAQ</a>
24 <li><a name="TOC9" href="#SEC9">EXAMPLE CODE</a>
25 <li><a name="TOC10" href="#SEC10">SEE ALSO</a>
26 <li><a name="TOC11" href="#SEC11">AUTHOR</a>
27 <li><a name="TOC12" href="#SEC12">REVISION</a>
28 </ul>
29 <br><a name="SEC1" href="#TOC1">PCRE JUST-IN-TIME COMPILER SUPPORT</a><br>
30 <P>
31 Just-in-time compiling is a heavyweight optimization that can greatly speed up
32 pattern matching. However, it comes at the cost of extra processing before the
33 match is performed. Therefore, it is of most benefit when the same pattern is
34 going to be matched many times. This does not necessarily mean many calls of
35 \fPpcre_exec()\fP; if the pattern is not anchored, matching attempts may take
36 place many times at various positions in the subject, even for a single call to
37 <b>pcre_exec()</b>. If the subject string is very long, it may still pay to use
38 JIT for one-off matches.
39 </P>
40 <P>
41 JIT support applies only to the traditional matching function,
42 <b>pcre_exec()</b>. It does not apply when <b>pcre_dfa_exec()</b> is being used.
43 The code for this support was written by Zoltan Herczeg.
44 </P>
45 <br><a name="SEC2" href="#TOC1">AVAILABILITY OF JIT SUPPORT</a><br>
46 <P>
47 JIT support is an optional feature of PCRE. The "configure" option --enable-jit
48 (or equivalent CMake option) must be set when PCRE is built if you want to use
49 JIT. The support is limited to the following hardware platforms:
50 <pre>
51 ARM v5, v7, and Thumb2
52 Intel x86 32-bit and 64-bit
53 MIPS 32-bit
54 Power PC 32-bit and 64-bit (experimental)
55 </pre>
56 The Power PC support is designated as experimental because it has not been
57 fully tested. If --enable-jit is set on an unsupported platform, compilation
58 fails.
59 </P>
60 <P>
61 A program that is linked with PCRE 8.20 or later can tell if JIT support is
62 available by calling <b>pcre_config()</b> with the PCRE_CONFIG_JIT option. The
63 result is 1 when JIT is available, and 0 otherwise. However, a simple program
64 does not need to check this in order to use JIT. The API is implemented in a
65 way that falls back to the ordinary PCRE code if JIT is not available.
66 </P>
67 <P>
68 If your program may sometimes be linked with versions of PCRE that are older
69 than 8.20, but you want to use JIT when it is available, you can test
70 the values of PCRE_MAJOR and PCRE_MINOR, or the existence of a JIT macro such
71 as PCRE_CONFIG_JIT, for compile-time control of your code.
72 </P>
73 <br><a name="SEC3" href="#TOC1">SIMPLE USE OF JIT</a><br>
74 <P>
75 You have to do two things to make use of the JIT support in the simplest way:
76 <pre>
77 (1) Call <b>pcre_study()</b> with the PCRE_STUDY_JIT_COMPILE option for
78 each compiled pattern, and pass the resulting <b>pcre_extra</b> block to
79 <b>pcre_exec()</b>.
80
81 (2) Use <b>pcre_free_study()</b> to free the <b>pcre_extra</b> block when it is
82 no longer needed instead of just freeing it yourself. This
83 ensures that any JIT data is also freed.
84 </pre>
85 For a program that may be linked with pre-8.20 versions of PCRE, you can insert
86 <pre>
87 #ifndef PCRE_STUDY_JIT_COMPILE
88 #define PCRE_STUDY_JIT_COMPILE 0
89 #endif
90 </pre>
91 so that no option is passed to <b>pcre_study()</b>, and then use something like
92 this to free the study data:
93 <pre>
94 #ifdef PCRE_CONFIG_JIT
95 pcre_free_study(study_ptr);
96 #else
97 pcre_free(study_ptr);
98 #endif
99 </pre>
100 In some circumstances you may need to call additional functions. These are
101 described in the section entitled
102 <a href="#stackcontrol">"Controlling the JIT stack"</a>
103 below.
104 </P>
105 <P>
106 If JIT support is not available, PCRE_STUDY_JIT_COMPILE is ignored, and no JIT
107 data is set up. Otherwise, the compiled pattern is passed to the JIT compiler,
108 which turns it into machine code that executes much faster than the normal
109 interpretive code. When <b>pcre_exec()</b> is passed a <b>pcre_extra</b> block
110 containing a pointer to JIT code, it obeys that instead of the normal code. The
111 result is identical, but the code runs much faster.
112 </P>
113 <P>
114 There are some <b>pcre_exec()</b> options that are not supported for JIT
115 execution. There are also some pattern items that JIT cannot handle. Details
116 are given below. In both cases, execution automatically falls back to the
117 interpretive code.
118 </P>
119 <P>
120 If the JIT compiler finds an unsupported item, no JIT data is generated. You
121 can find out if JIT execution is available after studying a pattern by calling
122 <b>pcre_fullinfo()</b> with the PCRE_INFO_JIT option. A result of 1 means that
123 JIT compilation was successful. A result of 0 means that JIT support is not
124 available, or the pattern was not studied with PCRE_STUDY_JIT_COMPILE, or the
125 JIT compiler was not able to handle the pattern.
126 </P>
127 <P>
128 Once a pattern has been studied, with or without JIT, it can be used as many
129 times as you like for matching different subject strings.
130 </P>
131 <br><a name="SEC4" href="#TOC1">UNSUPPORTED OPTIONS AND PATTERN ITEMS</a><br>
132 <P>
133 The only <b>pcre_exec()</b> options that are supported for JIT execution are
134 PCRE_NO_UTF8_CHECK, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, and
135 PCRE_NOTEMPTY_ATSTART. Note in particular that partial matching is not
136 supported.
137 </P>
138 <P>
139 The unsupported pattern items are:
140 <pre>
141 \C match a single byte; not supported in UTF-8 mode
142 (?Cn) callouts
143 (*COMMIT) )
144 (*MARK) )
145 (*PRUNE) ) the backtracking control verbs
146 (*SKIP) )
147 (*THEN) )
148 </pre>
149 Support for some of these may be added in future.
150 </P>
151 <br><a name="SEC5" href="#TOC1">RETURN VALUES FROM JIT EXECUTION</a><br>
152 <P>
153 When a pattern is matched using JIT execution, the return values are the same
154 as those given by the interpretive <b>pcre_exec()</b> code, with the addition of
155 one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means that the memory used
156 for the JIT stack was insufficient. See
157 <a href="#stackcontrol">"Controlling the JIT stack"</a>
158 below for a discussion of JIT stack usage. For compatibility with the
159 interpretive <b>pcre_exec()</b> code, no more than two-thirds of the
160 <i>ovector</i> argument is used for passing back captured substrings.
161 </P>
162 <P>
163 The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if searching a
164 very large pattern tree goes on for too long, as it is in the same circumstance
165 when JIT is not used, but the details of exactly what is counted are not the
166 same. The PCRE_ERROR_RECURSIONLIMIT error code is never returned by JIT
167 execution.
168 </P>
169 <br><a name="SEC6" href="#TOC1">SAVING AND RESTORING COMPILED PATTERNS</a><br>
170 <P>
171 The code that is generated by the JIT compiler is architecture-specific, and is
172 also position dependent. For those reasons it cannot be saved (in a file or
173 database) and restored later like the bytecode and other data of a compiled
174 pattern. Saving and restoring compiled patterns is not something many people
175 do. More detail about this facility is given in the
176 <a href="pcreprecompile.html"><b>pcreprecompile</b></a>
177 documentation. It should be possible to run <b>pcre_study()</b> on a saved and
178 restored pattern, and thereby recreate the JIT data, but because JIT
179 compilation uses significant resources, it is probably not worth doing this;
180 you might as well recompile the original pattern.
181 <a name="stackcontrol"></a></P>
182 <br><a name="SEC7" href="#TOC1">CONTROLLING THE JIT STACK</a><br>
183 <P>
184 When the compiled JIT code runs, it needs a block of memory to use as a stack.
185 By default, it uses 32K on the machine stack. However, some large or
186 complicated patterns need more than this. The error PCRE_ERROR_JIT_STACKLIMIT
187 is given when there is not enough stack. Three functions are provided for
188 managing blocks of memory for use as JIT stacks. There is further discussion
189 about the use of JIT stacks in the section entitled
190 <a href="#stackcontrol">"JIT stack FAQ"</a>
191 below.
192 </P>
193 <P>
194 The <b>pcre_jit_stack_alloc()</b> function creates a JIT stack. Its arguments
195 are a starting size and a maximum size, and it returns a pointer to an opaque
196 structure of type <b>pcre_jit_stack</b>, or NULL if there is an error. The
197 <b>pcre_jit_stack_free()</b> function can be used to free a stack that is no
198 longer needed. (For the technically minded: the address space is allocated by
199 mmap or VirtualAlloc.)
200 </P>
201 <P>
202 JIT uses far less memory for recursion than the interpretive code,
203 and a maximum stack size of 512K to 1M should be more than enough for any
204 pattern.
205 </P>
206 <P>
207 The <b>pcre_assign_jit_stack()</b> function specifies which stack JIT code
208 should use. Its arguments are as follows:
209 <pre>
210 pcre_extra *extra
211 pcre_jit_callback callback
212 void *data
213 </pre>
214 The <i>extra</i> argument must be the result of studying a pattern with
215 PCRE_STUDY_JIT_COMPILE. There are three cases for the values of the other two
216 options:
217 <pre>
218 (1) If <i>callback</i> is NULL and <i>data</i> is NULL, an internal 32K block
219 on the machine stack is used.
220
221 (2) If <i>callback</i> is NULL and <i>data</i> is not NULL, <i>data</i> must be
222 a valid JIT stack, the result of calling <b>pcre_jit_stack_alloc()</b>.
223
224 (3) If <i>callback</i> not NULL, it must point to a function that is called
225 with <i>data</i> as an argument at the start of matching, in order to
226 set up a JIT stack. If the result is NULL, the internal 32K stack
227 is used; otherwise the return value must be a valid JIT stack,
228 the result of calling <b>pcre_jit_stack_alloc()</b>.
229 </pre>
230 You may safely assign the same JIT stack to more than one pattern, as long as
231 they are all matched sequentially in the same thread. In a multithread
232 application, each thread must use its own JIT stack.
233 </P>
234 <P>
235 Strictly speaking, even more is allowed. You can assign the same stack to any
236 number of patterns as long as they are not used for matching by multiple
237 threads at the same time. For example, you can assign the same stack to all
238 compiled patterns, and use a global mutex in the callback to wait until the
239 stack is available for use. However, this is an inefficient solution, and
240 not recommended.
241 </P>
242 <P>
243 This is a suggestion for how a typical multithreaded program might operate:
244 <pre>
245 During thread initalization
246 thread_local_var = pcre_jit_stack_alloc(...)
247
248 During thread exit
249 pcre_jit_stack_free(thread_local_var)
250
251 Use a one-line callback function
252 return thread_local_var
253 </pre>
254 All the functions described in this section do nothing if JIT is not available,
255 and <b>pcre_assign_jit_stack()</b> does nothing unless the <b>extra</b> argument
256 is non-NULL and points to a <b>pcre_extra</b> block that is the result of a
257 successful study with PCRE_STUDY_JIT_COMPILE.
258 <a name="stackfaq"></a></P>
259 <br><a name="SEC8" href="#TOC1">JIT STACK FAQ</a><br>
260 <P>
261 (1) Why do we need JIT stacks?
262 <br>
263 <br>
264 PCRE (and JIT) is a recursive, depth-first engine, so it needs a stack where
265 the local data of the current node is pushed before checking its child nodes.
266 Allocating real machine stack on some platforms is difficult. For example, the
267 stack chain needs to be updated every time if we extend the stack on PowerPC.
268 Although it is possible, its updating time overhead decreases performance. So
269 we do the recursion in memory.
270 </P>
271 <P>
272 (2) Why don't we simply allocate blocks of memory with <b>malloc()</b>?
273 <br>
274 <br>
275 Modern operating systems have a nice feature: they can reserve an address space
276 instead of allocating memory. We can safely allocate memory pages inside this
277 address space, so the stack could grow without moving memory data (this is
278 important because of pointers). Thus we can allocate 1M address space, and use
279 only a single memory page (usually 4K) if that is enough. However, we can still
280 grow up to 1M anytime if needed.
281 </P>
282 <P>
283 (3) Who "owns" a JIT stack?
284 <br>
285 <br>
286 The owner of the stack is the user program, not the JIT studied pattern or
287 anything else. The user program must ensure that if a stack is used by
288 <b>pcre_exec()</b>, (that is, it is assigned to the pattern currently running),
289 that stack must not be used by any other threads (to avoid overwriting the same
290 memory area). The best practice for multithreaded programs is to allocate a
291 stack for each thread, and return this stack through the JIT callback function.
292 </P>
293 <P>
294 (4) When should a JIT stack be freed?
295 <br>
296 <br>
297 You can free a JIT stack at any time, as long as it will not be used by
298 <b>pcre_exec()</b> again. When you assign the stack to a pattern, only a pointer
299 is set. There is no reference counting or any other magic. You can free the
300 patterns and stacks in any order, anytime. Just <i>do not</i> call
301 <b>pcre_exec()</b> with a pattern pointing to an already freed stack, as that
302 will cause SEGFAULT. (Also, do not free a stack currently used by
303 <b>pcre_exec()</b> in another thread). You can also replace the stack for a
304 pattern at any time. You can even free the previous stack before assigning a
305 replacement.
306 </P>
307 <P>
308 (5) Should I allocate/free a stack every time before/after calling
309 <b>pcre_exec()</b>?
310 <br>
311 <br>
312 No, because this is too costly in terms of resources. However, you could
313 implement some clever idea which release the stack if it is not used in let's
314 say two minutes. The JIT callback can help to achive this without keeping a
315 list of the currently JIT studied patterns.
316 </P>
317 <P>
318 (6) OK, the stack is for long term memory allocation. But what happens if a
319 pattern causes stack overflow with a stack of 1M? Is that 1M kept until the
320 stack is freed?
321 <br>
322 <br>
323 Especially on embedded sytems, it might be a good idea to release
324 memory sometimes without freeing the stack. There is no API for this at the
325 moment. Probably a function call which returns with the currently allocated
326 memory for any stack and another which allows releasing memory (shrinking the
327 stack) would be a good idea if someone needs this.
328 </P>
329 <P>
330 (7) This is too much of a headache. Isn't there any better solution for JIT
331 stack handling?
332 <br>
333 <br>
334 No, thanks to Windows. If POSIX threads were used everywhere, we could throw
335 out this complicated API.
336 </P>
337 <br><a name="SEC9" href="#TOC1">EXAMPLE CODE</a><br>
338 <P>
339 This is a single-threaded example that specifies a JIT stack without using a
340 callback.
341 <pre>
342 int rc;
343 int ovector[30];
344 pcre *re;
345 pcre_extra *extra;
346 pcre_jit_stack *jit_stack;
347
348 re = pcre_compile(pattern, 0, &error, &erroffset, NULL);
349 /* Check for errors */
350 extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error);
351 jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024);
352 /* Check for error (NULL) */
353 pcre_assign_jit_stack(extra, NULL, jit_stack);
354 rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30);
355 /* Check results */
356 pcre_free(re);
357 pcre_free_study(extra);
358 pcre_jit_stack_free(jit_stack);
359
360 </PRE>
361 </P>
362 <br><a name="SEC10" href="#TOC1">SEE ALSO</a><br>
363 <P>
364 <b>pcreapi</b>(3)
365 </P>
366 <br><a name="SEC11" href="#TOC1">AUTHOR</a><br>
367 <P>
368 Philip Hazel (FAQ by Zoltan Herczeg)
369 <br>
370 University Computing Service
371 <br>
372 Cambridge CB2 3QH, England.
373 <br>
374 </P>
375 <br><a name="SEC12" href="#TOC1">REVISION</a><br>
376 <P>
377 Last updated: 26 November 2011
378 <br>
379 Copyright &copy; 1997-2011 University of Cambridge.
380 <br>
381 <p>
382 Return to the <a href="index.html">PCRE index page</a>.
383 </p>

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12