ViewVC logotype

Contents of /code/branches/pcre16/doc/pcrejit.3

Parent Directory Parent Directory | Revision Log Revision Log

Revision 801 - (show annotations)
Mon Dec 12 16:23:37 2011 UTC (9 years, 4 months ago) by ph10
File size: 14230 byte(s)
Merge changes from trunk r755 to r800 into the 16-bit branch.
3 PCRE - Perl-compatible regular expressions
5 .rs
6 .sp
7 Just-in-time compiling is a heavyweight optimization that can greatly speed up
8 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 \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 \fBpcre_exec()\fP. If the subject string is very long, it may still pay to use
14 JIT for one-off matches.
15 .P
16 JIT support applies only to the traditional matching function,
17 \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 .
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 Intel x86 32-bit and 64-bit
30 MIPS 32-bit
31 Power PC 32-bit and 64-bit (experimental)
32 .sp
33 The Power PC support is designated as experimental because it has not been
34 fully tested. If --enable-jit is set on an unsupported platform, compilation
35 fails.
36 .P
37 A program that is linked with PCRE 8.20 or later can tell if JIT support is
38 available by calling \fBpcre_config()\fP with the PCRE_CONFIG_JIT option. The
39 result is 1 when JIT is available, and 0 otherwise. However, a simple program
40 does not need to check this in order to use JIT. The API is implemented in a
41 way that falls back to the ordinary PCRE code if JIT is not available.
42 .P
43 If your program may sometimes be linked with versions of PCRE that are older
44 than 8.20, but you want to use JIT when it is available, you can test
45 the values of PCRE_MAJOR and PCRE_MINOR, or the existence of a JIT macro such
46 as PCRE_CONFIG_JIT, for compile-time control of your code.
47 .
48 .
50 .rs
51 .sp
52 You have to do two things to make use of the JIT support in the simplest way:
53 .sp
54 (1) Call \fBpcre_study()\fP with the PCRE_STUDY_JIT_COMPILE option for
55 each compiled pattern, and pass the resulting \fBpcre_extra\fP block to
56 \fBpcre_exec()\fP.
57 .sp
58 (2) Use \fBpcre_free_study()\fP to free the \fBpcre_extra\fP block when it is
59 no longer needed instead of just freeing it yourself. This
60 ensures that any JIT data is also freed.
61 .sp
62 For a program that may be linked with pre-8.20 versions of PCRE, you can insert
63 .sp
66 #endif
67 .sp
68 so that no option is passed to \fBpcre_study()\fP, and then use something like
69 this to free the study data:
70 .sp
72 pcre_free_study(study_ptr);
73 #else
74 pcre_free(study_ptr);
75 #endif
76 .sp
77 In some circumstances you may need to call additional functions. These are
78 described in the section entitled
79 .\" HTML <a href="#stackcontrol">
80 .\" </a>
81 "Controlling the JIT stack"
82 .\"
83 below.
84 .P
85 If JIT support is not available, PCRE_STUDY_JIT_COMPILE is ignored, and no JIT
86 data is set up. Otherwise, the compiled pattern is passed to the JIT compiler,
87 which turns it into machine code that executes much faster than the normal
88 interpretive code. When \fBpcre_exec()\fP is passed a \fBpcre_extra\fP block
89 containing a pointer to JIT code, it obeys that instead of the normal code. The
90 result is identical, but the code runs much faster.
91 .P
92 There are some \fBpcre_exec()\fP options that are not supported for JIT
93 execution. There are also some pattern items that JIT cannot handle. Details
94 are given below. In both cases, execution automatically falls back to the
95 interpretive code.
96 .P
97 If the JIT compiler finds an unsupported item, no JIT data is generated. You
98 can find out if JIT execution is available after studying a pattern by calling
99 \fBpcre_fullinfo()\fP with the PCRE_INFO_JIT option. A result of 1 means that
100 JIT compilation was successful. A result of 0 means that JIT support is not
101 available, or the pattern was not studied with PCRE_STUDY_JIT_COMPILE, or the
102 JIT compiler was not able to handle the pattern.
103 .P
104 Once a pattern has been studied, with or without JIT, it can be used as many
105 times as you like for matching different subject strings.
106 .
107 .
109 .rs
110 .sp
111 The only \fBpcre_exec()\fP options that are supported for JIT execution are
113 PCRE_NOTEMPTY_ATSTART. Note in particular that partial matching is not
114 supported.
115 .P
116 The unsupported pattern items are:
117 .sp
118 \eC match a single byte; not supported in UTF-8 mode
119 (?Cn) callouts
120 (*COMMIT) )
121 (*MARK) )
122 (*PRUNE) ) the backtracking control verbs
123 (*SKIP) )
124 (*THEN) )
125 .sp
126 Support for some of these may be added in future.
127 .
128 .
130 .rs
131 .sp
132 When a pattern is matched using JIT execution, the return values are the same
133 as those given by the interpretive \fBpcre_exec()\fP code, with the addition of
134 one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means that the memory used
135 for the JIT stack was insufficient. See
136 .\" HTML <a href="#stackcontrol">
137 .\" </a>
138 "Controlling the JIT stack"
139 .\"
140 below for a discussion of JIT stack usage. For compatibility with the
141 interpretive \fBpcre_exec()\fP code, no more than two-thirds of the
142 \fIovector\fP argument is used for passing back captured substrings.
143 .P
144 The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if searching a
145 very large pattern tree goes on for too long, as it is in the same circumstance
146 when JIT is not used, but the details of exactly what is counted are not the
147 same. The PCRE_ERROR_RECURSIONLIMIT error code is never returned by JIT
148 execution.
149 .
150 .
152 .rs
153 .sp
154 The code that is generated by the JIT compiler is architecture-specific, and is
155 also position dependent. For those reasons it cannot be saved (in a file or
156 database) and restored later like the bytecode and other data of a compiled
157 pattern. Saving and restoring compiled patterns is not something many people
158 do. More detail about this facility is given in the
159 .\" HREF
160 \fBpcreprecompile\fP
161 .\"
162 documentation. It should be possible to run \fBpcre_study()\fP on a saved and
163 restored pattern, and thereby recreate the JIT data, but because JIT
164 compilation uses significant resources, it is probably not worth doing this;
165 you might as well recompile the original pattern.
166 .
167 .
168 .\" HTML <a name="stackcontrol"></a>
170 .rs
171 .sp
172 When the compiled JIT code runs, it needs a block of memory to use as a stack.
173 By default, it uses 32K on the machine stack. However, some large or
174 complicated patterns need more than this. The error PCRE_ERROR_JIT_STACKLIMIT
175 is given when there is not enough stack. Three functions are provided for
176 managing blocks of memory for use as JIT stacks. There is further discussion
177 about the use of JIT stacks in the section entitled
178 .\" HTML <a href="#stackcontrol">
179 .\" </a>
180 "JIT stack FAQ"
181 .\"
182 below.
183 .P
184 The \fBpcre_jit_stack_alloc()\fP function creates a JIT stack. Its arguments
185 are a starting size and a maximum size, and it returns a pointer to an opaque
186 structure of type \fBpcre_jit_stack\fP, or NULL if there is an error. The
187 \fBpcre_jit_stack_free()\fP function can be used to free a stack that is no
188 longer needed. (For the technically minded: the address space is allocated by
189 mmap or VirtualAlloc.)
190 .P
191 JIT uses far less memory for recursion than the interpretive code,
192 and a maximum stack size of 512K to 1M should be more than enough for any
193 pattern.
194 .P
195 The \fBpcre_assign_jit_stack()\fP function specifies which stack JIT code
196 should use. Its arguments are as follows:
197 .sp
198 pcre_extra *extra
199 pcre_jit_callback callback
200 void *data
201 .sp
202 The \fIextra\fP argument must be the result of studying a pattern with
203 PCRE_STUDY_JIT_COMPILE. There are three cases for the values of the other two
204 options:
205 .sp
206 (1) If \fIcallback\fP is NULL and \fIdata\fP is NULL, an internal 32K block
207 on the machine stack is used.
208 .sp
209 (2) If \fIcallback\fP is NULL and \fIdata\fP is not NULL, \fIdata\fP must be
210 a valid JIT stack, the result of calling \fBpcre_jit_stack_alloc()\fP.
211 .sp
212 (3) If \fIcallback\fP not NULL, it must point to a function that is called
213 with \fIdata\fP as an argument at the start of matching, in order to
214 set up a JIT stack. If the result is NULL, the internal 32K stack
215 is used; otherwise the return value must be a valid JIT stack,
216 the result of calling \fBpcre_jit_stack_alloc()\fP.
217 .sp
218 You may safely assign the same JIT stack to more than one pattern, as long as
219 they are all matched sequentially in the same thread. In a multithread
220 application, each thread must use its own JIT stack.
221 .P
222 Strictly speaking, even more is allowed. You can assign the same stack to any
223 number of patterns as long as they are not used for matching by multiple
224 threads at the same time. For example, you can assign the same stack to all
225 compiled patterns, and use a global mutex in the callback to wait until the
226 stack is available for use. However, this is an inefficient solution, and
227 not recommended.
228 .P
229 This is a suggestion for how a typical multithreaded program might operate:
230 .sp
231 During thread initalization
232 thread_local_var = pcre_jit_stack_alloc(...)
233 .sp
234 During thread exit
235 pcre_jit_stack_free(thread_local_var)
236 .sp
237 Use a one-line callback function
238 return thread_local_var
239 .sp
240 All the functions described in this section do nothing if JIT is not available,
241 and \fBpcre_assign_jit_stack()\fP does nothing unless the \fBextra\fP argument
242 is non-NULL and points to a \fBpcre_extra\fP block that is the result of a
243 successful study with PCRE_STUDY_JIT_COMPILE.
244 .
245 .
246 .\" HTML <a name="stackfaq"></a>
248 .rs
249 .sp
250 (1) Why do we need JIT stacks?
251 .sp
252 PCRE (and JIT) is a recursive, depth-first engine, so it needs a stack where
253 the local data of the current node is pushed before checking its child nodes.
254 Allocating real machine stack on some platforms is difficult. For example, the
255 stack chain needs to be updated every time if we extend the stack on PowerPC.
256 Although it is possible, its updating time overhead decreases performance. So
257 we do the recursion in memory.
258 .P
259 (2) Why don't we simply allocate blocks of memory with \fBmalloc()\fP?
260 .sp
261 Modern operating systems have a nice feature: they can reserve an address space
262 instead of allocating memory. We can safely allocate memory pages inside this
263 address space, so the stack could grow without moving memory data (this is
264 important because of pointers). Thus we can allocate 1M address space, and use
265 only a single memory page (usually 4K) if that is enough. However, we can still
266 grow up to 1M anytime if needed.
267 .P
268 (3) Who "owns" a JIT stack?
269 .sp
270 The owner of the stack is the user program, not the JIT studied pattern or
271 anything else. The user program must ensure that if a stack is used by
272 \fBpcre_exec()\fP, (that is, it is assigned to the pattern currently running),
273 that stack must not be used by any other threads (to avoid overwriting the same
274 memory area). The best practice for multithreaded programs is to allocate a
275 stack for each thread, and return this stack through the JIT callback function.
276 .P
277 (4) When should a JIT stack be freed?
278 .sp
279 You can free a JIT stack at any time, as long as it will not be used by
280 \fBpcre_exec()\fP again. When you assign the stack to a pattern, only a pointer
281 is set. There is no reference counting or any other magic. You can free the
282 patterns and stacks in any order, anytime. Just \fIdo not\fP call
283 \fBpcre_exec()\fP with a pattern pointing to an already freed stack, as that
284 will cause SEGFAULT. (Also, do not free a stack currently used by
285 \fBpcre_exec()\fP in another thread). You can also replace the stack for a
286 pattern at any time. You can even free the previous stack before assigning a
287 replacement.
288 .P
289 (5) Should I allocate/free a stack every time before/after calling
290 \fBpcre_exec()\fP?
291 .sp
292 No, because this is too costly in terms of resources. However, you could
293 implement some clever idea which release the stack if it is not used in let's
294 say two minutes. The JIT callback can help to achive this without keeping a
295 list of the currently JIT studied patterns.
296 .P
297 (6) OK, the stack is for long term memory allocation. But what happens if a
298 pattern causes stack overflow with a stack of 1M? Is that 1M kept until the
299 stack is freed?
300 .sp
301 Especially on embedded sytems, it might be a good idea to release
302 memory sometimes without freeing the stack. There is no API for this at the
303 moment. Probably a function call which returns with the currently allocated
304 memory for any stack and another which allows releasing memory (shrinking the
305 stack) would be a good idea if someone needs this.
306 .P
307 (7) This is too much of a headache. Isn't there any better solution for JIT
308 stack handling?
309 .sp
310 No, thanks to Windows. If POSIX threads were used everywhere, we could throw
311 out this complicated API.
312 .
313 .
315 .rs
316 .sp
317 This is a single-threaded example that specifies a JIT stack without using a
318 callback.
319 .sp
320 int rc;
321 int ovector[30];
322 pcre *re;
323 pcre_extra *extra;
324 pcre_jit_stack *jit_stack;
325 .sp
326 re = pcre_compile(pattern, 0, &error, &erroffset, NULL);
327 /* Check for errors */
328 extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error);
329 jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024);
330 /* Check for error (NULL) */
331 pcre_assign_jit_stack(extra, NULL, jit_stack);
332 rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30);
333 /* Check results */
334 pcre_free(re);
335 pcre_free_study(extra);
336 pcre_jit_stack_free(jit_stack);
337 .sp
338 .
339 .
340 .SH "SEE ALSO"
341 .rs
342 .sp
343 \fBpcreapi\fP(3)
344 .
345 .
347 .rs
348 .sp
349 .nf
350 Philip Hazel (FAQ by Zoltan Herczeg)
351 University Computing Service
352 Cambridge CB2 3QH, England.
353 .fi
354 .
355 .
357 .rs
358 .sp
359 .nf
360 Last updated: 26 November 2011
361 Copyright (c) 1997-2011 University of Cambridge.
362 .fi

  ViewVC Help
Powered by ViewVC 1.1.5