1PCREJIT(3) Library Functions Manual PCREJIT(3)
2
6 PCRE - Perl-compatible regular expressions
7
9 Just-in-time compiling is a heavyweight optimization that can greatly
11 speed up pattern matching. However, it comes at the cost of extra pro‐
12 cessing before the match is performed. Therefore, it is of most benefit
13 when the same pattern is going to be matched many times. This does not
14 necessarily mean many calls of a matching function; if the pattern is
15 not anchored, matching attempts may take place many times at various
16 positions in the subject, even for a single call. Therefore, if the
17 subject string is very long, it may still pay to use JIT for one-off
18 matches.
19 JIT support applies only to the traditional Perl-compatible matching
21 function. It does not apply when the DFA matching function is being
22 used. The code for this support was written by Zoltan Herczeg.
23
25 JIT support is available for all of the 8-bit, 16-bit and 32-bit PCRE
27 libraries. To keep this documentation simple, only the 8-bit interface
28 is described in what follows. If you are using the 16-bit library, sub‐
29 stitute the 16-bit functions and 16-bit structures (for example,
30 pcre16_jit_stack instead of pcre_jit_stack). If you are using the
31 32-bit library, substitute the 32-bit functions and 32-bit structures
32 (for example, pcre32_jit_stack instead of pcre_jit_stack).
33
35 JIT support is an optional feature of PCRE. The "configure" option
37 --enable-jit (or equivalent CMake option) must be set when PCRE is
38 built if you want to use JIT. The support is limited to the following
39 hardware platforms:
40 ARM v5, v7, and Thumb2
42 Intel x86 32-bit and 64-bit
43 MIPS 32-bit
44 Power PC 32-bit and 64-bit
45 SPARC 32-bit (experimental)
46 If --enable-jit is set on an unsupported platform, compilation fails.
48 A program that is linked with PCRE 8.20 or later can tell if JIT sup‐
50 port is available by calling pcre_config() with the PCRE_CONFIG_JIT
51 option. The result is 1 when JIT is available, and 0 otherwise. How‐
52 ever, a simple program does not need to check this in order to use JIT.
53 The normal API is implemented in a way that falls back to the interpre‐
54 tive code if JIT is not available. For programs that need the best pos‐
55 sible performance, there is also a "fast path" API that is JIT-spe‐
56 cific.
57 If your program may sometimes be linked with versions of PCRE that are
59 older than 8.20, but you want to use JIT when it is available, you can
60 test the values of PCRE_MAJOR and PCRE_MINOR, or the existence of a JIT
61 macro such as PCRE_CONFIG_JIT, for compile-time control of your code.
62
64 You have to do two things to make use of the JIT support in the sim‐
66 plest way:
67 (1) Call pcre_study() with the PCRE_STUDY_JIT_COMPILE option for
69 each compiled pattern, and pass the resulting pcre_extra block to
70 pcre_exec().
71 (2) Use pcre_free_study() to free the pcre_extra block when it is
73 no longer needed, instead of just freeing it yourself. This
74 ensures that
75 any JIT data is also freed.
76 For a program that may be linked with pre-8.20 versions of PCRE, you
78 can insert
79 #ifndef PCRE_STUDY_JIT_COMPILE
81 #define PCRE_STUDY_JIT_COMPILE 0
82 #endif
83 so that no option is passed to pcre_study(), and then use something
85 like this to free the study data:
86 #ifdef PCRE_CONFIG_JIT
88 pcre_free_study(study_ptr);
89 #else
90 pcre_free(study_ptr);
91 #endif
92 PCRE_STUDY_JIT_COMPILE requests the JIT compiler to generate code for
94 complete matches. If you want to run partial matches using the
95 PCRE_PARTIAL_HARD or PCRE_PARTIAL_SOFT options of pcre_exec(), you
96 should set one or both of the following options in addition to, or
97 instead of, PCRE_STUDY_JIT_COMPILE when you call pcre_study():
98 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
100 PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
101 The JIT compiler generates different optimized code for each of the
103 three modes (normal, soft partial, hard partial). When pcre_exec() is
104 called, the appropriate code is run if it is available. Otherwise, the
105 pattern is matched using interpretive code.
106 In some circumstances you may need to call additional functions. These
108 are described in the section entitled "Controlling the JIT stack"
109 below.
110 If JIT support is not available, PCRE_STUDY_JIT_COMPILE etc. are
112 ignored, and no JIT data is created. Otherwise, the compiled pattern is
113 passed to the JIT compiler, which turns it into machine code that exe‐
114 cutes much faster than the normal interpretive code. When pcre_exec()
115 is passed a pcre_extra block containing a pointer to JIT code of the
116 appropriate mode (normal or hard/soft partial), it obeys that code
117 instead of running the interpreter. The result is identical, but the
118 compiled JIT code runs much faster.
119 There are some pcre_exec() options that are not supported for JIT exe‐
121 cution. There are also some pattern items that JIT cannot handle.
122 Details are given below. In both cases, execution automatically falls
123 back to the interpretive code. If you want to know whether JIT was
124 actually used for a particular match, you should arrange for a JIT
125 callback function to be set up as described in the section entitled
126 "Controlling the JIT stack" below, even if you do not need to supply a
127 non-default JIT stack. Such a callback function is called whenever JIT
128 code is about to be obeyed. If the execution options are not right for
129 JIT execution, the callback function is not obeyed.
130 If the JIT compiler finds an unsupported item, no JIT data is gener‐
132 ated. You can find out if JIT execution is available after studying a
133 pattern by calling pcre_fullinfo() with the PCRE_INFO_JIT option. A
134 result of 1 means that JIT compilation was successful. A result of 0
135 means that JIT support is not available, or the pattern was not studied
136 with PCRE_STUDY_JIT_COMPILE etc., or the JIT compiler was not able to
137 handle the pattern.
138 Once a pattern has been studied, with or without JIT, it can be used as
140 many times as you like for matching different subject strings.
141
143 The only pcre_exec() options that are supported for JIT execution are
145 PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK, PCRE_NO_UTF32_CHECK, PCRE_NOT‐
146 BOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART, PCRE_PAR‐
147 TIAL_HARD, and PCRE_PARTIAL_SOFT.
148 The unsupported pattern items are:
150 \C match a single byte; not supported in UTF-8 mode
152 (?Cn) callouts
153 (*PRUNE) )
154 (*SKIP) ) backtracking control verbs
155 (*THEN) )
156 Support for some of these may be added in future.
158
160 When a pattern is matched using JIT execution, the return values are
162 the same as those given by the interpretive pcre_exec() code, with the
163 addition of one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means
164 that the memory used for the JIT stack was insufficient. See "Control‐
165 ling the JIT stack" below for a discussion of JIT stack usage. For com‐
166 patibility with the interpretive pcre_exec() code, no more than two-
167 thirds of the ovector argument is used for passing back captured sub‐
168 strings.
169 The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if
171 searching a very large pattern tree goes on for too long, as it is in
172 the same circumstance when JIT is not used, but the details of exactly
173 what is counted are not the same. The PCRE_ERROR_RECURSIONLIMIT error
174 code is never returned by JIT execution.
175
177 The code that is generated by the JIT compiler is architecture-spe‐
179 cific, and is also position dependent. For those reasons it cannot be
180 saved (in a file or database) and restored later like the bytecode and
181 other data of a compiled pattern. Saving and restoring compiled pat‐
182 terns is not something many people do. More detail about this facility
183 is given in the pcreprecompile documentation. It should be possible to
184 run pcre_study() on a saved and restored pattern, and thereby recreate
185 the JIT data, but because JIT compilation uses significant resources,
186 it is probably not worth doing this; you might as well recompile the
187 original pattern.
188
190 When the compiled JIT code runs, it needs a block of memory to use as a
192 stack. By default, it uses 32K on the machine stack. However, some
193 large or complicated patterns need more than this. The error
194 PCRE_ERROR_JIT_STACKLIMIT is given when there is not enough stack.
195 Three functions are provided for managing blocks of memory for use as
196 JIT stacks. There is further discussion about the use of JIT stacks in
197 the section entitled "JIT stack FAQ" below.
198 The pcre_jit_stack_alloc() function creates a JIT stack. Its arguments
200 are a starting size and a maximum size, and it returns a pointer to an
201 opaque structure of type pcre_jit_stack, or NULL if there is an error.
202 The pcre_jit_stack_free() function can be used to free a stack that is
203 no longer needed. (For the technically minded: the address space is
204 allocated by mmap or VirtualAlloc.)
205 JIT uses far less memory for recursion than the interpretive code, and
207 a maximum stack size of 512K to 1M should be more than enough for any
208 pattern.
209 The pcre_assign_jit_stack() function specifies which stack JIT code
211 should use. Its arguments are as follows:
212 pcre_extra *extra
214 pcre_jit_callback callback
215 void *data
216 The extra argument must be the result of studying a pattern with
218 PCRE_STUDY_JIT_COMPILE etc. There are three cases for the values of the
219 other two options:
220 (1) If callback is NULL and data is NULL, an internal 32K block
222 on the machine stack is used.
223 (2) If callback is NULL and data is not NULL, data must be
225 a valid JIT stack, the result of calling pcre_jit_stack_alloc().
226 (3) If callback is not NULL, it must point to a function that is
228 called with data as an argument at the start of matching, in
229 order to set up a JIT stack. If the return from the callback
230 function is NULL, the internal 32K stack is used; otherwise the
231 return value must be a valid JIT stack, the result of calling
232 pcre_jit_stack_alloc().
233 A callback function is obeyed whenever JIT code is about to be run; it
235 is not obeyed when pcre_exec() is called with options that are incom‐
236 patible for JIT execution. A callback function can therefore be used to
237 determine whether a match operation was executed by JIT or by the
238 interpreter.
239 You may safely use the same JIT stack for more than one pattern (either
241 by assigning directly or by callback), as long as the patterns are all
242 matched sequentially in the same thread. In a multithread application,
243 if you do not specify a JIT stack, or if you assign or pass back NULL
244 from a callback, that is thread-safe, because each thread has its own
245 machine stack. However, if you assign or pass back a non-NULL JIT
246 stack, this must be a different stack for each thread so that the
247 application is thread-safe.
248 Strictly speaking, even more is allowed. You can assign the same non-
250 NULL stack to any number of patterns as long as they are not used for
251 matching by multiple threads at the same time. For example, you can
252 assign the same stack to all compiled patterns, and use a global mutex
253 in the callback to wait until the stack is available for use. However,
254 this is an inefficient solution, and not recommended.
255 This is a suggestion for how a multithreaded program that needs to set
257 up non-default JIT stacks might operate:
258 During thread initalization
260 thread_local_var = pcre_jit_stack_alloc(...)
261 During thread exit
263 pcre_jit_stack_free(thread_local_var)
264 Use a one-line callback function
266 return thread_local_var
267 All the functions described in this section do nothing if JIT is not
269 available, and pcre_assign_jit_stack() does nothing unless the extra
270 argument is non-NULL and points to a pcre_extra block that is the
271 result of a successful study with PCRE_STUDY_JIT_COMPILE etc.
272
274 (1) Why do we need JIT stacks?
276 PCRE (and JIT) is a recursive, depth-first engine, so it needs a stack
278 where the local data of the current node is pushed before checking its
279 child nodes. Allocating real machine stack on some platforms is diffi‐
280 cult. For example, the stack chain needs to be updated every time if we
281 extend the stack on PowerPC. Although it is possible, its updating
282 time overhead decreases performance. So we do the recursion in memory.
283 (2) Why don't we simply allocate blocks of memory with malloc()?
285 Modern operating systems have a nice feature: they can reserve an
287 address space instead of allocating memory. We can safely allocate mem‐
288 ory pages inside this address space, so the stack could grow without
289 moving memory data (this is important because of pointers). Thus we can
290 allocate 1M address space, and use only a single memory page (usually
291 4K) if that is enough. However, we can still grow up to 1M anytime if
292 needed.
293 (3) Who "owns" a JIT stack?
295 The owner of the stack is the user program, not the JIT studied pattern
297 or anything else. The user program must ensure that if a stack is used
298 by pcre_exec(), (that is, it is assigned to the pattern currently run‐
299 ning), that stack must not be used by any other threads (to avoid over‐
300 writing the same memory area). The best practice for multithreaded pro‐
301 grams is to allocate a stack for each thread, and return this stack
302 through the JIT callback function.
303 (4) When should a JIT stack be freed?
305 You can free a JIT stack at any time, as long as it will not be used by
307 pcre_exec() again. When you assign the stack to a pattern, only a
308 pointer is set. There is no reference counting or any other magic. You
309 can free the patterns and stacks in any order, anytime. Just do not
310 call pcre_exec() with a pattern pointing to an already freed stack, as
311 that will cause SEGFAULT. (Also, do not free a stack currently used by
312 pcre_exec() in another thread). You can also replace the stack for a
313 pattern at any time. You can even free the previous stack before
314 assigning a replacement.
315 (5) Should I allocate/free a stack every time before/after calling
317 pcre_exec()?
318 No, because this is too costly in terms of resources. However, you
320 could implement some clever idea which release the stack if it is not
321 used in let's say two minutes. The JIT callback can help to achieve
322 this without keeping a list of the currently JIT studied patterns.
323 (6) OK, the stack is for long term memory allocation. But what happens
325 if a pattern causes stack overflow with a stack of 1M? Is that 1M kept
326 until the stack is freed?
327 Especially on embedded sytems, it might be a good idea to release mem‐
329 ory sometimes without freeing the stack. There is no API for this at
330 the moment. Probably a function call which returns with the currently
331 allocated memory for any stack and another which allows releasing mem‐
332 ory (shrinking the stack) would be a good idea if someone needs this.
333 (7) This is too much of a headache. Isn't there any better solution for
335 JIT stack handling?
336 No, thanks to Windows. If POSIX threads were used everywhere, we could
338 throw out this complicated API.
339
341 This is a single-threaded example that specifies a JIT stack without
343 using a callback.
344 int rc;
346 int ovector[30];
347 pcre *re;
348 pcre_extra *extra;
349 pcre_jit_stack *jit_stack;
350 re = pcre_compile(pattern, 0, &error, &erroffset, NULL);
352 /* Check for errors */
353 extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error);
354 jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024);
355 /* Check for error (NULL) */
356 pcre_assign_jit_stack(extra, NULL, jit_stack);
357 rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30);
358 /* Check results */
359 pcre_free(re);
360 pcre_free_study(extra);
361 pcre_jit_stack_free(jit_stack);
362
365 Because the API described above falls back to interpreted execution
367 when JIT is not available, it is convenient for programs that are writ‐
368 ten for general use in many environments. However, calling JIT via
369 pcre_exec() does have a performance impact. Programs that are written
370 for use where JIT is known to be available, and which need the best
371 possible performance, can instead use a "fast path" API to call JIT
372 execution directly instead of calling pcre_exec() (obviously only for
373 patterns that have been successfully studied by JIT).
374 The fast path function is called pcre_jit_exec(), and it takes exactly
376 the same arguments as pcre_exec(), plus one additional argument that
377 must point to a JIT stack. The JIT stack arrangements described above
378 do not apply. The return values are the same as for pcre_exec().
379 When you call pcre_exec(), as well as testing for invalid options, a
381 number of other sanity checks are performed on the arguments. For exam‐
382 ple, if the subject pointer is NULL, or its length is negative, an
383 immediate error is given. Also, unless PCRE_NO_UTF[8|16|32] is set, a
384 UTF subject string is tested for validity. In the interests of speed,
385 these checks do not happen on the JIT fast path, and if invalid data is
386 passed, the result is undefined.
387 Bypassing the sanity checks and the pcre_exec() wrapping can give
389 speedups of more than 10%.
390
392 pcreapi(3)
394
396 Philip Hazel (FAQ by Zoltan Herczeg)
398 University Computing Service
399 Cambridge CB2 3QH, England.
400
402 Last updated: 31 October 2012
404 Copyright (c) 1997-2012 University of Cambridge.
405PCRE 8.32 31 October 2012 PCREJIT(3)