You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1257 lines
56 KiB
HTML

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
<html>
<!-- Copyright (C) 1988-2018 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. -->
<!-- Created by GNU Texinfo 6.4, http://www.gnu.org/software/texinfo/ -->
<head>
<title>Extended Asm (Using the GNU Compiler Collection (GCC))</title>
<meta name="description" content="Extended Asm (Using the GNU Compiler Collection (GCC))">
<meta name="keywords" content="Extended Asm (Using the GNU Compiler Collection (GCC))">
<meta name="resource-type" content="document">
<meta name="distribution" content="global">
<meta name="Generator" content="makeinfo">
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<link href="index.html#Top" rel="start" title="Top">
<link href="Option-Index.html#Option-Index" rel="index" title="Option Index">
<link href="index.html#SEC_Contents" rel="contents" title="Table of Contents">
<link href="Using-Assembly-Language-with-C.html#Using-Assembly-Language-with-C" rel="up" title="Using Assembly Language with C">
<link href="Constraints.html#Constraints" rel="next" title="Constraints">
<link href="Basic-Asm.html#Basic-Asm" rel="prev" title="Basic Asm">
<style type="text/css">
<!--
a.summary-letter {text-decoration: none}
blockquote.indentedblock {margin-right: 0em}
blockquote.smallindentedblock {margin-right: 0em; font-size: smaller}
blockquote.smallquotation {font-size: smaller}
div.display {margin-left: 3.2em}
div.example {margin-left: 3.2em}
div.lisp {margin-left: 3.2em}
div.smalldisplay {margin-left: 3.2em}
div.smallexample {margin-left: 3.2em}
div.smalllisp {margin-left: 3.2em}
kbd {font-style: oblique}
pre.display {font-family: inherit}
pre.format {font-family: inherit}
pre.menu-comment {font-family: serif}
pre.menu-preformatted {font-family: serif}
pre.smalldisplay {font-family: inherit; font-size: smaller}
pre.smallexample {font-size: smaller}
pre.smallformat {font-family: inherit; font-size: smaller}
pre.smalllisp {font-size: smaller}
span.nolinebreak {white-space: nowrap}
span.roman {font-family: initial; font-weight: normal}
span.sansserif {font-family: sans-serif; font-weight: normal}
ul.no-bullet {list-style: none}
-->
</style>
</head>
<body lang="en">
<a name="Extended-Asm"></a>
<div class="header">
<p>
Next: <a href="Constraints.html#Constraints" accesskey="n" rel="next">Constraints</a>, Previous: <a href="Basic-Asm.html#Basic-Asm" accesskey="p" rel="prev">Basic Asm</a>, Up: <a href="Using-Assembly-Language-with-C.html#Using-Assembly-Language-with-C" accesskey="u" rel="up">Using Assembly Language with C</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Option-Index.html#Option-Index" title="Index" rel="index">Index</a>]</p>
</div>
<hr>
<a name="Extended-Asm-_002d-Assembler-Instructions-with-C-Expression-Operands"></a>
<h4 class="subsection">6.45.2 Extended Asm - Assembler Instructions with C Expression Operands</h4>
<a name="index-extended-asm"></a>
<a name="index-assembly-language-in-C_002c-extended"></a>
<p>With extended <code>asm</code> you can read and write C variables from
assembler and perform jumps from assembler code to C labels.
Extended <code>asm</code> syntax uses colons (&lsquo;<samp>:</samp>&rsquo;) to delimit
the operand parameters after the assembler template:
</p>
<div class="example">
<pre class="example">asm <var>asm-qualifiers</var> ( <var>AssemblerTemplate</var>
: <var>OutputOperands</var>
<span class="roman">[</span> : <var>InputOperands</var>
<span class="roman">[</span> : <var>Clobbers</var> <span class="roman">]</span> <span class="roman">]</span>)
asm <var>asm-qualifiers</var> ( <var>AssemblerTemplate</var>
:
: <var>InputOperands</var>
: <var>Clobbers</var>
: <var>GotoLabels</var>)
</pre></div>
<p>where in the last form, <var>asm-qualifiers</var> contains <code>goto</code> (and in the
first form, not).
</p>
<p>The <code>asm</code> keyword is a GNU extension.
When writing code that can be compiled with <samp>-ansi</samp> and the
various <samp>-std</samp> options, use <code>__asm__</code> instead of
<code>asm</code> (see <a href="Alternate-Keywords.html#Alternate-Keywords">Alternate Keywords</a>).
</p>
<a name="Qualifiers-2"></a>
<h4 class="subsubheading">Qualifiers</h4>
<dl compact="compact">
<dt><code>volatile</code></dt>
<dd><p>The typical use of extended <code>asm</code> statements is to manipulate input
values to produce output values. However, your <code>asm</code> statements may
also produce side effects. If so, you may need to use the <code>volatile</code>
qualifier to disable certain optimizations. See <a href="#Volatile">Volatile</a>.
</p>
</dd>
<dt><code>inline</code></dt>
<dd><p>If you use the <code>inline</code> qualifier, then for inlining purposes the size
of the asm is taken as the smallest size possible (see <a href="Size-of-an-asm.html#Size-of-an-asm">Size of an asm</a>).
</p>
</dd>
<dt><code>goto</code></dt>
<dd><p>This qualifier informs the compiler that the <code>asm</code> statement may
perform a jump to one of the labels listed in the <var>GotoLabels</var>.
See <a href="#GotoLabels">GotoLabels</a>.
</p></dd>
</dl>
<a name="Parameters-1"></a>
<h4 class="subsubheading">Parameters</h4>
<dl compact="compact">
<dt><var>AssemblerTemplate</var></dt>
<dd><p>This is a literal string that is the template for the assembler code. It is a
combination of fixed text and tokens that refer to the input, output,
and goto parameters. See <a href="#AssemblerTemplate">AssemblerTemplate</a>.
</p>
</dd>
<dt><var>OutputOperands</var></dt>
<dd><p>A comma-separated list of the C variables modified by the instructions in the
<var>AssemblerTemplate</var>. An empty list is permitted. See <a href="#OutputOperands">OutputOperands</a>.
</p>
</dd>
<dt><var>InputOperands</var></dt>
<dd><p>A comma-separated list of C expressions read by the instructions in the
<var>AssemblerTemplate</var>. An empty list is permitted. See <a href="#InputOperands">InputOperands</a>.
</p>
</dd>
<dt><var>Clobbers</var></dt>
<dd><p>A comma-separated list of registers or other values changed by the
<var>AssemblerTemplate</var>, beyond those listed as outputs.
An empty list is permitted. See <a href="#Clobbers-and-Scratch-Registers">Clobbers and Scratch Registers</a>.
</p>
</dd>
<dt><var>GotoLabels</var></dt>
<dd><p>When you are using the <code>goto</code> form of <code>asm</code>, this section contains
the list of all C labels to which the code in the
<var>AssemblerTemplate</var> may jump.
See <a href="#GotoLabels">GotoLabels</a>.
</p>
<p><code>asm</code> statements may not perform jumps into other <code>asm</code> statements,
only to the listed <var>GotoLabels</var>.
GCC&rsquo;s optimizers do not know about other jumps; therefore they cannot take
account of them when deciding how to optimize.
</p></dd>
</dl>
<p>The total number of input + output + goto operands is limited to 30.
</p>
<a name="Remarks-1"></a>
<h4 class="subsubheading">Remarks</h4>
<p>The <code>asm</code> statement allows you to include assembly instructions directly
within C code. This may help you to maximize performance in time-sensitive
code or to access assembly instructions that are not readily available to C
programs.
</p>
<p>Note that extended <code>asm</code> statements must be inside a function. Only
basic <code>asm</code> may be outside functions (see <a href="Basic-Asm.html#Basic-Asm">Basic Asm</a>).
Functions declared with the <code>naked</code> attribute also require basic
<code>asm</code> (see <a href="Function-Attributes.html#Function-Attributes">Function Attributes</a>).
</p>
<p>While the uses of <code>asm</code> are many and varied, it may help to think of an
<code>asm</code> statement as a series of low-level instructions that convert input
parameters to output parameters. So a simple (if not particularly useful)
example for i386 using <code>asm</code> might look like this:
</p>
<div class="example">
<pre class="example">int src = 1;
int dst;
asm (&quot;mov %1, %0\n\t&quot;
&quot;add $1, %0&quot;
: &quot;=r&quot; (dst)
: &quot;r&quot; (src));
printf(&quot;%d\n&quot;, dst);
</pre></div>
<p>This code copies <code>src</code> to <code>dst</code> and add 1 to <code>dst</code>.
</p>
<a name="Volatile"></a><a name="Volatile-1"></a>
<h4 class="subsubsection">6.45.2.1 Volatile</h4>
<a name="index-volatile-asm"></a>
<a name="index-asm-volatile"></a>
<p>GCC&rsquo;s optimizers sometimes discard <code>asm</code> statements if they determine
there is no need for the output variables. Also, the optimizers may move
code out of loops if they believe that the code will always return the same
result (i.e. none of its input values change between calls). Using the
<code>volatile</code> qualifier disables these optimizations. <code>asm</code> statements
that have no output operands, including <code>asm goto</code> statements,
are implicitly volatile.
</p>
<p>This i386 code demonstrates a case that does not use (or require) the
<code>volatile</code> qualifier. If it is performing assertion checking, this code
uses <code>asm</code> to perform the validation. Otherwise, <code>dwRes</code> is
unreferenced by any code. As a result, the optimizers can discard the
<code>asm</code> statement, which in turn removes the need for the entire
<code>DoCheck</code> routine. By omitting the <code>volatile</code> qualifier when it
isn&rsquo;t needed you allow the optimizers to produce the most efficient code
possible.
</p>
<div class="example">
<pre class="example">void DoCheck(uint32_t dwSomeValue)
{
uint32_t dwRes;
// Assumes dwSomeValue is not zero.
asm (&quot;bsfl %1,%0&quot;
: &quot;=r&quot; (dwRes)
: &quot;r&quot; (dwSomeValue)
: &quot;cc&quot;);
assert(dwRes &gt; 3);
}
</pre></div>
<p>The next example shows a case where the optimizers can recognize that the input
(<code>dwSomeValue</code>) never changes during the execution of the function and can
therefore move the <code>asm</code> outside the loop to produce more efficient code.
Again, using <code>volatile</code> disables this type of optimization.
</p>
<div class="example">
<pre class="example">void do_print(uint32_t dwSomeValue)
{
uint32_t dwRes;
for (uint32_t x=0; x &lt; 5; x++)
{
// Assumes dwSomeValue is not zero.
asm (&quot;bsfl %1,%0&quot;
: &quot;=r&quot; (dwRes)
: &quot;r&quot; (dwSomeValue)
: &quot;cc&quot;);
printf(&quot;%u: %u %u\n&quot;, x, dwSomeValue, dwRes);
}
}
</pre></div>
<p>The following example demonstrates a case where you need to use the
<code>volatile</code> qualifier.
It uses the x86 <code>rdtsc</code> instruction, which reads
the computer&rsquo;s time-stamp counter. Without the <code>volatile</code> qualifier,
the optimizers might assume that the <code>asm</code> block will always return the
same value and therefore optimize away the second call.
</p>
<div class="example">
<pre class="example">uint64_t msr;
asm volatile ( &quot;rdtsc\n\t&quot; // Returns the time in EDX:EAX.
&quot;shl $32, %%rdx\n\t&quot; // Shift the upper bits left.
&quot;or %%rdx, %0&quot; // 'Or' in the lower bits.
: &quot;=a&quot; (msr)
:
: &quot;rdx&quot;);
printf(&quot;msr: %llx\n&quot;, msr);
// Do other work...
// Reprint the timestamp
asm volatile ( &quot;rdtsc\n\t&quot; // Returns the time in EDX:EAX.
&quot;shl $32, %%rdx\n\t&quot; // Shift the upper bits left.
&quot;or %%rdx, %0&quot; // 'Or' in the lower bits.
: &quot;=a&quot; (msr)
:
: &quot;rdx&quot;);
printf(&quot;msr: %llx\n&quot;, msr);
</pre></div>
<p>GCC&rsquo;s optimizers do not treat this code like the non-volatile code in the
earlier examples. They do not move it out of loops or omit it on the
assumption that the result from a previous call is still valid.
</p>
<p>Note that the compiler can move even volatile <code>asm</code> instructions relative
to other code, including across jump instructions. For example, on many
targets there is a system register that controls the rounding mode of
floating-point operations. Setting it with a volatile <code>asm</code>, as in the
following PowerPC example, does not work reliably.
</p>
<div class="example">
<pre class="example">asm volatile(&quot;mtfsf 255, %0&quot; : : &quot;f&quot; (fpenv));
sum = x + y;
</pre></div>
<p>The compiler may move the addition back before the volatile <code>asm</code>. To
make it work as expected, add an artificial dependency to the <code>asm</code> by
referencing a variable in the subsequent code, for example:
</p>
<div class="example">
<pre class="example">asm volatile (&quot;mtfsf 255,%1&quot; : &quot;=X&quot; (sum) : &quot;f&quot; (fpenv));
sum = x + y;
</pre></div>
<p>Under certain circumstances, GCC may duplicate (or remove duplicates of) your
assembly code when optimizing. This can lead to unexpected duplicate symbol
errors during compilation if your asm code defines symbols or labels.
Using &lsquo;<samp>%=</samp>&rsquo;
(see <a href="#AssemblerTemplate">AssemblerTemplate</a>) may help resolve this problem.
</p>
<a name="AssemblerTemplate"></a><a name="Assembler-Template"></a>
<h4 class="subsubsection">6.45.2.2 Assembler Template</h4>
<a name="index-asm-assembler-template"></a>
<p>An assembler template is a literal string containing assembler instructions.
The compiler replaces tokens in the template that refer
to inputs, outputs, and goto labels,
and then outputs the resulting string to the assembler. The
string can contain any instructions recognized by the assembler, including
directives. GCC does not parse the assembler instructions
themselves and does not know what they mean or even whether they are valid
assembler input. However, it does count the statements
(see <a href="Size-of-an-asm.html#Size-of-an-asm">Size of an asm</a>).
</p>
<p>You may place multiple assembler instructions together in a single <code>asm</code>
string, separated by the characters normally used in assembly code for the
system. A combination that works in most places is a newline to break the
line, plus a tab character to move to the instruction field (written as
&lsquo;<samp>\n\t</samp>&rsquo;).
Some assemblers allow semicolons as a line separator. However, note
that some assembler dialects use semicolons to start a comment.
</p>
<p>Do not expect a sequence of <code>asm</code> statements to remain perfectly
consecutive after compilation, even when you are using the <code>volatile</code>
qualifier. If certain instructions need to remain consecutive in the output,
put them in a single multi-instruction asm statement.
</p>
<p>Accessing data from C programs without using input/output operands (such as
by using global symbols directly from the assembler template) may not work as
expected. Similarly, calling functions directly from an assembler template
requires a detailed understanding of the target assembler and ABI.
</p>
<p>Since GCC does not parse the assembler template,
it has no visibility of any
symbols it references. This may result in GCC discarding those symbols as
unreferenced unless they are also listed as input, output, or goto operands.
</p>
<a name="Special-format-strings"></a>
<h4 class="subsubheading">Special format strings</h4>
<p>In addition to the tokens described by the input, output, and goto operands,
these tokens have special meanings in the assembler template:
</p>
<dl compact="compact">
<dt>&lsquo;<samp>%%</samp>&rsquo;</dt>
<dd><p>Outputs a single &lsquo;<samp>%</samp>&rsquo; into the assembler code.
</p>
</dd>
<dt>&lsquo;<samp>%=</samp>&rsquo;</dt>
<dd><p>Outputs a number that is unique to each instance of the <code>asm</code>
statement in the entire compilation. This option is useful when creating local
labels and referring to them multiple times in a single template that
generates multiple assembler instructions.
</p>
</dd>
<dt>&lsquo;<samp>%{</samp>&rsquo;</dt>
<dt>&lsquo;<samp>%|</samp>&rsquo;</dt>
<dt>&lsquo;<samp>%}</samp>&rsquo;</dt>
<dd><p>Outputs &lsquo;<samp>{</samp>&rsquo;, &lsquo;<samp>|</samp>&rsquo;, and &lsquo;<samp>}</samp>&rsquo; characters (respectively)
into the assembler code. When unescaped, these characters have special
meaning to indicate multiple assembler dialects, as described below.
</p></dd>
</dl>
<a name="Multiple-assembler-dialects-in-asm-templates"></a>
<h4 class="subsubheading">Multiple assembler dialects in <code>asm</code> templates</h4>
<p>On targets such as x86, GCC supports multiple assembler dialects.
The <samp>-masm</samp> option controls which dialect GCC uses as its
default for inline assembler. The target-specific documentation for the
<samp>-masm</samp> option contains the list of supported dialects, as well as the
default dialect if the option is not specified. This information may be
important to understand, since assembler code that works correctly when
compiled using one dialect will likely fail if compiled using another.
See <a href="x86-Options.html#x86-Options">x86 Options</a>.
</p>
<p>If your code needs to support multiple assembler dialects (for example, if
you are writing public headers that need to support a variety of compilation
options), use constructs of this form:
</p>
<div class="example">
<pre class="example">{ dialect0 | dialect1 | dialect2... }
</pre></div>
<p>This construct outputs <code>dialect0</code>
when using dialect #0 to compile the code,
<code>dialect1</code> for dialect #1, etc. If there are fewer alternatives within the
braces than the number of dialects the compiler supports, the construct
outputs nothing.
</p>
<p>For example, if an x86 compiler supports two dialects
(&lsquo;<samp>att</samp>&rsquo;, &lsquo;<samp>intel</samp>&rsquo;), an
assembler template such as this:
</p>
<div class="example">
<pre class="example">&quot;bt{l %[Offset],%[Base] | %[Base],%[Offset]}; jc %l2&quot;
</pre></div>
<p>is equivalent to one of
</p>
<div class="example">
<pre class="example">&quot;btl %[Offset],%[Base] ; jc %l2&quot; <span class="roman">/* att dialect */</span>
&quot;bt %[Base],%[Offset]; jc %l2&quot; <span class="roman">/* intel dialect */</span>
</pre></div>
<p>Using that same compiler, this code:
</p>
<div class="example">
<pre class="example">&quot;xchg{l}\t{%%}ebx, %1&quot;
</pre></div>
<p>corresponds to either
</p>
<div class="example">
<pre class="example">&quot;xchgl\t%%ebx, %1&quot; <span class="roman">/* att dialect */</span>
&quot;xchg\tebx, %1&quot; <span class="roman">/* intel dialect */</span>
</pre></div>
<p>There is no support for nesting dialect alternatives.
</p>
<a name="OutputOperands"></a><a name="Output-Operands"></a>
<h4 class="subsubsection">6.45.2.3 Output Operands</h4>
<a name="index-asm-output-operands"></a>
<p>An <code>asm</code> statement has zero or more output operands indicating the names
of C variables modified by the assembler code.
</p>
<p>In this i386 example, <code>old</code> (referred to in the template string as
<code>%0</code>) and <code>*Base</code> (as <code>%1</code>) are outputs and <code>Offset</code>
(<code>%2</code>) is an input:
</p>
<div class="example">
<pre class="example">bool old;
__asm__ (&quot;btsl %2,%1\n\t&quot; // Turn on zero-based bit #Offset in Base.
&quot;sbb %0,%0&quot; // Use the CF to calculate old.
: &quot;=r&quot; (old), &quot;+rm&quot; (*Base)
: &quot;Ir&quot; (Offset)
: &quot;cc&quot;);
return old;
</pre></div>
<p>Operands are separated by commas. Each operand has this format:
</p>
<div class="example">
<pre class="example"><span class="roman">[</span> [<var>asmSymbolicName</var>] <span class="roman">]</span> <var>constraint</var> (<var>cvariablename</var>)
</pre></div>
<dl compact="compact">
<dt><var>asmSymbolicName</var></dt>
<dd><p>Specifies a symbolic name for the operand.
Reference the name in the assembler template
by enclosing it in square brackets
(i.e. &lsquo;<samp>%[Value]</samp>&rsquo;). The scope of the name is the <code>asm</code> statement
that contains the definition. Any valid C variable name is acceptable,
including names already defined in the surrounding code. No two operands
within the same <code>asm</code> statement can use the same symbolic name.
</p>
<p>When not using an <var>asmSymbolicName</var>, use the (zero-based) position
of the operand
in the list of operands in the assembler template. For example if there are
three output operands, use &lsquo;<samp>%0</samp>&rsquo; in the template to refer to the first,
&lsquo;<samp>%1</samp>&rsquo; for the second, and &lsquo;<samp>%2</samp>&rsquo; for the third.
</p>
</dd>
<dt><var>constraint</var></dt>
<dd><p>A string constant specifying constraints on the placement of the operand;
See <a href="Constraints.html#Constraints">Constraints</a>, for details.
</p>
<p>Output constraints must begin with either &lsquo;<samp>=</samp>&rsquo; (a variable overwriting an
existing value) or &lsquo;<samp>+</samp>&rsquo; (when reading and writing). When using
&lsquo;<samp>=</samp>&rsquo;, do not assume the location contains the existing value
on entry to the <code>asm</code>, except
when the operand is tied to an input; see <a href="#InputOperands">Input Operands</a>.
</p>
<p>After the prefix, there must be one or more additional constraints
(see <a href="Constraints.html#Constraints">Constraints</a>) that describe where the value resides. Common
constraints include &lsquo;<samp>r</samp>&rsquo; for register and &lsquo;<samp>m</samp>&rsquo; for memory.
When you list more than one possible location (for example, <code>&quot;=rm&quot;</code>),
the compiler chooses the most efficient one based on the current context.
If you list as many alternates as the <code>asm</code> statement allows, you permit
the optimizers to produce the best possible code.
If you must use a specific register, but your Machine Constraints do not
provide sufficient control to select the specific register you want,
local register variables may provide a solution (see <a href="Local-Register-Variables.html#Local-Register-Variables">Local Register Variables</a>).
</p>
</dd>
<dt><var>cvariablename</var></dt>
<dd><p>Specifies a C lvalue expression to hold the output, typically a variable name.
The enclosing parentheses are a required part of the syntax.
</p>
</dd>
</dl>
<p>When the compiler selects the registers to use to
represent the output operands, it does not use any of the clobbered registers
(see <a href="#Clobbers-and-Scratch-Registers">Clobbers and Scratch Registers</a>).
</p>
<p>Output operand expressions must be lvalues. The compiler cannot check whether
the operands have data types that are reasonable for the instruction being
executed. For output expressions that are not directly addressable (for
example a bit-field), the constraint must allow a register. In that case, GCC
uses the register as the output of the <code>asm</code>, and then stores that
register into the output.
</p>
<p>Operands using the &lsquo;<samp>+</samp>&rsquo; constraint modifier count as two operands
(that is, both as input and output) towards the total maximum of 30 operands
per <code>asm</code> statement.
</p>
<p>Use the &lsquo;<samp>&amp;</samp>&rsquo; constraint modifier (see <a href="Modifiers.html#Modifiers">Modifiers</a>) on all output
operands that must not overlap an input. Otherwise,
GCC may allocate the output operand in the same register as an unrelated
input operand, on the assumption that the assembler code consumes its
inputs before producing outputs. This assumption may be false if the assembler
code actually consists of more than one instruction.
</p>
<p>The same problem can occur if one output parameter (<var>a</var>) allows a register
constraint and another output parameter (<var>b</var>) allows a memory constraint.
The code generated by GCC to access the memory address in <var>b</var> can contain
registers which <em>might</em> be shared by <var>a</var>, and GCC considers those
registers to be inputs to the asm. As above, GCC assumes that such input
registers are consumed before any outputs are written. This assumption may
result in incorrect behavior if the asm writes to <var>a</var> before using
<var>b</var>. Combining the &lsquo;<samp>&amp;</samp>&rsquo; modifier with the register constraint on <var>a</var>
ensures that modifying <var>a</var> does not affect the address referenced by
<var>b</var>. Otherwise, the location of <var>b</var>
is undefined if <var>a</var> is modified before using <var>b</var>.
</p>
<p><code>asm</code> supports operand modifiers on operands (for example &lsquo;<samp>%k2</samp>&rsquo;
instead of simply &lsquo;<samp>%2</samp>&rsquo;). Typically these qualifiers are hardware
dependent. The list of supported modifiers for x86 is found at
<a href="#x86Operandmodifiers">x86 Operand modifiers</a>.
</p>
<p>If the C code that follows the <code>asm</code> makes no use of any of the output
operands, use <code>volatile</code> for the <code>asm</code> statement to prevent the
optimizers from discarding the <code>asm</code> statement as unneeded
(see <a href="#Volatile">Volatile</a>).
</p>
<p>This code makes no use of the optional <var>asmSymbolicName</var>. Therefore it
references the first output operand as <code>%0</code> (were there a second, it
would be <code>%1</code>, etc). The number of the first input operand is one greater
than that of the last output operand. In this i386 example, that makes
<code>Mask</code> referenced as <code>%1</code>:
</p>
<div class="example">
<pre class="example">uint32_t Mask = 1234;
uint32_t Index;
asm (&quot;bsfl %1, %0&quot;
: &quot;=r&quot; (Index)
: &quot;r&quot; (Mask)
: &quot;cc&quot;);
</pre></div>
<p>That code overwrites the variable <code>Index</code> (&lsquo;<samp>=</samp>&rsquo;),
placing the value in a register (&lsquo;<samp>r</samp>&rsquo;).
Using the generic &lsquo;<samp>r</samp>&rsquo; constraint instead of a constraint for a specific
register allows the compiler to pick the register to use, which can result
in more efficient code. This may not be possible if an assembler instruction
requires a specific register.
</p>
<p>The following i386 example uses the <var>asmSymbolicName</var> syntax.
It produces the
same result as the code above, but some may consider it more readable or more
maintainable since reordering index numbers is not necessary when adding or
removing operands. The names <code>aIndex</code> and <code>aMask</code>
are only used in this example to emphasize which
names get used where.
It is acceptable to reuse the names <code>Index</code> and <code>Mask</code>.
</p>
<div class="example">
<pre class="example">uint32_t Mask = 1234;
uint32_t Index;
asm (&quot;bsfl %[aMask], %[aIndex]&quot;
: [aIndex] &quot;=r&quot; (Index)
: [aMask] &quot;r&quot; (Mask)
: &quot;cc&quot;);
</pre></div>
<p>Here are some more examples of output operands.
</p>
<div class="example">
<pre class="example">uint32_t c = 1;
uint32_t d;
uint32_t *e = &amp;c;
asm (&quot;mov %[e], %[d]&quot;
: [d] &quot;=rm&quot; (d)
: [e] &quot;rm&quot; (*e));
</pre></div>
<p>Here, <code>d</code> may either be in a register or in memory. Since the compiler
might already have the current value of the <code>uint32_t</code> location
pointed to by <code>e</code>
in a register, you can enable it to choose the best location
for <code>d</code> by specifying both constraints.
</p>
<a name="FlagOutputOperands"></a><a name="Flag-Output-Operands"></a>
<h4 class="subsubsection">6.45.2.4 Flag Output Operands</h4>
<a name="index-asm-flag-output-operands"></a>
<p>Some targets have a special register that holds the &ldquo;flags&rdquo; for the
result of an operation or comparison. Normally, the contents of that
register are either unmodifed by the asm, or the asm is considered to
clobber the contents.
</p>
<p>On some targets, a special form of output operand exists by which
conditions in the flags register may be outputs of the asm. The set of
conditions supported are target specific, but the general rule is that
the output variable must be a scalar integer, and the value is boolean.
When supported, the target defines the preprocessor symbol
<code>__GCC_ASM_FLAG_OUTPUTS__</code>.
</p>
<p>Because of the special nature of the flag output operands, the constraint
may not include alternatives.
</p>
<p>Most often, the target has only one flags register, and thus is an implied
operand of many instructions. In this case, the operand should not be
referenced within the assembler template via <code>%0</code> etc, as there&rsquo;s
no corresponding text in the assembly language.
</p>
<dl compact="compact">
<dt>x86 family</dt>
<dd><p>The flag output constraints for the x86 family are of the form
&lsquo;<samp>=@cc<var>cond</var></samp>&rsquo; where <var>cond</var> is one of the standard
conditions defined in the ISA manual for <code>j<var>cc</var></code> or
<code>set<var>cc</var></code>.
</p>
<dl compact="compact">
<dt><code>a</code></dt>
<dd><p>&ldquo;above&rdquo; or unsigned greater than
</p></dd>
<dt><code>ae</code></dt>
<dd><p>&ldquo;above or equal&rdquo; or unsigned greater than or equal
</p></dd>
<dt><code>b</code></dt>
<dd><p>&ldquo;below&rdquo; or unsigned less than
</p></dd>
<dt><code>be</code></dt>
<dd><p>&ldquo;below or equal&rdquo; or unsigned less than or equal
</p></dd>
<dt><code>c</code></dt>
<dd><p>carry flag set
</p></dd>
<dt><code>e</code></dt>
<dt><code>z</code></dt>
<dd><p>&ldquo;equal&rdquo; or zero flag set
</p></dd>
<dt><code>g</code></dt>
<dd><p>signed greater than
</p></dd>
<dt><code>ge</code></dt>
<dd><p>signed greater than or equal
</p></dd>
<dt><code>l</code></dt>
<dd><p>signed less than
</p></dd>
<dt><code>le</code></dt>
<dd><p>signed less than or equal
</p></dd>
<dt><code>o</code></dt>
<dd><p>overflow flag set
</p></dd>
<dt><code>p</code></dt>
<dd><p>parity flag set
</p></dd>
<dt><code>s</code></dt>
<dd><p>sign flag set
</p></dd>
<dt><code>na</code></dt>
<dt><code>nae</code></dt>
<dt><code>nb</code></dt>
<dt><code>nbe</code></dt>
<dt><code>nc</code></dt>
<dt><code>ne</code></dt>
<dt><code>ng</code></dt>
<dt><code>nge</code></dt>
<dt><code>nl</code></dt>
<dt><code>nle</code></dt>
<dt><code>no</code></dt>
<dt><code>np</code></dt>
<dt><code>ns</code></dt>
<dt><code>nz</code></dt>
<dd><p>&ldquo;not&rdquo; <var>flag</var>, or inverted versions of those above
</p></dd>
</dl>
</dd>
</dl>
<a name="InputOperands"></a><a name="Input-Operands"></a>
<h4 class="subsubsection">6.45.2.5 Input Operands</h4>
<a name="index-asm-input-operands"></a>
<a name="index-asm-expressions"></a>
<p>Input operands make values from C variables and expressions available to the
assembly code.
</p>
<p>Operands are separated by commas. Each operand has this format:
</p>
<div class="example">
<pre class="example"><span class="roman">[</span> [<var>asmSymbolicName</var>] <span class="roman">]</span> <var>constraint</var> (<var>cexpression</var>)
</pre></div>
<dl compact="compact">
<dt><var>asmSymbolicName</var></dt>
<dd><p>Specifies a symbolic name for the operand.
Reference the name in the assembler template
by enclosing it in square brackets
(i.e. &lsquo;<samp>%[Value]</samp>&rsquo;). The scope of the name is the <code>asm</code> statement
that contains the definition. Any valid C variable name is acceptable,
including names already defined in the surrounding code. No two operands
within the same <code>asm</code> statement can use the same symbolic name.
</p>
<p>When not using an <var>asmSymbolicName</var>, use the (zero-based) position
of the operand
in the list of operands in the assembler template. For example if there are
two output operands and three inputs,
use &lsquo;<samp>%2</samp>&rsquo; in the template to refer to the first input operand,
&lsquo;<samp>%3</samp>&rsquo; for the second, and &lsquo;<samp>%4</samp>&rsquo; for the third.
</p>
</dd>
<dt><var>constraint</var></dt>
<dd><p>A string constant specifying constraints on the placement of the operand;
See <a href="Constraints.html#Constraints">Constraints</a>, for details.
</p>
<p>Input constraint strings may not begin with either &lsquo;<samp>=</samp>&rsquo; or &lsquo;<samp>+</samp>&rsquo;.
When you list more than one possible location (for example, &lsquo;<samp>&quot;irm&quot;</samp>&rsquo;),
the compiler chooses the most efficient one based on the current context.
If you must use a specific register, but your Machine Constraints do not
provide sufficient control to select the specific register you want,
local register variables may provide a solution (see <a href="Local-Register-Variables.html#Local-Register-Variables">Local Register Variables</a>).
</p>
<p>Input constraints can also be digits (for example, <code>&quot;0&quot;</code>). This indicates
that the specified input must be in the same place as the output constraint
at the (zero-based) index in the output constraint list.
When using <var>asmSymbolicName</var> syntax for the output operands,
you may use these names (enclosed in brackets &lsquo;<samp>[]</samp>&rsquo;) instead of digits.
</p>
</dd>
<dt><var>cexpression</var></dt>
<dd><p>This is the C variable or expression being passed to the <code>asm</code> statement
as input. The enclosing parentheses are a required part of the syntax.
</p>
</dd>
</dl>
<p>When the compiler selects the registers to use to represent the input
operands, it does not use any of the clobbered registers
(see <a href="#Clobbers-and-Scratch-Registers">Clobbers and Scratch Registers</a>).
</p>
<p>If there are no output operands but there are input operands, place two
consecutive colons where the output operands would go:
</p>
<div class="example">
<pre class="example">__asm__ (&quot;some instructions&quot;
: /* No outputs. */
: &quot;r&quot; (Offset / 8));
</pre></div>
<p><strong>Warning:</strong> Do <em>not</em> modify the contents of input-only operands
(except for inputs tied to outputs). The compiler assumes that on exit from
the <code>asm</code> statement these operands contain the same values as they
had before executing the statement.
It is <em>not</em> possible to use clobbers
to inform the compiler that the values in these inputs are changing. One
common work-around is to tie the changing input variable to an output variable
that never gets used. Note, however, that if the code that follows the
<code>asm</code> statement makes no use of any of the output operands, the GCC
optimizers may discard the <code>asm</code> statement as unneeded
(see <a href="#Volatile">Volatile</a>).
</p>
<p><code>asm</code> supports operand modifiers on operands (for example &lsquo;<samp>%k2</samp>&rsquo;
instead of simply &lsquo;<samp>%2</samp>&rsquo;). Typically these qualifiers are hardware
dependent. The list of supported modifiers for x86 is found at
<a href="#x86Operandmodifiers">x86 Operand modifiers</a>.
</p>
<p>In this example using the fictitious <code>combine</code> instruction, the
constraint <code>&quot;0&quot;</code> for input operand 1 says that it must occupy the same
location as output operand 0. Only input operands may use numbers in
constraints, and they must each refer to an output operand. Only a number (or
the symbolic assembler name) in the constraint can guarantee that one operand
is in the same place as another. The mere fact that <code>foo</code> is the value of
both operands is not enough to guarantee that they are in the same place in
the generated assembler code.
</p>
<div class="example">
<pre class="example">asm (&quot;combine %2, %0&quot;
: &quot;=r&quot; (foo)
: &quot;0&quot; (foo), &quot;g&quot; (bar));
</pre></div>
<p>Here is an example using symbolic names.
</p>
<div class="example">
<pre class="example">asm (&quot;cmoveq %1, %2, %[result]&quot;
: [result] &quot;=r&quot;(result)
: &quot;r&quot; (test), &quot;r&quot; (new), &quot;[result]&quot; (old));
</pre></div>
<a name="Clobbers-and-Scratch-Registers"></a><a name="Clobbers-and-Scratch-Registers-1"></a>
<h4 class="subsubsection">6.45.2.6 Clobbers and Scratch Registers</h4>
<a name="index-asm-clobbers"></a>
<a name="index-asm-scratch-registers"></a>
<p>While the compiler is aware of changes to entries listed in the output
operands, the inline <code>asm</code> code may modify more than just the outputs. For
example, calculations may require additional registers, or the processor may
overwrite a register as a side effect of a particular assembler instruction.
In order to inform the compiler of these changes, list them in the clobber
list. Clobber list items are either register names or the special clobbers
(listed below). Each clobber list item is a string constant
enclosed in double quotes and separated by commas.
</p>
<p>Clobber descriptions may not in any way overlap with an input or output
operand. For example, you may not have an operand describing a register class
with one member when listing that register in the clobber list. Variables
declared to live in specific registers (see <a href="Explicit-Register-Variables.html#Explicit-Register-Variables">Explicit Register Variables</a>) and used
as <code>asm</code> input or output operands must have no part mentioned in the
clobber description. In particular, there is no way to specify that input
operands get modified without also specifying them as output operands.
</p>
<p>When the compiler selects which registers to use to represent input and output
operands, it does not use any of the clobbered registers. As a result,
clobbered registers are available for any use in the assembler code.
</p>
<p>Here is a realistic example for the VAX showing the use of clobbered
registers:
</p>
<div class="example">
<pre class="example">asm volatile (&quot;movc3 %0, %1, %2&quot;
: /* No outputs. */
: &quot;g&quot; (from), &quot;g&quot; (to), &quot;g&quot; (count)
: &quot;r0&quot;, &quot;r1&quot;, &quot;r2&quot;, &quot;r3&quot;, &quot;r4&quot;, &quot;r5&quot;, &quot;memory&quot;);
</pre></div>
<p>Also, there are two special clobber arguments:
</p>
<dl compact="compact">
<dt><code>&quot;cc&quot;</code></dt>
<dd><p>The <code>&quot;cc&quot;</code> clobber indicates that the assembler code modifies the flags
register. On some machines, GCC represents the condition codes as a specific
hardware register; <code>&quot;cc&quot;</code> serves to name this register.
On other machines, condition code handling is different,
and specifying <code>&quot;cc&quot;</code> has no effect. But
it is valid no matter what the target.
</p>
</dd>
<dt><code>&quot;memory&quot;</code></dt>
<dd><p>The <code>&quot;memory&quot;</code> clobber tells the compiler that the assembly code
performs memory
reads or writes to items other than those listed in the input and output
operands (for example, accessing the memory pointed to by one of the input
parameters). To ensure memory contains correct values, GCC may need to flush
specific register values to memory before executing the <code>asm</code>. Further,
the compiler does not assume that any values read from memory before an
<code>asm</code> remain unchanged after that <code>asm</code>; it reloads them as
needed.
Using the <code>&quot;memory&quot;</code> clobber effectively forms a read/write
memory barrier for the compiler.
</p>
<p>Note that this clobber does not prevent the <em>processor</em> from doing
speculative reads past the <code>asm</code> statement. To prevent that, you need
processor-specific fence instructions.
</p>
</dd>
</dl>
<p>Flushing registers to memory has performance implications and may be
an issue for time-sensitive code. You can provide better information
to GCC to avoid this, as shown in the following examples. At a
minimum, aliasing rules allow GCC to know what memory <em>doesn&rsquo;t</em>
need to be flushed.
</p>
<p>Here is a fictitious sum of squares instruction, that takes two
pointers to floating point values in memory and produces a floating
point register output.
Notice that <code>x</code>, and <code>y</code> both appear twice in the <code>asm</code>
parameters, once to specify memory accessed, and once to specify a
base register used by the <code>asm</code>. You won&rsquo;t normally be wasting a
register by doing this as GCC can use the same register for both
purposes. However, it would be foolish to use both <code>%1</code> and
<code>%3</code> for <code>x</code> in this <code>asm</code> and expect them to be the
same. In fact, <code>%3</code> may well not be a register. It might be a
symbolic memory reference to the object pointed to by <code>x</code>.
</p>
<div class="smallexample">
<pre class="smallexample">asm (&quot;sumsq %0, %1, %2&quot;
: &quot;+f&quot; (result)
: &quot;r&quot; (x), &quot;r&quot; (y), &quot;m&quot; (*x), &quot;m&quot; (*y));
</pre></div>
<p>Here is a fictitious <code>*z++ = *x++ * *y++</code> instruction.
Notice that the <code>x</code>, <code>y</code> and <code>z</code> pointer registers
must be specified as input/output because the <code>asm</code> modifies
them.
</p>
<div class="smallexample">
<pre class="smallexample">asm (&quot;vecmul %0, %1, %2&quot;
: &quot;+r&quot; (z), &quot;+r&quot; (x), &quot;+r&quot; (y), &quot;=m&quot; (*z)
: &quot;m&quot; (*x), &quot;m&quot; (*y));
</pre></div>
<p>An x86 example where the string memory argument is of unknown length.
</p>
<div class="smallexample">
<pre class="smallexample">asm(&quot;repne scasb&quot;
: &quot;=c&quot; (count), &quot;+D&quot; (p)
: &quot;m&quot; (*(const char (*)[]) p), &quot;0&quot; (-1), &quot;a&quot; (0));
</pre></div>
<p>If you know the above will only be reading a ten byte array then you
could instead use a memory input like:
<code>&quot;m&quot; (*(const char (*)[10]) p)</code>.
</p>
<p>Here is an example of a PowerPC vector scale implemented in assembly,
complete with vector and condition code clobbers, and some initialized
offset registers that are unchanged by the <code>asm</code>.
</p>
<div class="smallexample">
<pre class="smallexample">void
dscal (size_t n, double *x, double alpha)
{
asm (&quot;/* lots of asm here */&quot;
: &quot;+m&quot; (*(double (*)[n]) x), &quot;+&amp;r&quot; (n), &quot;+b&quot; (x)
: &quot;d&quot; (alpha), &quot;b&quot; (32), &quot;b&quot; (48), &quot;b&quot; (64),
&quot;b&quot; (80), &quot;b&quot; (96), &quot;b&quot; (112)
: &quot;cr0&quot;,
&quot;vs32&quot;,&quot;vs33&quot;,&quot;vs34&quot;,&quot;vs35&quot;,&quot;vs36&quot;,&quot;vs37&quot;,&quot;vs38&quot;,&quot;vs39&quot;,
&quot;vs40&quot;,&quot;vs41&quot;,&quot;vs42&quot;,&quot;vs43&quot;,&quot;vs44&quot;,&quot;vs45&quot;,&quot;vs46&quot;,&quot;vs47&quot;);
}
</pre></div>
<p>Rather than allocating fixed registers via clobbers to provide scratch
registers for an <code>asm</code> statement, an alternative is to define a
variable and make it an early-clobber output as with <code>a2</code> and
<code>a3</code> in the example below. This gives the compiler register
allocator more freedom. You can also define a variable and make it an
output tied to an input as with <code>a0</code> and <code>a1</code>, tied
respectively to <code>ap</code> and <code>lda</code>. Of course, with tied
outputs your <code>asm</code> can&rsquo;t use the input value after modifying the
output register since they are one and the same register. What&rsquo;s
more, if you omit the early-clobber on the output, it is possible that
GCC might allocate the same register to another of the inputs if GCC
could prove they had the same value on entry to the <code>asm</code>. This
is why <code>a1</code> has an early-clobber. Its tied input, <code>lda</code>
might conceivably be known to have the value 16 and without an
early-clobber share the same register as <code>%11</code>. On the other
hand, <code>ap</code> can&rsquo;t be the same as any of the other inputs, so an
early-clobber on <code>a0</code> is not needed. It is also not desirable in
this case. An early-clobber on <code>a0</code> would cause GCC to allocate
a separate register for the <code>&quot;m&quot; (*(const double (*)[]) ap)</code>
input. Note that tying an input to an output is the way to set up an
initialized temporary register modified by an <code>asm</code> statement.
An input not tied to an output is assumed by GCC to be unchanged, for
example <code>&quot;b&quot; (16)</code> below sets up <code>%11</code> to 16, and GCC might
use that register in following code if the value 16 happened to be
needed. You can even use a normal <code>asm</code> output for a scratch if
all inputs that might share the same register are consumed before the
scratch is used. The VSX registers clobbered by the <code>asm</code>
statement could have used this technique except for GCC&rsquo;s limit on the
number of <code>asm</code> parameters.
</p>
<div class="smallexample">
<pre class="smallexample">static void
dgemv_kernel_4x4 (long n, const double *ap, long lda,
const double *x, double *y, double alpha)
{
double *a0;
double *a1;
double *a2;
double *a3;
__asm__
(
/* lots of asm here */
&quot;#n=%1 ap=%8=%12 lda=%13 x=%7=%10 y=%0=%2 alpha=%9 o16=%11\n&quot;
&quot;#a0=%3 a1=%4 a2=%5 a3=%6&quot;
:
&quot;+m&quot; (*(double (*)[n]) y),
&quot;+&amp;r&quot; (n), // 1
&quot;+b&quot; (y), // 2
&quot;=b&quot; (a0), // 3
&quot;=&amp;b&quot; (a1), // 4
&quot;=&amp;b&quot; (a2), // 5
&quot;=&amp;b&quot; (a3) // 6
:
&quot;m&quot; (*(const double (*)[n]) x),
&quot;m&quot; (*(const double (*)[]) ap),
&quot;d&quot; (alpha), // 9
&quot;r&quot; (x), // 10
&quot;b&quot; (16), // 11
&quot;3&quot; (ap), // 12
&quot;4&quot; (lda) // 13
:
&quot;cr0&quot;,
&quot;vs32&quot;,&quot;vs33&quot;,&quot;vs34&quot;,&quot;vs35&quot;,&quot;vs36&quot;,&quot;vs37&quot;,
&quot;vs40&quot;,&quot;vs41&quot;,&quot;vs42&quot;,&quot;vs43&quot;,&quot;vs44&quot;,&quot;vs45&quot;,&quot;vs46&quot;,&quot;vs47&quot;
);
}
</pre></div>
<a name="GotoLabels"></a><a name="Goto-Labels"></a>
<h4 class="subsubsection">6.45.2.7 Goto Labels</h4>
<a name="index-asm-goto-labels"></a>
<p><code>asm goto</code> allows assembly code to jump to one or more C labels. The
<var>GotoLabels</var> section in an <code>asm goto</code> statement contains
a comma-separated
list of all C labels to which the assembler code may jump. GCC assumes that
<code>asm</code> execution falls through to the next statement (if this is not the
case, consider using the <code>__builtin_unreachable</code> intrinsic after the
<code>asm</code> statement). Optimization of <code>asm goto</code> may be improved by
using the <code>hot</code> and <code>cold</code> label attributes (see <a href="Label-Attributes.html#Label-Attributes">Label Attributes</a>).
</p>
<p>An <code>asm goto</code> statement cannot have outputs.
This is due to an internal restriction of
the compiler: control transfer instructions cannot have outputs.
If the assembler code does modify anything, use the <code>&quot;memory&quot;</code> clobber
to force the
optimizers to flush all register values to memory and reload them if
necessary after the <code>asm</code> statement.
</p>
<p>Also note that an <code>asm goto</code> statement is always implicitly
considered volatile.
</p>
<p>To reference a label in the assembler template,
prefix it with &lsquo;<samp>%l</samp>&rsquo; (lowercase &lsquo;<samp>L</samp>&rsquo;) followed
by its (zero-based) position in <var>GotoLabels</var> plus the number of input
operands. For example, if the <code>asm</code> has three inputs and references two
labels, refer to the first label as &lsquo;<samp>%l3</samp>&rsquo; and the second as &lsquo;<samp>%l4</samp>&rsquo;).
</p>
<p>Alternately, you can reference labels using the actual C label name enclosed
in brackets. For example, to reference a label named <code>carry</code>, you can
use &lsquo;<samp>%l[carry]</samp>&rsquo;. The label must still be listed in the <var>GotoLabels</var>
section when using this approach.
</p>
<p>Here is an example of <code>asm goto</code> for i386:
</p>
<div class="example">
<pre class="example">asm goto (
&quot;btl %1, %0\n\t&quot;
&quot;jc %l2&quot;
: /* No outputs. */
: &quot;r&quot; (p1), &quot;r&quot; (p2)
: &quot;cc&quot;
: carry);
return 0;
carry:
return 1;
</pre></div>
<p>The following example shows an <code>asm goto</code> that uses a memory clobber.
</p>
<div class="example">
<pre class="example">int frob(int x)
{
int y;
asm goto (&quot;frob %%r5, %1; jc %l[error]; mov (%2), %%r5&quot;
: /* No outputs. */
: &quot;r&quot;(x), &quot;r&quot;(&amp;y)
: &quot;r5&quot;, &quot;memory&quot;
: error);
return y;
error:
return -1;
}
</pre></div>
<a name="x86Operandmodifiers"></a><a name="x86-Operand-Modifiers"></a>
<h4 class="subsubsection">6.45.2.8 x86 Operand Modifiers</h4>
<p>References to input, output, and goto operands in the assembler template
of extended <code>asm</code> statements can use
modifiers to affect the way the operands are formatted in
the code output to the assembler. For example, the
following code uses the &lsquo;<samp>h</samp>&rsquo; and &lsquo;<samp>b</samp>&rsquo; modifiers for x86:
</p>
<div class="example">
<pre class="example">uint16_t num;
asm volatile (&quot;xchg %h0, %b0&quot; : &quot;+a&quot; (num) );
</pre></div>
<p>These modifiers generate this assembler code:
</p>
<div class="example">
<pre class="example">xchg %ah, %al
</pre></div>
<p>The rest of this discussion uses the following code for illustrative purposes.
</p>
<div class="example">
<pre class="example">int main()
{
int iInt = 1;
top:
asm volatile goto (&quot;some assembler instructions here&quot;
: /* No outputs. */
: &quot;q&quot; (iInt), &quot;X&quot; (sizeof(unsigned char) + 1)
: /* No clobbers. */
: top);
}
</pre></div>
<p>With no modifiers, this is what the output from the operands would be for the
&lsquo;<samp>att</samp>&rsquo; and &lsquo;<samp>intel</samp>&rsquo; dialects of assembler:
</p>
<table>
<thead><tr><th>Operand</th><th>&lsquo;<samp>att</samp>&rsquo;</th><th>&lsquo;<samp>intel</samp>&rsquo;</th></tr></thead>
<tr><td><code>%0</code></td><td><code>%eax</code></td><td><code>eax</code></td></tr>
<tr><td><code>%1</code></td><td><code>$2</code></td><td><code>2</code></td></tr>
<tr><td><code>%2</code></td><td><code>$.L2</code></td><td><code>OFFSET FLAT:.L2</code></td></tr>
</table>
<p>The table below shows the list of supported modifiers and their effects.
</p>
<table>
<thead><tr><th>Modifier</th><th>Description</th><th>Operand</th><th>&lsquo;<samp>att</samp>&rsquo;</th><th>&lsquo;<samp>intel</samp>&rsquo;</th></tr></thead>
<tr><td><code>z</code></td><td>Print the opcode suffix for the size of the current integer operand (one of <code>b</code>/<code>w</code>/<code>l</code>/<code>q</code>).</td><td><code>%z0</code></td><td><code>l</code></td><td></td></tr>
<tr><td><code>b</code></td><td>Print the QImode name of the register.</td><td><code>%b0</code></td><td><code>%al</code></td><td><code>al</code></td></tr>
<tr><td><code>h</code></td><td>Print the QImode name for a &ldquo;high&rdquo; register.</td><td><code>%h0</code></td><td><code>%ah</code></td><td><code>ah</code></td></tr>
<tr><td><code>w</code></td><td>Print the HImode name of the register.</td><td><code>%w0</code></td><td><code>%ax</code></td><td><code>ax</code></td></tr>
<tr><td><code>k</code></td><td>Print the SImode name of the register.</td><td><code>%k0</code></td><td><code>%eax</code></td><td><code>eax</code></td></tr>
<tr><td><code>q</code></td><td>Print the DImode name of the register.</td><td><code>%q0</code></td><td><code>%rax</code></td><td><code>rax</code></td></tr>
<tr><td><code>l</code></td><td>Print the label name with no punctuation.</td><td><code>%l2</code></td><td><code>.L2</code></td><td><code>.L2</code></td></tr>
<tr><td><code>c</code></td><td>Require a constant operand and print the constant expression with no punctuation.</td><td><code>%c1</code></td><td><code>2</code></td><td><code>2</code></td></tr>
</table>
<p><code>V</code> is a special modifier which prints the name of the full integer
register without <code>%</code>.
</p>
<a name="x86floatingpointasmoperands"></a><a name="x86-Floating_002dPoint-asm-Operands"></a>
<h4 class="subsubsection">6.45.2.9 x86 Floating-Point <code>asm</code> Operands</h4>
<p>On x86 targets, there are several rules on the usage of stack-like registers
in the operands of an <code>asm</code>. These rules apply only to the operands
that are stack-like registers:
</p>
<ol>
<li> Given a set of input registers that die in an <code>asm</code>, it is
necessary to know which are implicitly popped by the <code>asm</code>, and
which must be explicitly popped by GCC.
<p>An input register that is implicitly popped by the <code>asm</code> must be
explicitly clobbered, unless it is constrained to match an
output operand.
</p>
</li><li> For any input register that is implicitly popped by an <code>asm</code>, it is
necessary to know how to adjust the stack to compensate for the pop.
If any non-popped input is closer to the top of the reg-stack than
the implicitly popped register, it would not be possible to know what the
stack looked like&mdash;it&rsquo;s not clear how the rest of the stack &ldquo;slides
up&rdquo;.
<p>All implicitly popped input registers must be closer to the top of
the reg-stack than any input that is not implicitly popped.
</p>
<p>It is possible that if an input dies in an <code>asm</code>, the compiler might
use the input register for an output reload. Consider this example:
</p>
<div class="smallexample">
<pre class="smallexample">asm (&quot;foo&quot; : &quot;=t&quot; (a) : &quot;f&quot; (b));
</pre></div>
<p>This code says that input <code>b</code> is not popped by the <code>asm</code>, and that
the <code>asm</code> pushes a result onto the reg-stack, i.e., the stack is one
deeper after the <code>asm</code> than it was before. But, it is possible that
reload may think that it can use the same register for both the input and
the output.
</p>
<p>To prevent this from happening,
if any input operand uses the &lsquo;<samp>f</samp>&rsquo; constraint, all output register
constraints must use the &lsquo;<samp>&amp;</samp>&rsquo; early-clobber modifier.
</p>
<p>The example above is correctly written as:
</p>
<div class="smallexample">
<pre class="smallexample">asm (&quot;foo&quot; : &quot;=&amp;t&quot; (a) : &quot;f&quot; (b));
</pre></div>
</li><li> Some operands need to be in particular places on the stack. All
output operands fall in this category&mdash;GCC has no other way to
know which registers the outputs appear in unless you indicate
this in the constraints.
<p>Output operands must specifically indicate which register an output
appears in after an <code>asm</code>. &lsquo;<samp>=f</samp>&rsquo; is not allowed: the operand
constraints must select a class with a single register.
</p>
</li><li> Output operands may not be &ldquo;inserted&rdquo; between existing stack registers.
Since no 387 opcode uses a read/write operand, all output operands
are dead before the <code>asm</code>, and are pushed by the <code>asm</code>.
It makes no sense to push anywhere but the top of the reg-stack.
<p>Output operands must start at the top of the reg-stack: output
operands may not &ldquo;skip&rdquo; a register.
</p>
</li><li> Some <code>asm</code> statements may need extra stack space for internal
calculations. This can be guaranteed by clobbering stack registers
unrelated to the inputs and outputs.
</li></ol>
<p>This <code>asm</code>
takes one input, which is internally popped, and produces two outputs.
</p>
<div class="smallexample">
<pre class="smallexample">asm (&quot;fsincos&quot; : &quot;=t&quot; (cos), &quot;=u&quot; (sin) : &quot;0&quot; (inp));
</pre></div>
<p>This <code>asm</code> takes two inputs, which are popped by the <code>fyl2xp1</code> opcode,
and replaces them with one output. The <code>st(1)</code> clobber is necessary
for the compiler to know that <code>fyl2xp1</code> pops both inputs.
</p>
<div class="smallexample">
<pre class="smallexample">asm (&quot;fyl2xp1&quot; : &quot;=t&quot; (result) : &quot;0&quot; (x), &quot;u&quot; (y) : &quot;st(1)&quot;);
</pre></div>
<hr>
<div class="header">
<p>
Next: <a href="Constraints.html#Constraints" accesskey="n" rel="next">Constraints</a>, Previous: <a href="Basic-Asm.html#Basic-Asm" accesskey="p" rel="prev">Basic Asm</a>, Up: <a href="Using-Assembly-Language-with-C.html#Using-Assembly-Language-with-C" accesskey="u" rel="up">Using Assembly Language with C</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Option-Index.html#Option-Index" title="Index" rel="index">Index</a>]</p>
</div>
</body>
</html>