]> rtime.felk.cvut.cz Git - l4.git/blob - l4/pkg/libstdc++-v3/contrib/libstdc++-v3-4.3.3/doc/html/manual/concurrency.html
update
[l4.git] / l4 / pkg / libstdc++-v3 / contrib / libstdc++-v3-4.3.3 / doc / html / manual / concurrency.html
1 <?xml version="1.0" encoding="UTF-8" standalone="no"?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
3 <html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>Chapter 40. Concurrency</title><meta name="generator" content="DocBook XSL Stylesheets V1.73.2" /><meta name="keywords" content="&#10;      ISO C++&#10;    , &#10;      library&#10;    " /><meta name="keywords" content="&#10;      ISO C++&#10;    , &#10;      library&#10;    " /><link rel="start" href="../spine.html" title="The GNU C++ Library Documentation" /><link rel="up" href="extensions.html" title="Part XII. Extensions" /><link rel="prev" href="bk01pt12ch39.html" title="Chapter 39. Demangling" /><link rel="next" href="bk01pt12ch40s02.html" title="Implementation" /></head><body><div class="navheader"><table width="100%" summary="Navigation header"><tr><th colspan="3" align="center">Chapter 40. Concurrency</th></tr><tr><td width="20%" align="left"><a accesskey="p" href="bk01pt12ch39.html">Prev</a> </td><th width="60%" align="center">Part XII. Extensions</th><td width="20%" align="right"> <a accesskey="n" href="bk01pt12ch40s02.html">Next</a></td></tr></table><hr /></div><div class="chapter" lang="en" xml:lang="en"><div class="titlepage"><div><div><h2 class="title"><a id="manual.ext.concurrency"></a>Chapter 40. Concurrency</h2></div></div></div><div class="toc"><p><b>Table of Contents</b></p><dl><dt><span class="sect1"><a href="concurrency.html#manual.ext.concurrency.design">Design</a></span></dt><dd><dl><dt><span class="sect2"><a href="concurrency.html#manual.ext.concurrency.design.threads">Interface to Locks and Mutexes</a></span></dt><dt><span class="sect2"><a href="concurrency.html#manual.ext.concurrency.design.atomics">Interface to Atomic Functions</a></span></dt></dl></dd><dt><span class="sect1"><a href="bk01pt12ch40s02.html">Implementation</a></span></dt><dd><dl><dt><span class="sect2"><a href="bk01pt12ch40s02.html#manual.ext.concurrency.impl.atomic_fallbacks">Using Builtin Atomic Functions</a></span></dt><dt><span class="sect2"><a href="bk01pt12ch40s02.html#manual.ext.concurrency.impl.thread">Thread Abstraction</a></span></dt></dl></dd><dt><span class="sect1"><a href="bk01pt12ch40s03.html">Use</a></span></dt></dl></div><div class="sect1" lang="en" xml:lang="en"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="manual.ext.concurrency.design"></a>Design</h2></div></div></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.concurrency.design.threads"></a>Interface to Locks and Mutexes</h3></div></div></div><p>The file &lt;ext/concurrence.h&gt; contains all the higher-level
4 constructs for playing with threads. In contrast to the atomics layer,
5 the concurrence layer consists largely of types. All types are defined within <code class="code">namespace __gnu_cxx</code>.
6 </p><p>
7 These types can be used in a portable manner, regardless of the
8 specific environment. They are carefully designed to provide optimum
9 efficiency and speed, abstracting out underlying thread calls and
10 accesses when compiling for single-threaded situations (even on hosts
11 that support multiple threads.)
12 </p><p>The enumerated type <code class="code">_Lock_policy</code> details the set of
13 available locking
14 policies: <code class="code">_S_single</code>, <code class="code">_S_mutex</code>,
15 and <code class="code">_S_atomic</code>.
16 </p><div class="itemizedlist"><ul type="disc"><li><p><code class="code">_S_single</code></p><p>Indicates single-threaded code that does not need locking.
17 </p></li><li><p><code class="code">_S_mutex</code></p><p>Indicates multi-threaded code using thread-layer abstractions.
18 </p></li><li><p><code class="code">_S_atomic</code></p><p>Indicates multi-threaded code using atomic operations.
19 </p></li></ul></div><p>The compile-time constant <code class="code">__default_lock_policy</code> is set
20 to one of the three values above, depending on characteristics of the
21 host environment and the current compilation flags.
22 </p><p>Two more datatypes make up the rest of the
23 interface: <code class="code">__mutex</code>, and <code class="code">__scoped_lock</code>.
24 </p><p>
25 </p><p>The scoped lock idiom is well-discussed within the C++
26 community. This version takes a <code class="code">__mutex</code> reference, and
27 locks it during construction of <code class="code">__scoped_locke</code> and
28 unlocks it during destruction. This is an efficient way of locking
29 critical sections, while retaining exception-safety.
30 </p></div><div class="sect2" lang="en" xml:lang="en"><div class="titlepage"><div><div><h3 class="title"><a id="manual.ext.concurrency.design.atomics"></a>Interface to Atomic Functions</h3></div></div></div><p>
31 Two functions and one type form the base of atomic support. 
32 </p><p>The type <code class="code">_Atomic_word</code> is a signed integral type
33 supporting atomic operations.
34 </p><p>
35 The two functions functions are:
36 </p><pre class="programlisting">
37 _Atomic_word
38 __exchange_and_add_dispatch(volatile _Atomic_word*, int);
39
40 void
41 __atomic_add_dispatch(volatile _Atomic_word*, int);
42 </pre><p>Both of these functions are declared in the header file
43 &lt;ext/atomicity.h&gt;, and are in <code class="code">namespace __gnu_cxx</code>.
44 </p><div class="itemizedlist"><ul type="disc"><li><p>
45 <code class="code">
46 __exchange_and_add_dispatch
47 </code>
48 </p><p>Adds the second argument's value to the first argument. Returns the old value.
49 </p></li><li><p>
50 <code class="code">
51 __atomic_add_dispatch
52 </code>
53 </p><p>Adds the second argument's value to the first argument. Has no return value.
54 </p></li></ul></div><p>
55 These functions forward to one of several specialized helper
56 functions, depending on the circumstances. For instance, 
57 </p><p>
58 <code class="code">
59 __exchange_and_add_dispatch
60 </code>
61 </p><p>
62 Calls through to either of:
63 </p><div class="itemizedlist"><ul type="disc"><li><p><code class="code">__exchange_and_add</code>
64 </p><p>Multi-thread version. Inlined if compiler-generated builtin atomics
65 can be used, otherwise resolved at link time to a non-builtin code
66 sequence.
67 </p></li><li><p><code class="code">__exchange_and_add_single</code> 
68 </p><p>Single threaded version. Inlined.</p></li></ul></div><p>However, only <code class="code">__exchange_and_add_dispatch</code>
69 and <code class="code">__atomic_add_dispatch</code> should be used. These functions
70 can be used in a portable manner, regardless of the specific
71 environment. They are carefully designed to provide optimum efficiency
72 and speed, abstracting out atomic accesses when they are not required
73 (even on hosts that support compiler intrinsics for atomic
74 operations.)
75 </p><p>
76 In addition, there are two macros
77 </p><p>
78 <code class="code">
79 _GLIBCXX_READ_MEM_BARRIER 
80 </code>
81 </p><p>
82 <code class="code">
83 _GLIBCXX_WRITE_MEM_BARRIER 
84 </code>
85 </p><p>
86 Which expand to the appropriate write and read barrier required by the
87 host hardware and operating system.
88 </p></div></div></div><div class="navfooter"><hr /><table width="100%" summary="Navigation footer"><tr><td width="40%" align="left"><a accesskey="p" href="bk01pt12ch39.html">Prev</a> </td><td width="20%" align="center"><a accesskey="u" href="extensions.html">Up</a></td><td width="40%" align="right"> <a accesskey="n" href="bk01pt12ch40s02.html">Next</a></td></tr><tr><td width="40%" align="left" valign="top">Chapter 39. Demangling </td><td width="20%" align="center"><a accesskey="h" href="../spine.html">Home</a></td><td width="40%" align="right" valign="top"> Implementation</td></tr></table></div></body></html>