doc/html/pcrejit.html

<html>
<head>
<title>pcrejit specification</title>
</head>
<body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB">
<h1>pcrejit man page</h1>
<p>
Return to the <a href="index.html">PCRE index page</a>.
</p>
<p>
This page is part of the PCRE HTML documentation. It was generated automatically
from the original man page. If there is any nonsense in it, please consult the
man page, in case the conversion went wrong.
<br>
<ul>
<li><a name="TOC1" href="#SEC1">PCRE JUST-IN-TIME COMPILER SUPPORT</a>
<li><a name="TOC2" href="#SEC2">8-BIT, 16-BIT AND 32-BIT SUPPORT</a>
<li><a name="TOC3" href="#SEC3">AVAILABILITY OF JIT SUPPORT</a>
<li><a name="TOC4" href="#SEC4">SIMPLE USE OF JIT</a>
<li><a name="TOC5" href="#SEC5">UNSUPPORTED OPTIONS AND PATTERN ITEMS</a>
<li><a name="TOC6" href="#SEC6">RETURN VALUES FROM JIT EXECUTION</a>
<li><a name="TOC7" href="#SEC7">SAVING AND RESTORING COMPILED PATTERNS</a>
<li><a name="TOC8" href="#SEC8">CONTROLLING THE JIT STACK</a>
<li><a name="TOC9" href="#SEC9">JIT STACK FAQ</a>
<li><a name="TOC10" href="#SEC10">EXAMPLE CODE</a>
<li><a name="TOC11" href="#SEC11">JIT FAST PATH API</a>
<li><a name="TOC12" href="#SEC12">SEE ALSO</a>
<li><a name="TOC13" href="#SEC13">AUTHOR</a>
<li><a name="TOC14" href="#SEC14">REVISION</a>
</ul>
<br><a name="SEC1" href="#TOC1">PCRE JUST-IN-TIME COMPILER SUPPORT</a><br>
<P>
Just-in-time compiling is a heavyweight optimization that can greatly speed up
pattern matching. However, it comes at the cost of extra processing before the
match is performed. Therefore, it is of most benefit when the same pattern is
going to be matched many times. This does not necessarily mean many calls of a
matching function; if the pattern is not anchored, matching attempts may take
place many times at various positions in the subject, even for a single call.
Therefore, if the subject string is very long, it may still pay to use JIT for
one-off matches.
</P>
<P>
JIT support applies only to the traditional Perl-compatible matching function.
It does not apply when the DFA matching function is being used. The code for
this support was written by Zoltan Herczeg.
</P>
<br><a name="SEC2" href="#TOC1">8-BIT, 16-BIT AND 32-BIT SUPPORT</a><br>
<P>
JIT support is available for all of the 8-bit, 16-bit and 32-bit PCRE
libraries. To keep this documentation simple, only the 8-bit interface is
described in what follows. If you are using the 16-bit library, substitute the
16-bit functions and 16-bit structures (for example, <i>pcre16_jit_stack</i>
instead of <i>pcre_jit_stack</i>). If you are using the 32-bit library,
substitute the 32-bit functions and 32-bit structures (for example,
<i>pcre32_jit_stack</i> instead of <i>pcre_jit_stack</i>).
</P>
<br><a name="SEC3" href="#TOC1">AVAILABILITY OF JIT SUPPORT</a><br>
<P>
JIT support is an optional feature of PCRE. The "configure" option --enable-jit
(or equivalent CMake option) must be set when PCRE is built if you want to use
JIT. The support is limited to the following hardware platforms:
<pre>
  ARM v5, v7, and Thumb2
  Intel x86 32-bit and 64-bit
  MIPS 32-bit
  Power PC 32-bit and 64-bit
  SPARC 32-bit (experimental)
</pre>
If --enable-jit is set on an unsupported platform, compilation fails.
</P>
<P>
A program that is linked with PCRE 8.20 or later can tell if JIT support is
available by calling <b>pcre_config()</b> with the PCRE_CONFIG_JIT option. The
result is 1 when JIT is available, and 0 otherwise. However, a simple program
does not need to check this in order to use JIT. The normal API is implemented
in a way that falls back to the interpretive code if JIT is not available. For
programs that need the best possible performance, there is also a "fast path"
API that is JIT-specific.
</P>
<P>
If your program may sometimes be linked with versions of PCRE that are older
than 8.20, but you want to use JIT when it is available, you can test the
values of PCRE_MAJOR and PCRE_MINOR, or the existence of a JIT macro such as
PCRE_CONFIG_JIT, for compile-time control of your code. Also beware that the
<b>pcre_jit_exec()</b> function was not available at all before 8.32,
and may not be available at all if PCRE isn't compiled with
--enable-jit. See the "JIT FAST PATH API" section below for details.
</P>
<br><a name="SEC4" href="#TOC1">SIMPLE USE OF JIT</a><br>
<P>
You have to do two things to make use of the JIT support in the simplest way:
<pre>
  (1) Call <b>pcre_study()</b> with the PCRE_STUDY_JIT_COMPILE option for
      each compiled pattern, and pass the resulting <b>pcre_extra</b> block to
      <b>pcre_exec()</b>.

  (2) Use <b>pcre_free_study()</b> to free the <b>pcre_extra</b> block when it is
      no longer needed, instead of just freeing it yourself. This ensures that
      any JIT data is also freed.
</pre>
For a program that may be linked with pre-8.20 versions of PCRE, you can insert
<pre>
  #ifndef PCRE_STUDY_JIT_COMPILE
  #define PCRE_STUDY_JIT_COMPILE 0
  #endif
</pre>
so that no option is passed to <b>pcre_study()</b>, and then use something like
this to free the study data:
<pre>
  #ifdef PCRE_CONFIG_JIT
      pcre_free_study(study_ptr);
  #else
      pcre_free(study_ptr);
  #endif
</pre>
PCRE_STUDY_JIT_COMPILE requests the JIT compiler to generate code for complete
matches. If you want to run partial matches using the PCRE_PARTIAL_HARD or
PCRE_PARTIAL_SOFT options of <b>pcre_exec()</b>, you should set one or both of
the following options in addition to, or instead of, PCRE_STUDY_JIT_COMPILE
when you call <b>pcre_study()</b>:
<pre>
  PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
  PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
</pre>
If using <b>pcre_jit_exec()</b> and supporting a pre-8.32 version of
PCRE, you can insert:
<pre>
   #if PCRE_MAJOR &#62;= 8 && PCRE_MINOR &#62;= 32
   pcre_jit_exec(...);
   #else
   pcre_exec(...)
   #endif
</pre>
but as described in the "JIT FAST PATH API" section below this assumes
version 8.32 and later are compiled with --enable-jit, which may
break.
<br>
<br>
The JIT compiler generates different optimized code for each of the three
modes (normal, soft partial, hard partial). When <b>pcre_exec()</b> is called,
the appropriate code is run if it is available. Otherwise, the pattern is
matched using interpretive code.
</P>
<P>
In some circumstances you may need to call additional functions. These are
described in the section entitled
<a href="#stackcontrol">"Controlling the JIT stack"</a>
below.
</P>
<P>
If JIT support is not available, PCRE_STUDY_JIT_COMPILE etc. are ignored, and
no JIT data is created. Otherwise, the compiled pattern is passed to the JIT
compiler, which turns it into machine code that executes much faster than the
normal interpretive code. When <b>pcre_exec()</b> is passed a <b>pcre_extra</b>
block containing a pointer to JIT code of the appropriate mode (normal or
hard/soft partial), it obeys that code instead of running the interpreter. The
result is identical, but the compiled JIT code runs much faster.
</P>
<P>
There are some <b>pcre_exec()</b> options that are not supported for JIT
execution. There are also some pattern items that JIT cannot handle. Details
are given below. In both cases, execution automatically falls back to the
interpretive code. If you want to know whether JIT was actually used for a
particular match, you should arrange for a JIT callback function to be set up
as described in the section entitled
<a href="#stackcontrol">"Controlling the JIT stack"</a>
below, even if you do not need to supply a non-default JIT stack. Such a
callback function is called whenever JIT code is about to be obeyed. If the
execution options are not right for JIT execution, the callback function is not
obeyed.
</P>
<P>
If the JIT compiler finds an unsupported item, no JIT data is generated. You
can find out if JIT execution is available after studying a pattern by calling
<b>pcre_fullinfo()</b> with the PCRE_INFO_JIT option. A result of 1 means that
JIT compilation was successful. A result of 0 means that JIT support is not
available, or the pattern was not studied with PCRE_STUDY_JIT_COMPILE etc., or
the JIT compiler was not able to handle the pattern.
</P>
<P>
Once a pattern has been studied, with or without JIT, it can be used as many
times as you like for matching different subject strings.
</P>
<br><a name="SEC5" href="#TOC1">UNSUPPORTED OPTIONS AND PATTERN ITEMS</a><br>
<P>
The only <b>pcre_exec()</b> options that are supported for JIT execution are
PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK, PCRE_NO_UTF32_CHECK, PCRE_NOTBOL,
PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART, PCRE_PARTIAL_HARD, and
PCRE_PARTIAL_SOFT.
</P>
<P>
The only unsupported pattern items are \C (match a single data unit) when
running in a UTF mode, and a callout immediately before an assertion condition
in a conditional group.
</P>
<br><a name="SEC6" href="#TOC1">RETURN VALUES FROM JIT EXECUTION</a><br>
<P>
When a pattern is matched using JIT execution, the return values are the same
as those given by the interpretive <b>pcre_exec()</b> code, with the addition of
one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means that the memory used
for the JIT stack was insufficient. See
<a href="#stackcontrol">"Controlling the JIT stack"</a>
below for a discussion of JIT stack usage. For compatibility with the
interpretive <b>pcre_exec()</b> code, no more than two-thirds of the
<i>ovector</i> argument is used for passing back captured substrings.
</P>
<P>
The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if searching a
very large pattern tree goes on for too long, as it is in the same circumstance
when JIT is not used, but the details of exactly what is counted are not the
same. The PCRE_ERROR_RECURSIONLIMIT error code is never returned by JIT
execution.
</P>
<br><a name="SEC7" href="#TOC1">SAVING AND RESTORING COMPILED PATTERNS</a><br>
<P>
The code that is generated by the JIT compiler is architecture-specific, and is
also position dependent. For those reasons it cannot be saved (in a file or
database) and restored later like the bytecode and other data of a compiled
pattern. Saving and restoring compiled patterns is not something many people
do. More detail about this facility is given in the
<a href="pcreprecompile.html"><b>pcreprecompile</b></a>
documentation. It should be possible to run <b>pcre_study()</b> on a saved and
restored pattern, and thereby recreate the JIT data, but because JIT
compilation uses significant resources, it is probably not worth doing this;
you might as well recompile the original pattern.
<a name="stackcontrol"></a></P>
<br><a name="SEC8" href="#TOC1">CONTROLLING THE JIT STACK</a><br>
<P>
When the compiled JIT code runs, it needs a block of memory to use as a stack.
By default, it uses 32K on the machine stack. However, some large or
complicated patterns need more than this. The error PCRE_ERROR_JIT_STACKLIMIT
is given when there is not enough stack. Three functions are provided for
managing blocks of memory for use as JIT stacks. There is further discussion
about the use of JIT stacks in the section entitled
<a href="#stackcontrol">"JIT stack FAQ"</a>
below.
</P>
<P>
The <b>pcre_jit_stack_alloc()</b> function creates a JIT stack. Its arguments
are a starting size and a maximum size, and it returns a pointer to an opaque
structure of type <b>pcre_jit_stack</b>, or NULL if there is an error. The
<b>pcre_jit_stack_free()</b> function can be used to free a stack that is no
longer needed. (For the technically minded: the address space is allocated by
mmap or VirtualAlloc.)
</P>
<P>
JIT uses far less memory for recursion than the interpretive code,
and a maximum stack size of 512K to 1M should be more than enough for any
pattern.
</P>
<P>
The <b>pcre_assign_jit_stack()</b> function specifies which stack JIT code
should use. Its arguments are as follows:
<pre>
  pcre_extra         *extra
  pcre_jit_callback  callback
  void               *data
</pre>
The <i>extra</i> argument must be the result of studying a pattern with
PCRE_STUDY_JIT_COMPILE etc. There are three cases for the values of the other
two options:
<pre>
  (1) If <i>callback</i> is NULL and <i>data</i> is NULL, an internal 32K block
      on the machine stack is used.

  (2) If <i>callback</i> is NULL and <i>data</i> is not NULL, <i>data</i> must be
      a valid JIT stack, the result of calling <b>pcre_jit_stack_alloc()</b>.

  (3) If <i>callback</i> is not NULL, it must point to a function that is
      called with <i>data</i> as an argument at the start of matching, in
      order to set up a JIT stack. If the return from the callback
      function is NULL, the internal 32K stack is used; otherwise the
      return value must be a valid JIT stack, the result of calling
      <b>pcre_jit_stack_alloc()</b>.
</pre>
A callback function is obeyed whenever JIT code is about to be run; it is not
obeyed when <b>pcre_exec()</b> is called with options that are incompatible for
JIT execution. A callback function can therefore be used to determine whether a
match operation was executed by JIT or by the interpreter.
</P>
<P>
You may safely use the same JIT stack for more than one pattern (either by
assigning directly or by callback), as long as the patterns are all matched
sequentially in the same thread. In a multithread application, if you do not
specify a JIT stack, or if you assign or pass back NULL from a callback, that
is thread-safe, because each thread has its own machine stack. However, if you
assign or pass back a non-NULL JIT stack, this must be a different stack for
each thread so that the application is thread-safe.
</P>
<P>
Strictly speaking, even more is allowed. You can assign the same non-NULL stack
to any number of patterns as long as they are not used for matching by multiple
threads at the same time. For example, you can assign the same stack to all
compiled patterns, and use a global mutex in the callback to wait until the
stack is available for use. However, this is an inefficient solution, and not
recommended.
</P>
<P>
This is a suggestion for how a multithreaded program that needs to set up
non-default JIT stacks might operate:
<pre>
  During thread initalization
    thread_local_var = pcre_jit_stack_alloc(...)

  During thread exit
    pcre_jit_stack_free(thread_local_var)

  Use a one-line callback function
    return thread_local_var
</pre>
All the functions described in this section do nothing if JIT is not available,
and <b>pcre_assign_jit_stack()</b> does nothing unless the <b>extra</b> argument
is non-NULL and points to a <b>pcre_extra</b> block that is the result of a
successful study with PCRE_STUDY_JIT_COMPILE etc.
<a name="stackfaq"></a></P>
<br><a name="SEC9" href="#TOC1">JIT STACK FAQ</a><br>
<P>
(1) Why do we need JIT stacks?
<br>
<br>
PCRE (and JIT) is a recursive, depth-first engine, so it needs a stack where
the local data of the current node is pushed before checking its child nodes.
Allocating real machine stack on some platforms is difficult. For example, the
stack chain needs to be updated every time if we extend the stack on PowerPC.
Although it is possible, its updating time overhead decreases performance. So
we do the recursion in memory.
</P>
<P>
(2) Why don't we simply allocate blocks of memory with <b>malloc()</b>?
<br>
<br>
Modern operating systems have a nice feature: they can reserve an address space
instead of allocating memory. We can safely allocate memory pages inside this
address space, so the stack could grow without moving memory data (this is
important because of pointers). Thus we can allocate 1M address space, and use
only a single memory page (usually 4K) if that is enough. However, we can still
grow up to 1M anytime if needed.
</P>
<P>
(3) Who "owns" a JIT stack?
<br>
<br>
The owner of the stack is the user program, not the JIT studied pattern or
anything else. The user program must ensure that if a stack is used by
<b>pcre_exec()</b>, (that is, it is assigned to the pattern currently running),
that stack must not be used by any other threads (to avoid overwriting the same
memory area). The best practice for multithreaded programs is to allocate a
stack for each thread, and return this stack through the JIT callback function.
</P>
<P>
(4) When should a JIT stack be freed?
<br>
<br>
You can free a JIT stack at any time, as long as it will not be used by
<b>pcre_exec()</b> again. When you assign the stack to a pattern, only a pointer
is set. There is no reference counting or any other magic. You can free the
patterns and stacks in any order, anytime. Just <i>do not</i> call
<b>pcre_exec()</b> with a pattern pointing to an already freed stack, as that
will cause SEGFAULT. (Also, do not free a stack currently used by
<b>pcre_exec()</b> in another thread). You can also replace the stack for a
pattern at any time. You can even free the previous stack before assigning a
replacement.
</P>
<P>
(5) Should I allocate/free a stack every time before/after calling
<b>pcre_exec()</b>?
<br>
<br>
No, because this is too costly in terms of resources. However, you could
implement some clever idea which release the stack if it is not used in let's
say two minutes. The JIT callback can help to achieve this without keeping a
list of the currently JIT studied patterns.
</P>
<P>
(6) OK, the stack is for long term memory allocation. But what happens if a
pattern causes stack overflow with a stack of 1M? Is that 1M kept until the
stack is freed?
<br>
<br>
Especially on embedded sytems, it might be a good idea to release memory
sometimes without freeing the stack. There is no API for this at the moment.
Probably a function call which returns with the currently allocated memory for
any stack and another which allows releasing memory (shrinking the stack) would
be a good idea if someone needs this.
</P>
<P>
(7) This is too much of a headache. Isn't there any better solution for JIT
stack handling?
<br>
<br>
No, thanks to Windows. If POSIX threads were used everywhere, we could throw
out this complicated API.
</P>
<br><a name="SEC10" href="#TOC1">EXAMPLE CODE</a><br>
<P>
This is a single-threaded example that specifies a JIT stack without using a
callback.
<pre>
  int rc;
  int ovector[30];
  pcre *re;
  pcre_extra *extra;
  pcre_jit_stack *jit_stack;

  re = pcre_compile(pattern, 0, &error, &erroffset, NULL);
  /* Check for errors */
  extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error);
  jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024);
  /* Check for error (NULL) */
  pcre_assign_jit_stack(extra, NULL, jit_stack);
  rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30);
  /* Check results */
  pcre_free(re);
  pcre_free_study(extra);
  pcre_jit_stack_free(jit_stack);

</PRE>
</P>
<br><a name="SEC11" href="#TOC1">JIT FAST PATH API</a><br>
<P>
Because the API described above falls back to interpreted execution when JIT is
not available, it is convenient for programs that are written for general use
in many environments. However, calling JIT via <b>pcre_exec()</b> does have a
performance impact. Programs that are written for use where JIT is known to be
available, and which need the best possible performance, can instead use a
"fast path" API to call JIT execution directly instead of calling
<b>pcre_exec()</b> (obviously only for patterns that have been successfully
studied by JIT).
</P>
<P>
The fast path function is called <b>pcre_jit_exec()</b>, and it takes exactly
the same arguments as <b>pcre_exec()</b>, plus one additional argument that
must point to a JIT stack. The JIT stack arrangements described above do not
apply. The return values are the same as for <b>pcre_exec()</b>.
</P>
<P>
When you call <b>pcre_exec()</b>, as well as testing for invalid options, a
number of other sanity checks are performed on the arguments. For example, if
the subject pointer is NULL, or its length is negative, an immediate error is
given. Also, unless PCRE_NO_UTF[8|16|32] is set, a UTF subject string is tested
for validity. In the interests of speed, these checks do not happen on the JIT
fast path, and if invalid data is passed, the result is undefined.
</P>
<P>
Bypassing the sanity checks and the <b>pcre_exec()</b> wrapping can give
speedups of more than 10%.
</P>
<P>
Note that the <b>pcre_jit_exec()</b> function is not available in versions of
PCRE before 8.32 (released in November 2012). If you need to support versions
that old you must either use the slower <b>pcre_exec()</b>, or switch between
the two codepaths by checking the values of PCRE_MAJOR and PCRE_MINOR.
</P>
<P>
Due to an unfortunate implementation oversight, even in versions 8.32
and later there will be no <b>pcre_jit_exec()</b> stub function defined
when PCRE is compiled with --disable-jit, which is the default, and
there's no way to detect whether PCRE was compiled with --enable-jit
via a macro.
</P>
<P>
If you need to support versions older than 8.32, or versions that may
not build with --enable-jit, you must either use the slower
<b>pcre_exec()</b>, or switch between the two codepaths by checking the
values of PCRE_MAJOR and PCRE_MINOR.
</P>
<P>
Switching between the two by checking the version assumes that all the
versions being targeted are built with --enable-jit. To also support
builds that may use --disable-jit either <b>pcre_exec()</b> must be
used, or a compile-time check for JIT via <b>pcre_config()</b> (which
assumes the runtime environment will be the same), or as the Git
project decided to do, simply assume that <b>pcre_jit_exec()</b> is
present in 8.32 or later unless a compile-time flag is provided, see
the "grep: un-break building with PCRE &#62;= 8.32 without --enable-jit"
commit in git.git for an example of that.
</P>
<br><a name="SEC12" href="#TOC1">SEE ALSO</a><br>
<P>
<b>pcreapi</b>(3)
</P>
<br><a name="SEC13" href="#TOC1">AUTHOR</a><br>
<P>
Philip Hazel (FAQ by Zoltan Herczeg)
<br>
University Computing Service
<br>
Cambridge CB2 3QH, England.
<br>
</P>
<br><a name="SEC14" href="#TOC1">REVISION</a><br>
<P>
Last updated: 05 July 2017
<br>
Copyright &copy; 1997-2017 University of Cambridge.
<br>
<p>
Return to the <a href="index.html">PCRE index page</a>.
</p>