ViewVC logotype

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

Parent Directory Parent Directory | Revision Log Revision Log

Revision 975 - (hide annotations) (download) (as text)
Sat Jun 2 11:03:06 2012 UTC (2 years, 10 months ago) by ph10
File MIME type: text/html
File size: 18233 byte(s)
Document update for 8.31-RC1 test release.

1 ph10 678 <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 ph10 691 <li><a name="TOC1" href="#SEC1">PCRE JUST-IN-TIME COMPILER SUPPORT</a>
17 ph10 869 <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 ph10 678 </ul>
30 ph10 691 <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 ph10 869 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 ph10 691 </P>
41     <P>
42 ph10 869 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 ph10 691 </P>
46 ph10 869 <br><a name="SEC2" href="#TOC1">8-BIT and 16-BIT SUPPORT</a><br>
47 ph10 691 <P>
48 ph10 903 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 ph10 869 <i>pcre_jit_stack</i>).
53     </P>
54     <br><a name="SEC3" href="#TOC1">AVAILABILITY OF JIT SUPPORT</a><br>
55     <P>
56 ph10 691 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 ph10 869 Power PC 32-bit and 64-bit
64 ph10 691 </pre>
65 ph10 930 If --enable-jit is set on an unsupported platform, compilation fails.
66 ph10 691 </P>
67     <P>
68 ph10 836 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 ph10 930 way that falls back to the interpretive code if JIT is not available.
73 ph10 691 </P>
74 ph10 836 <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 ph10 869 <br><a name="SEC4" href="#TOC1">SIMPLE USE OF JIT</a><br>
81 ph10 691 <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>.
88     (2) Use <b>pcre_free_study()</b> to free the <b>pcre_extra</b> block when it is
89 ph10 930 no longer needed, instead of just freeing it yourself. This
90 ph10 691 ensures that any JIT data is also freed.
91     </pre>
92 ph10 836 For a program that may be linked with pre-8.20 versions of PCRE, you can insert
93     <pre>
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 ph10 975 PCRE_STUDY_JIT_COMPILE requests the JIT compiler to generate code for complete
108 ph10 930 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>
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 ph10 691 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 ph10 930 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 ph10 975 block containing a pointer to JIT code of the appropriate mode (normal or
133 ph10 930 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 ph10 691 </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 ph10 930 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 ph10 691 </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 ph10 708 JIT compilation was successful. A result of 0 means that JIT support is not
154 ph10 930 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 ph10 691 </P>
157 ph10 708 <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 ph10 869 <br><a name="SEC5" href="#TOC1">UNSUPPORTED OPTIONS AND PATTERN ITEMS</a><br>
162 ph10 691 <P>
163     The only <b>pcre_exec()</b> options that are supported for JIT execution are
166 ph10 691 </P>
167     <P>
168     The unsupported pattern items are:
169     <pre>
170 ph10 836 \C match a single byte; not supported in UTF-8 mode
171 ph10 691 (?Cn) callouts
172 ph10 975 (*PRUNE) )
173     (*SKIP) ) backtracking control verbs
174 ph10 691 (*THEN) )
175     </pre>
176     Support for some of these may be added in future.
177     </P>
178 ph10 869 <br><a name="SEC6" href="#TOC1">RETURN VALUES FROM JIT EXECUTION</a><br>
179 ph10 691 <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 ph10 869 <br><a name="SEC7" href="#TOC1">SAVING AND RESTORING COMPILED PATTERNS</a><br>
197 ph10 691 <P>
198     The code that is generated by the JIT compiler is architecture-specific, and is
199 ph10 708 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 ph10 691 <a name="stackcontrol"></a></P>
209 ph10 869 <br><a name="SEC8" href="#TOC1">CONTROLLING THE JIT STACK</a><br>
210 ph10 691 <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 ph10 836 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 ph10 691 </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 ph10 930 PCRE_STUDY_JIT_COMPILE etc. There are three cases for the values of the other
243     two options:
244 ph10 691 <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.
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>.
251 ph10 975 (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 ph10 930 <b>pcre_jit_stack_alloc()</b>.
257 ph10 691 </pre>
258 ph10 975 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 ph10 930 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 ph10 691 </P>
263     <P>
264 ph10 930 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 ph10 975 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 ph10 930 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 ph10 691 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 ph10 930 stack is available for use. However, this is an inefficient solution, and not
278     recommended.
279 ph10 691 </P>
280     <P>
281 ph10 930 This is a suggestion for how a multithreaded program that needs to set up
282     non-default JIT stacks might operate:
283 ph10 691 <pre>
284     During thread initalization
285     thread_local_var = pcre_jit_stack_alloc(...)
287     During thread exit
288     pcre_jit_stack_free(thread_local_var)
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 ph10 930 successful study with PCRE_STUDY_JIT_COMPILE etc.
297 ph10 836 <a name="stackfaq"></a></P>
298 ph10 869 <br><a name="SEC9" href="#TOC1">JIT STACK FAQ</a><br>
299 ph10 836 <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 ph10 691 </P>
310     <P>
311 ph10 836 (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 ph10 930 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 ph10 836 </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 ph10 869 <br><a name="SEC10" href="#TOC1">EXAMPLE CODE</a><br>
377 ph10 836 <P>
378 ph10 691 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;
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);
399     </PRE>
400     </P>
401 ph10 869 <br><a name="SEC11" href="#TOC1">SEE ALSO</a><br>
402 ph10 691 <P>
403     <b>pcreapi</b>(3)
404     </P>
405 ph10 869 <br><a name="SEC12" href="#TOC1">AUTHOR</a><br>
406 ph10 691 <P>
407 ph10 836 Philip Hazel (FAQ by Zoltan Herczeg)
408 ph10 691 <br>
409     University Computing Service
410     <br>
411     Cambridge CB2 3QH, England.
412     <br>
413     </P>
414 ph10 869 <br><a name="SEC13" href="#TOC1">REVISION</a><br>
415 ph10 691 <P>
416 ph10 975 Last updated: 04 May 2012
417 ph10 691 <br>
418 ph10 869 Copyright &copy; 1997-2012 University of Cambridge.
419 ph10 691 <br>
420 ph10 678 <p>
421     Return to the <a href="index.html">PCRE index page</a>.
422     </p>

ViewVC Help
Powered by ViewVC 1.1.12