<html lang="en">
<head>
<title>__atomic Builtins - Using the GNU Compiler Collection (GCC)</title>
<meta http-equiv="Content-Type" content="text/html">
<meta name="description" content="Using the GNU Compiler Collection (GCC)">
<meta name="generator" content="makeinfo 4.13">
<link title="Top" rel="start" href="index.html#Top">
<link rel="up" href="C-Extensions.html#C-Extensions" title="C Extensions">
<link rel="prev" href="_005f_005fsync-Builtins.html#g_t_005f_005fsync-Builtins" title="__sync Builtins">
<link rel="next" href="Integer-Overflow-Builtins.html#Integer-Overflow-Builtins" title="Integer Overflow Builtins">
<link href="http://www.gnu.org/software/texinfo/" rel="generator-home" title="Texinfo Homepage">
<!--
Copyright (C) 1988-2015 Free Software Foundation, Inc.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with the
Invariant Sections being ``Funding Free Software'', the Front-Cover
Texts being (a) (see below), and with the Back-Cover Texts being (b)
(see below). A copy of the license is included in the section entitled
``GNU Free Documentation License''.
(a) The FSF's Front-Cover Text is:
A GNU Manual
(b) The FSF's Back-Cover Text is:
You have freedom to copy and modify this GNU Manual, like GNU
software. Copies published by the Free Software Foundation raise
funds for GNU development.-->
<meta http-equiv="Content-Style-Type" content="text/css">
<style type="text/css"><!--
pre.display { font-family:inherit }
pre.format { font-family:inherit }
pre.smalldisplay { font-family:inherit; font-size:smaller }
pre.smallformat { font-family:inherit; font-size:smaller }
pre.smallexample { font-size:smaller }
pre.smalllisp { font-size:smaller }
span.sc { font-variant:small-caps }
span.roman { font-family:serif; font-weight:normal; }
span.sansserif { font-family:sans-serif; font-weight:normal; }
--></style>
</head>
<body>
<div class="node">
<a name="__atomic-Builtins"></a>
<a name="g_t_005f_005fatomic-Builtins"></a>
<p>
Next: <a rel="next" accesskey="n" href="Integer-Overflow-Builtins.html#Integer-Overflow-Builtins">Integer Overflow Builtins</a>,
Previous: <a rel="previous" accesskey="p" href="_005f_005fsync-Builtins.html#g_t_005f_005fsync-Builtins">__sync Builtins</a>,
Up: <a rel="up" accesskey="u" href="C-Extensions.html#C-Extensions">C Extensions</a>
<hr>
</div>
<h3 class="section">6.51 Built-in Functions for Memory Model Aware Atomic Operations</h3>
<p>The following built-in functions approximately match the requirements for
C++11 memory model. Many are similar to the ‘<samp><span class="samp">__sync</span></samp>’ prefixed built-in
functions, but all also have a memory model parameter. These are all
identified by being prefixed with ‘<samp><span class="samp">__atomic</span></samp>’, and most are overloaded
such that they work with multiple types.
<p>GCC allows any integral scalar or pointer type that is 1, 2, 4, or 8
bytes in length. 16-byte integral types are also allowed if
‘<samp><span class="samp">__int128</span></samp>’ (see <a href="_005f_005fint128.html#g_t_005f_005fint128">__int128</a>) is supported by the architecture.
<p>Target architectures are encouraged to provide their own patterns for
each of these built-in functions. If no target is provided, the original
non-memory model set of ‘<samp><span class="samp">__sync</span></samp>’ atomic built-in functions are
utilized, along with any required synchronization fences surrounding it in
order to achieve the proper behavior. Execution in this case is subject
to the same restrictions as those built-in functions.
<p>If there is no pattern or mechanism to provide a lock free instruction
sequence, a call is made to an external routine with the same parameters
to be resolved at run time.
<p>The four non-arithmetic functions (load, store, exchange, and
compare_exchange) all have a generic version as well. This generic
version works on any data type. If the data type size maps to one
of the integral sizes that may have lock free support, the generic
version utilizes the lock free built-in function. Otherwise an
external call is left to be resolved at run time. This external call is
the same format with the addition of a ‘<samp><span class="samp">size_t</span></samp>’ parameter inserted
as the first parameter indicating the size of the object being pointed to.
All objects must be the same size.
<p>There are 6 different memory models that can be specified. These map
to the same names in the C++11 standard. Refer there or to the
<a href="http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync">GCC wiki on atomic synchronization</a> for more detailed definitions. These memory
models integrate both barriers to code motion as well as synchronization
requirements with other threads. These are listed in approximately
ascending order of strength. It is also possible to use target specific
flags for memory model flags, like Hardware Lock Elision.
<dl>
<dt><code>__ATOMIC_RELAXED</code><dd>No barriers or synchronization.
<br><dt><code>__ATOMIC_CONSUME</code><dd>Data dependency only for both barrier and synchronization with another
thread.
<br><dt><code>__ATOMIC_ACQUIRE</code><dd>Barrier to hoisting of code and synchronizes with release (or stronger)
semantic stores from another thread.
<br><dt><code>__ATOMIC_RELEASE</code><dd>Barrier to sinking of code and synchronizes with acquire (or stronger)
semantic loads from another thread.
<br><dt><code>__ATOMIC_ACQ_REL</code><dd>Full barrier in both directions and synchronizes with acquire loads and
release stores in another thread.
<br><dt><code>__ATOMIC_SEQ_CST</code><dd>Full barrier in both directions and synchronizes with acquire loads and
release stores in all threads.
</dl>
<p>When implementing patterns for these built-in functions, the memory model
parameter can be ignored as long as the pattern implements the most
restrictive <code>__ATOMIC_SEQ_CST</code> model. Any of the other memory models
execute correctly with this memory model but they may not execute as
efficiently as they could with a more appropriate implementation of the
relaxed requirements.
<p>Note that the C++11 standard allows for the memory model parameter to be
determined at run time rather than at compile time. These built-in
functions map any run-time value to <code>__ATOMIC_SEQ_CST</code> rather
than invoke a runtime library call or inline a switch statement. This is
standard compliant, safe, and the simplest approach for now.
<p>The memory model parameter is a signed int, but only the lower 16 bits are
reserved for the memory model. The remainder of the signed int is reserved
for target use and should be 0. Use of the predefined atomic values
ensures proper usage.
<div class="defun">
— Built-in Function: <var>type</var> <b>__atomic_load_n</b> (<var>type *ptr, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fload_005fn-3536"></a></var><br>
<blockquote><p>This built-in function implements an atomic load operation. It returns the
contents of <code>*</code><var>ptr</var>.
<p>The valid memory model variants are
<code>__ATOMIC_RELAXED</code>, <code>__ATOMIC_SEQ_CST</code>, <code>__ATOMIC_ACQUIRE</code>,
and <code>__ATOMIC_CONSUME</code>.
</blockquote></div>
<div class="defun">
— Built-in Function: void <b>__atomic_load</b> (<var>type *ptr, type *ret, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fload-3537"></a></var><br>
<blockquote><p>This is the generic version of an atomic load. It returns the
contents of <code>*</code><var>ptr</var> in <code>*</code><var>ret</var>.
</blockquote></div>
<div class="defun">
— Built-in Function: void <b>__atomic_store_n</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fstore_005fn-3538"></a></var><br>
<blockquote><p>This built-in function implements an atomic store operation. It writes
<var>val</var> into <code>*</code><var>ptr</var>.
<p>The valid memory model variants are
<code>__ATOMIC_RELAXED</code>, <code>__ATOMIC_SEQ_CST</code>, and <code>__ATOMIC_RELEASE</code>.
</blockquote></div>
<div class="defun">
— Built-in Function: void <b>__atomic_store</b> (<var>type *ptr, type *val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fstore-3539"></a></var><br>
<blockquote><p>This is the generic version of an atomic store. It stores the value
of <code>*</code><var>val</var> into <code>*</code><var>ptr</var>.
</blockquote></div>
<div class="defun">
— Built-in Function: <var>type</var> <b>__atomic_exchange_n</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fexchange_005fn-3540"></a></var><br>
<blockquote><p>This built-in function implements an atomic exchange operation. It writes
<var>val</var> into <code>*</code><var>ptr</var>, and returns the previous contents of
<code>*</code><var>ptr</var>.
<p>The valid memory model variants are
<code>__ATOMIC_RELAXED</code>, <code>__ATOMIC_SEQ_CST</code>, <code>__ATOMIC_ACQUIRE</code>,
<code>__ATOMIC_RELEASE</code>, and <code>__ATOMIC_ACQ_REL</code>.
</blockquote></div>
<div class="defun">
— Built-in Function: void <b>__atomic_exchange</b> (<var>type *ptr, type *val, type *ret, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fexchange-3541"></a></var><br>
<blockquote><p>This is the generic version of an atomic exchange. It stores the
contents of <code>*</code><var>val</var> into <code>*</code><var>ptr</var>. The original value
of <code>*</code><var>ptr</var> is copied into <code>*</code><var>ret</var>.
</blockquote></div>
<div class="defun">
— Built-in Function: bool <b>__atomic_compare_exchange_n</b> (<var>type *ptr, type *expected, type desired, bool weak, int success_memmodel, int failure_memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fcompare_005fexchange_005fn-3542"></a></var><br>
<blockquote><p>This built-in function implements an atomic compare and exchange operation.
This compares the contents of <code>*</code><var>ptr</var> with the contents of
<code>*</code><var>expected</var> and if equal, writes <var>desired</var> into
<code>*</code><var>ptr</var>. If they are not equal, the current contents of
<code>*</code><var>ptr</var> is written into <code>*</code><var>expected</var>. <var>weak</var> is true
for weak compare_exchange, and false for the strong variation. Many targets
only offer the strong variation and ignore the parameter. When in doubt, use
the strong variation.
<p>True is returned if <var>desired</var> is written into
<code>*</code><var>ptr</var> and the execution is considered to conform to the
memory model specified by <var>success_memmodel</var>. There are no
restrictions on what memory model can be used here.
<p>False is returned otherwise, and the execution is considered to conform
to <var>failure_memmodel</var>. This memory model cannot be
<code>__ATOMIC_RELEASE</code> nor <code>__ATOMIC_ACQ_REL</code>. It also cannot be a
stronger model than that specified by <var>success_memmodel</var>.
</blockquote></div>
<div class="defun">
— Built-in Function: bool <b>__atomic_compare_exchange</b> (<var>type *ptr, type *expected, type *desired, bool weak, int success_memmodel, int failure_memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fcompare_005fexchange-3543"></a></var><br>
<blockquote><p>This built-in function implements the generic version of
<code>__atomic_compare_exchange</code>. The function is virtually identical to
<code>__atomic_compare_exchange_n</code>, except the desired value is also a
pointer.
</blockquote></div>
<div class="defun">
— Built-in Function: <var>type</var> <b>__atomic_add_fetch</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fadd_005ffetch-3544"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_sub_fetch</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fsub_005ffetch-3545"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_and_fetch</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fand_005ffetch-3546"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_xor_fetch</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fxor_005ffetch-3547"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_or_fetch</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005for_005ffetch-3548"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_nand_fetch</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fnand_005ffetch-3549"></a></var><br>
<blockquote><p>These built-in functions perform the operation suggested by the name, and
return the result of the operation. That is,
<pre class="smallexample"> { *ptr <var>op</var>= val; return *ptr; }
</pre>
<p>All memory models are valid.
</blockquote></div>
<div class="defun">
— Built-in Function: <var>type</var> <b>__atomic_fetch_add</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005ffetch_005fadd-3550"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_fetch_sub</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005ffetch_005fsub-3551"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_fetch_and</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005ffetch_005fand-3552"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_fetch_xor</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005ffetch_005fxor-3553"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_fetch_or</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005ffetch_005for-3554"></a></var><br>
— Built-in Function: <var>type</var> <b>__atomic_fetch_nand</b> (<var>type *ptr, type val, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005ffetch_005fnand-3555"></a></var><br>
<blockquote><p>These built-in functions perform the operation suggested by the name, and
return the value that had previously been in <code>*</code><var>ptr</var>. That is,
<pre class="smallexample"> { tmp = *ptr; *ptr <var>op</var>= val; return tmp; }
</pre>
<p>All memory models are valid.
</blockquote></div>
<div class="defun">
— Built-in Function: bool <b>__atomic_test_and_set</b> (<var>void *ptr, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005ftest_005fand_005fset-3556"></a></var><br>
<blockquote>
<p>This built-in function performs an atomic test-and-set operation on
the byte at <code>*</code><var>ptr</var>. The byte is set to some implementation
defined nonzero “set” value and the return value is <code>true</code> if and only
if the previous contents were “set”.
It should be only used for operands of type <code>bool</code> or <code>char</code>. For
other types only part of the value may be set.
<p>All memory models are valid.
</blockquote></div>
<div class="defun">
— Built-in Function: void <b>__atomic_clear</b> (<var>bool *ptr, int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fclear-3557"></a></var><br>
<blockquote>
<p>This built-in function performs an atomic clear operation on
<code>*</code><var>ptr</var>. After the operation, <code>*</code><var>ptr</var> contains 0.
It should be only used for operands of type <code>bool</code> or <code>char</code> and
in conjunction with <code>__atomic_test_and_set</code>.
For other types it may only clear partially. If the type is not <code>bool</code>
prefer using <code>__atomic_store</code>.
<p>The valid memory model variants are
<code>__ATOMIC_RELAXED</code>, <code>__ATOMIC_SEQ_CST</code>, and
<code>__ATOMIC_RELEASE</code>.
</blockquote></div>
<div class="defun">
— Built-in Function: void <b>__atomic_thread_fence</b> (<var>int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fthread_005ffence-3558"></a></var><br>
<blockquote>
<p>This built-in function acts as a synchronization fence between threads
based on the specified memory model.
<p>All memory orders are valid.
</blockquote></div>
<div class="defun">
— Built-in Function: void <b>__atomic_signal_fence</b> (<var>int memmodel</var>)<var><a name="index-g_t_005f_005fatomic_005fsignal_005ffence-3559"></a></var><br>
<blockquote>
<p>This built-in function acts as a synchronization fence between a thread
and signal handlers based in the same thread.
<p>All memory orders are valid.
</blockquote></div>
<div class="defun">
— Built-in Function: bool <b>__atomic_always_lock_free</b> (<var>size_t size, void *ptr</var>)<var><a name="index-g_t_005f_005fatomic_005falways_005flock_005ffree-3560"></a></var><br>
<blockquote>
<p>This built-in function returns true if objects of <var>size</var> bytes always
generate lock free atomic instructions for the target architecture.
<var>size</var> must resolve to a compile-time constant and the result also
resolves to a compile-time constant.
<p><var>ptr</var> is an optional pointer to the object that may be used to determine
alignment. A value of 0 indicates typical alignment should be used. The
compiler may also ignore this parameter.
<pre class="smallexample"> if (_atomic_always_lock_free (sizeof (long long), 0))
</pre>
</blockquote></div>
<div class="defun">
— Built-in Function: bool <b>__atomic_is_lock_free</b> (<var>size_t size, void *ptr</var>)<var><a name="index-g_t_005f_005fatomic_005fis_005flock_005ffree-3561"></a></var><br>
<blockquote>
<p>This built-in function returns true if objects of <var>size</var> bytes always
generate lock free atomic instructions for the target architecture. If
it is not known to be lock free a call is made to a runtime routine named
<code>__atomic_is_lock_free</code>.
<p><var>ptr</var> is an optional pointer to the object that may be used to determine
alignment. A value of 0 indicates typical alignment should be used. The
compiler may also ignore this parameter.
</p></blockquote></div>
</body></html>