/[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 975 - (show annotations) (download) (as text)
Sat Jun 2 11:03:06 2012 UTC (2 years, 3 months ago) by ph10
File MIME type: text/html
File size: 18233 byte(s)
Document update for 8.31-RC1 test release.

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">8-BIT and 16-BIT SUPPORT</a>
18 <li><a name="TOC3" href="#SEC3">AVAILABILITY OF JIT SUPPORT</a>
19 <li><a name="TOC4" href="#SEC4">SIMPLE USE OF JIT</a>
20 <li><a name="TOC5" href="#SEC5">UNSUPPORTED OPTIONS AND PATTERN ITEMS</a>
21 <li><a name="TOC6" href="#SEC6">RETURN VALUES FROM JIT EXECUTION</a>
22 <li><a name="TOC7" href="#SEC7">SAVING AND RESTORING COMPILED PATTERNS</a>
23 <li><a name="TOC8" href="#SEC8">CONTROLLING THE JIT STACK</a>
24 <li><a name="TOC9" href="#SEC9">JIT STACK FAQ</a>
25 <li><a name="TOC10" href="#SEC10">EXAMPLE CODE</a>
26 <li><a name="TOC11" href="#SEC11">SEE ALSO</a>
27 <li><a name="TOC12" href="#SEC12">AUTHOR</a>
28 <li><a name="TOC13" href="#SEC13">REVISION</a>
29 </ul>
30 <br><a name="SEC1" href="#TOC1">PCRE JUST-IN-TIME COMPILER SUPPORT</a><br>
31 <P>
32 Just-in-time compiling is a heavyweight optimization that can greatly speed up
33 pattern matching. However, it comes at the cost of extra processing before the
34 match is performed. Therefore, it is of most benefit when the same pattern is
35 going to be matched many times. This does not necessarily mean many calls of a
36 matching function; if the pattern is not anchored, matching attempts may take
37 place many times at various positions in the subject, even for a single call.
38 Therefore, if the subject string is very long, it may still pay to use JIT for
39 one-off matches.
40 </P>
41 <P>
42 JIT support applies only to the traditional Perl-compatible matching function.
43 It does not apply when the DFA matching function is being used. The code for
44 this support was written by Zoltan Herczeg.
45 </P>
46 <br><a name="SEC2" href="#TOC1">8-BIT and 16-BIT SUPPORT</a><br>
47 <P>
48 JIT support is available for both the 8-bit and 16-bit PCRE libraries. To keep
49 this documentation simple, only the 8-bit interface is described in what
50 follows. If you are using the 16-bit library, substitute the 16-bit functions
51 and 16-bit structures (for example, <i>pcre16_jit_stack</i> instead of
52 <i>pcre_jit_stack</i>).
53 </P>
54 <br><a name="SEC3" href="#TOC1">AVAILABILITY OF JIT SUPPORT</a><br>
55 <P>
56 JIT support is an optional feature of PCRE. The "configure" option --enable-jit
57 (or equivalent CMake option) must be set when PCRE is built if you want to use
58 JIT. The support is limited to the following hardware platforms:
59 <pre>
60 ARM v5, v7, and Thumb2
61 Intel x86 32-bit and 64-bit
62 MIPS 32-bit
63 Power PC 32-bit and 64-bit
64 </pre>
65 If --enable-jit is set on an unsupported platform, compilation fails.
66 </P>
67 <P>
68 A program that is linked with PCRE 8.20 or later can tell if JIT support is
69 available by calling <b>pcre_config()</b> with the PCRE_CONFIG_JIT option. The
70 result is 1 when JIT is available, and 0 otherwise. However, a simple program
71 does not need to check this in order to use JIT. The API is implemented in a
72 way that falls back to the interpretive code if JIT is not available.
73 </P>
74 <P>
75 If your program may sometimes be linked with versions of PCRE that are older
76 than 8.20, but you want to use JIT when it is available, you can test
77 the values of PCRE_MAJOR and PCRE_MINOR, or the existence of a JIT macro such
78 as PCRE_CONFIG_JIT, for compile-time control of your code.
79 </P>
80 <br><a name="SEC4" href="#TOC1">SIMPLE USE OF JIT</a><br>
81 <P>
82 You have to do two things to make use of the JIT support in the simplest way:
83 <pre>
84 (1) Call <b>pcre_study()</b> with the PCRE_STUDY_JIT_COMPILE option for
85 each compiled pattern, and pass the resulting <b>pcre_extra</b> block to
86 <b>pcre_exec()</b>.
87
88 (2) Use <b>pcre_free_study()</b> to free the <b>pcre_extra</b> block when it is
89 no longer needed, instead of just freeing it yourself. This
90 ensures that any JIT data is also freed.
91 </pre>
92 For a program that may be linked with pre-8.20 versions of PCRE, you can insert
93 <pre>
94 #ifndef PCRE_STUDY_JIT_COMPILE
95 #define PCRE_STUDY_JIT_COMPILE 0
96 #endif
97 </pre>
98 so that no option is passed to <b>pcre_study()</b>, and then use something like
99 this to free the study data:
100 <pre>
101 #ifdef PCRE_CONFIG_JIT
102 pcre_free_study(study_ptr);
103 #else
104 pcre_free(study_ptr);
105 #endif
106 </pre>
107 PCRE_STUDY_JIT_COMPILE requests the JIT compiler to generate code for complete
108 matches. If you want to run partial matches using the PCRE_PARTIAL_HARD or
109 PCRE_PARTIAL_SOFT options of <b>pcre_exec()</b>, you should set one or both of
110 the following options in addition to, or instead of, PCRE_STUDY_JIT_COMPILE
111 when you call <b>pcre_study()</b>:
112 <pre>
113 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
114 PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
115 </pre>
116 The JIT compiler generates different optimized code for each of the three
117 modes (normal, soft partial, hard partial). When <b>pcre_exec()</b> is called,
118 the appropriate code is run if it is available. Otherwise, the pattern is
119 matched using interpretive code.
120 </P>
121 <P>
122 In some circumstances you may need to call additional functions. These are
123 described in the section entitled
124 <a href="#stackcontrol">"Controlling the JIT stack"</a>
125 below.
126 </P>
127 <P>
128 If JIT support is not available, PCRE_STUDY_JIT_COMPILE etc. are ignored, and
129 no JIT data is created. Otherwise, the compiled pattern is passed to the JIT
130 compiler, which turns it into machine code that executes much faster than the
131 normal interpretive code. When <b>pcre_exec()</b> is passed a <b>pcre_extra</b>
132 block containing a pointer to JIT code of the appropriate mode (normal or
133 hard/soft partial), it obeys that code instead of running the interpreter. The
134 result is identical, but the compiled JIT code runs much faster.
135 </P>
136 <P>
137 There are some <b>pcre_exec()</b> options that are not supported for JIT
138 execution. There are also some pattern items that JIT cannot handle. Details
139 are given below. In both cases, execution automatically falls back to the
140 interpretive code. If you want to know whether JIT was actually used for a
141 particular match, you should arrange for a JIT callback function to be set up
142 as described in the section entitled
143 <a href="#stackcontrol">"Controlling the JIT stack"</a>
144 below, even if you do not need to supply a non-default JIT stack. Such a
145 callback function is called whenever JIT code is about to be obeyed. If the
146 execution options are not right for JIT execution, the callback function is not
147 obeyed.
148 </P>
149 <P>
150 If the JIT compiler finds an unsupported item, no JIT data is generated. You
151 can find out if JIT execution is available after studying a pattern by calling
152 <b>pcre_fullinfo()</b> with the PCRE_INFO_JIT option. A result of 1 means that
153 JIT compilation was successful. A result of 0 means that JIT support is not
154 available, or the pattern was not studied with PCRE_STUDY_JIT_COMPILE etc., or
155 the JIT compiler was not able to handle the pattern.
156 </P>
157 <P>
158 Once a pattern has been studied, with or without JIT, it can be used as many
159 times as you like for matching different subject strings.
160 </P>
161 <br><a name="SEC5" href="#TOC1">UNSUPPORTED OPTIONS AND PATTERN ITEMS</a><br>
162 <P>
163 The only <b>pcre_exec()</b> options that are supported for JIT execution are
164 PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK, PCRE_NOTBOL, PCRE_NOTEOL,
165 PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART, PCRE_PARTIAL_HARD, and PCRE_PARTIAL_SOFT.
166 </P>
167 <P>
168 The unsupported pattern items are:
169 <pre>
170 \C match a single byte; not supported in UTF-8 mode
171 (?Cn) callouts
172 (*PRUNE) )
173 (*SKIP) ) backtracking control verbs
174 (*THEN) )
175 </pre>
176 Support for some of these may be added in future.
177 </P>
178 <br><a name="SEC6" href="#TOC1">RETURN VALUES FROM JIT EXECUTION</a><br>
179 <P>
180 When a pattern is matched using JIT execution, the return values are the same
181 as those given by the interpretive <b>pcre_exec()</b> code, with the addition of
182 one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means that the memory used
183 for the JIT stack was insufficient. See
184 <a href="#stackcontrol">"Controlling the JIT stack"</a>
185 below for a discussion of JIT stack usage. For compatibility with the
186 interpretive <b>pcre_exec()</b> code, no more than two-thirds of the
187 <i>ovector</i> argument is used for passing back captured substrings.
188 </P>
189 <P>
190 The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if searching a
191 very large pattern tree goes on for too long, as it is in the same circumstance
192 when JIT is not used, but the details of exactly what is counted are not the
193 same. The PCRE_ERROR_RECURSIONLIMIT error code is never returned by JIT
194 execution.
195 </P>
196 <br><a name="SEC7" href="#TOC1">SAVING AND RESTORING COMPILED PATTERNS</a><br>
197 <P>
198 The code that is generated by the JIT compiler is architecture-specific, and is
199 also position dependent. For those reasons it cannot be saved (in a file or
200 database) and restored later like the bytecode and other data of a compiled
201 pattern. Saving and restoring compiled patterns is not something many people
202 do. More detail about this facility is given in the
203 <a href="pcreprecompile.html"><b>pcreprecompile</b></a>
204 documentation. It should be possible to run <b>pcre_study()</b> on a saved and
205 restored pattern, and thereby recreate the JIT data, but because JIT
206 compilation uses significant resources, it is probably not worth doing this;
207 you might as well recompile the original pattern.
208 <a name="stackcontrol"></a></P>
209 <br><a name="SEC8" href="#TOC1">CONTROLLING THE JIT STACK</a><br>
210 <P>
211 When the compiled JIT code runs, it needs a block of memory to use as a stack.
212 By default, it uses 32K on the machine stack. However, some large or
213 complicated patterns need more than this. The error PCRE_ERROR_JIT_STACKLIMIT
214 is given when there is not enough stack. Three functions are provided for
215 managing blocks of memory for use as JIT stacks. There is further discussion
216 about the use of JIT stacks in the section entitled
217 <a href="#stackcontrol">"JIT stack FAQ"</a>
218 below.
219 </P>
220 <P>
221 The <b>pcre_jit_stack_alloc()</b> function creates a JIT stack. Its arguments
222 are a starting size and a maximum size, and it returns a pointer to an opaque
223 structure of type <b>pcre_jit_stack</b>, or NULL if there is an error. The
224 <b>pcre_jit_stack_free()</b> function can be used to free a stack that is no
225 longer needed. (For the technically minded: the address space is allocated by
226 mmap or VirtualAlloc.)
227 </P>
228 <P>
229 JIT uses far less memory for recursion than the interpretive code,
230 and a maximum stack size of 512K to 1M should be more than enough for any
231 pattern.
232 </P>
233 <P>
234 The <b>pcre_assign_jit_stack()</b> function specifies which stack JIT code
235 should use. Its arguments are as follows:
236 <pre>
237 pcre_extra *extra
238 pcre_jit_callback callback
239 void *data
240 </pre>
241 The <i>extra</i> argument must be the result of studying a pattern with
242 PCRE_STUDY_JIT_COMPILE etc. There are three cases for the values of the other
243 two options:
244 <pre>
245 (1) If <i>callback</i> is NULL and <i>data</i> is NULL, an internal 32K block
246 on the machine stack is used.
247
248 (2) If <i>callback</i> is NULL and <i>data</i> is not NULL, <i>data</i> must be
249 a valid JIT stack, the result of calling <b>pcre_jit_stack_alloc()</b>.
250
251 (3) If <i>callback</i> is not NULL, it must point to a function that is
252 called with <i>data</i> as an argument at the start of matching, in
253 order to set up a JIT stack. If the return from the callback
254 function is NULL, the internal 32K stack is used; otherwise the
255 return value must be a valid JIT stack, the result of calling
256 <b>pcre_jit_stack_alloc()</b>.
257 </pre>
258 A callback function is obeyed whenever JIT code is about to be run; it is not
259 obeyed when <b>pcre_exec()</b> is called with options that are incompatible for
260 JIT execution. A callback function can therefore be used to determine whether a
261 match operation was executed by JIT or by the interpreter.
262 </P>
263 <P>
264 You may safely use the same JIT stack for more than one pattern (either by
265 assigning directly or by callback), as long as the patterns are all matched
266 sequentially in the same thread. In a multithread application, if you do not
267 specify a JIT stack, or if you assign or pass back NULL from a callback, that
268 is thread-safe, because each thread has its own machine stack. However, if you
269 assign or pass back a non-NULL JIT stack, this must be a different stack for
270 each thread so that the application is thread-safe.
271 </P>
272 <P>
273 Strictly speaking, even more is allowed. You can assign the same non-NULL stack
274 to any number of patterns as long as they are not used for matching by multiple
275 threads at the same time. For example, you can assign the same stack to all
276 compiled patterns, and use a global mutex in the callback to wait until the
277 stack is available for use. However, this is an inefficient solution, and not
278 recommended.
279 </P>
280 <P>
281 This is a suggestion for how a multithreaded program that needs to set up
282 non-default JIT stacks might operate:
283 <pre>
284 During thread initalization
285 thread_local_var = pcre_jit_stack_alloc(...)
286
287 During thread exit
288 pcre_jit_stack_free(thread_local_var)
289
290 Use a one-line callback function
291 return thread_local_var
292 </pre>
293 All the functions described in this section do nothing if JIT is not available,
294 and <b>pcre_assign_jit_stack()</b> does nothing unless the <b>extra</b> argument
295 is non-NULL and points to a <b>pcre_extra</b> block that is the result of a
296 successful study with PCRE_STUDY_JIT_COMPILE etc.
297 <a name="stackfaq"></a></P>
298 <br><a name="SEC9" href="#TOC1">JIT STACK FAQ</a><br>
299 <P>
300 (1) Why do we need JIT stacks?
301 <br>
302 <br>
303 PCRE (and JIT) is a recursive, depth-first engine, so it needs a stack where
304 the local data of the current node is pushed before checking its child nodes.
305 Allocating real machine stack on some platforms is difficult. For example, the
306 stack chain needs to be updated every time if we extend the stack on PowerPC.
307 Although it is possible, its updating time overhead decreases performance. So
308 we do the recursion in memory.
309 </P>
310 <P>
311 (2) Why don't we simply allocate blocks of memory with <b>malloc()</b>?
312 <br>
313 <br>
314 Modern operating systems have a nice feature: they can reserve an address space
315 instead of allocating memory. We can safely allocate memory pages inside this
316 address space, so the stack could grow without moving memory data (this is
317 important because of pointers). Thus we can allocate 1M address space, and use
318 only a single memory page (usually 4K) if that is enough. However, we can still
319 grow up to 1M anytime if needed.
320 </P>
321 <P>
322 (3) Who "owns" a JIT stack?
323 <br>
324 <br>
325 The owner of the stack is the user program, not the JIT studied pattern or
326 anything else. The user program must ensure that if a stack is used by
327 <b>pcre_exec()</b>, (that is, it is assigned to the pattern currently running),
328 that stack must not be used by any other threads (to avoid overwriting the same
329 memory area). The best practice for multithreaded programs is to allocate a
330 stack for each thread, and return this stack through the JIT callback function.
331 </P>
332 <P>
333 (4) When should a JIT stack be freed?
334 <br>
335 <br>
336 You can free a JIT stack at any time, as long as it will not be used by
337 <b>pcre_exec()</b> again. When you assign the stack to a pattern, only a pointer
338 is set. There is no reference counting or any other magic. You can free the
339 patterns and stacks in any order, anytime. Just <i>do not</i> call
340 <b>pcre_exec()</b> with a pattern pointing to an already freed stack, as that
341 will cause SEGFAULT. (Also, do not free a stack currently used by
342 <b>pcre_exec()</b> in another thread). You can also replace the stack for a
343 pattern at any time. You can even free the previous stack before assigning a
344 replacement.
345 </P>
346 <P>
347 (5) Should I allocate/free a stack every time before/after calling
348 <b>pcre_exec()</b>?
349 <br>
350 <br>
351 No, because this is too costly in terms of resources. However, you could
352 implement some clever idea which release the stack if it is not used in let's
353 say two minutes. The JIT callback can help to achive this without keeping a
354 list of the currently JIT studied patterns.
355 </P>
356 <P>
357 (6) OK, the stack is for long term memory allocation. But what happens if a
358 pattern causes stack overflow with a stack of 1M? Is that 1M kept until the
359 stack is freed?
360 <br>
361 <br>
362 Especially on embedded sytems, it might be a good idea to release memory
363 sometimes without freeing the stack. There is no API for this at the moment.
364 Probably a function call which returns with the currently allocated memory for
365 any stack and another which allows releasing memory (shrinking the stack) would
366 be a good idea if someone needs this.
367 </P>
368 <P>
369 (7) This is too much of a headache. Isn't there any better solution for JIT
370 stack handling?
371 <br>
372 <br>
373 No, thanks to Windows. If POSIX threads were used everywhere, we could throw
374 out this complicated API.
375 </P>
376 <br><a name="SEC10" href="#TOC1">EXAMPLE CODE</a><br>
377 <P>
378 This is a single-threaded example that specifies a JIT stack without using a
379 callback.
380 <pre>
381 int rc;
382 int ovector[30];
383 pcre *re;
384 pcre_extra *extra;
385 pcre_jit_stack *jit_stack;
386
387 re = pcre_compile(pattern, 0, &error, &erroffset, NULL);
388 /* Check for errors */
389 extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error);
390 jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024);
391 /* Check for error (NULL) */
392 pcre_assign_jit_stack(extra, NULL, jit_stack);
393 rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30);
394 /* Check results */
395 pcre_free(re);
396 pcre_free_study(extra);
397 pcre_jit_stack_free(jit_stack);
398
399 </PRE>
400 </P>
401 <br><a name="SEC11" href="#TOC1">SEE ALSO</a><br>
402 <P>
403 <b>pcreapi</b>(3)
404 </P>
405 <br><a name="SEC12" href="#TOC1">AUTHOR</a><br>
406 <P>
407 Philip Hazel (FAQ by Zoltan Herczeg)
408 <br>
409 University Computing Service
410 <br>
411 Cambridge CB2 3QH, England.
412 <br>
413 </P>
414 <br><a name="SEC13" href="#TOC1">REVISION</a><br>
415 <P>
416 Last updated: 04 May 2012
417 <br>
418 Copyright &copy; 1997-2012 University of Cambridge.
419 <br>
420 <p>
421 Return to the <a href="index.html">PCRE index page</a>.
422 </p>

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12