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.

2957 lines
146 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>Standard Names (GNU Compiler Collection (GCC) Internals)</title>
<meta name="description" content="Standard Names (GNU Compiler Collection (GCC) Internals)">
<meta name="keywords" content="Standard Names (GNU Compiler Collection (GCC) Internals)">
<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="Machine-Desc.html#Machine-Desc" rel="up" title="Machine Desc">
<link href="Pattern-Ordering.html#Pattern-Ordering" rel="next" title="Pattern Ordering">
<link href="C-Constraint-Interface.html#C-Constraint-Interface" rel="prev" title="C Constraint Interface">
<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="Standard-Names"></a>
<div class="header">
<p>
Next: <a href="Pattern-Ordering.html#Pattern-Ordering" accesskey="n" rel="next">Pattern Ordering</a>, Previous: <a href="Constraints.html#Constraints" accesskey="p" rel="prev">Constraints</a>, Up: <a href="Machine-Desc.html#Machine-Desc" accesskey="u" rel="up">Machine Desc</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="Standard-Pattern-Names-For-Generation"></a>
<h3 class="section">17.9 Standard Pattern Names For Generation</h3>
<a name="index-standard-pattern-names"></a>
<a name="index-pattern-names"></a>
<a name="index-names_002c-pattern"></a>
<p>Here is a table of the instruction names that are meaningful in the RTL
generation pass of the compiler. Giving one of these names to an
instruction pattern tells the RTL generation pass that it can use the
pattern to accomplish a certain task.
</p>
<dl compact="compact">
<dd><a name="index-movm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>mov<var>m</var></samp>&rsquo;</dt>
<dd><p>Here <var>m</var> stands for a two-letter machine mode name, in lowercase.
This instruction pattern moves data with that machine mode from operand
1 to operand 0. For example, &lsquo;<samp>movsi</samp>&rsquo; moves full-word data.
</p>
<p>If operand 0 is a <code>subreg</code> with mode <var>m</var> of a register whose
own mode is wider than <var>m</var>, the effect of this instruction is
to store the specified value in the part of the register that corresponds
to mode <var>m</var>. Bits outside of <var>m</var>, but which are within the
same target word as the <code>subreg</code> are undefined. Bits which are
outside the target word are left unchanged.
</p>
<p>This class of patterns is special in several ways. First of all, each
of these names up to and including full word size <em>must</em> be defined,
because there is no other way to copy a datum from one place to another.
If there are patterns accepting operands in larger modes,
&lsquo;<samp>mov<var>m</var></samp>&rsquo; must be defined for integer modes of those sizes.
</p>
<p>Second, these patterns are not used solely in the RTL generation pass.
Even the reload pass can generate move insns to copy values from stack
slots into temporary registers. When it does so, one of the operands is
a hard register and the other is an operand that can need to be reloaded
into a register.
</p>
<a name="index-force_005freg"></a>
<p>Therefore, when given such a pair of operands, the pattern must generate
RTL which needs no reloading and needs no temporary registers&mdash;no
registers other than the operands. For example, if you support the
pattern with a <code>define_expand</code>, then in such a case the
<code>define_expand</code> mustn&rsquo;t call <code>force_reg</code> or any other such
function which might generate new pseudo registers.
</p>
<p>This requirement exists even for subword modes on a RISC machine where
fetching those modes from memory normally requires several insns and
some temporary registers.
</p>
<a name="index-change_005faddress"></a>
<p>During reload a memory reference with an invalid address may be passed
as an operand. Such an address will be replaced with a valid address
later in the reload pass. In this case, nothing may be done with the
address except to use it as it stands. If it is copied, it will not be
replaced with a valid address. No attempt should be made to make such
an address into a valid address and no routine (such as
<code>change_address</code>) that will do so may be called. Note that
<code>general_operand</code> will fail when applied to such an address.
</p>
<a name="index-reload_005fin_005fprogress"></a>
<p>The global variable <code>reload_in_progress</code> (which must be explicitly
declared if required) can be used to determine whether such special
handling is required.
</p>
<p>The variety of operands that have reloads depends on the rest of the
machine description, but typically on a RISC machine these can only be
pseudo registers that did not get hard registers, while on other
machines explicit memory references will get optional reloads.
</p>
<p>If a scratch register is required to move an object to or from memory,
it can be allocated using <code>gen_reg_rtx</code> prior to life analysis.
</p>
<p>If there are cases which need scratch registers during or after reload,
you must provide an appropriate secondary_reload target hook.
</p>
<a name="index-can_005fcreate_005fpseudo_005fp"></a>
<p>The macro <code>can_create_pseudo_p</code> can be used to determine if it
is unsafe to create new pseudo registers. If this variable is nonzero, then
it is unsafe to call <code>gen_reg_rtx</code> to allocate a new pseudo.
</p>
<p>The constraints on a &lsquo;<samp>mov<var>m</var></samp>&rsquo; must permit moving any hard
register to any other hard register provided that
<code>TARGET_HARD_REGNO_MODE_OK</code> permits mode <var>m</var> in both registers and
<code>TARGET_REGISTER_MOVE_COST</code> applied to their classes returns a value
of 2.
</p>
<p>It is obligatory to support floating point &lsquo;<samp>mov<var>m</var></samp>&rsquo;
instructions into and out of any registers that can hold fixed point
values, because unions and structures (which have modes <code>SImode</code> or
<code>DImode</code>) can be in those registers and they may have floating
point members.
</p>
<p>There may also be a need to support fixed point &lsquo;<samp>mov<var>m</var></samp>&rsquo;
instructions in and out of floating point registers. Unfortunately, I
have forgotten why this was so, and I don&rsquo;t know whether it is still
true. If <code>TARGET_HARD_REGNO_MODE_OK</code> rejects fixed point values in
floating point registers, then the constraints of the fixed point
&lsquo;<samp>mov<var>m</var></samp>&rsquo; instructions must be designed to avoid ever trying to
reload into a floating point register.
</p>
<a name="index-reload_005fin-instruction-pattern"></a>
<a name="index-reload_005fout-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>reload_in<var>m</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>reload_out<var>m</var></samp>&rsquo;</dt>
<dd><p>These named patterns have been obsoleted by the target hook
<code>secondary_reload</code>.
</p>
<p>Like &lsquo;<samp>mov<var>m</var></samp>&rsquo;, but used when a scratch register is required to
move between operand 0 and operand 1. Operand 2 describes the scratch
register. See the discussion of the <code>SECONDARY_RELOAD_CLASS</code>
macro in see <a href="Register-Classes.html#Register-Classes">Register Classes</a>.
</p>
<p>There are special restrictions on the form of the <code>match_operand</code>s
used in these patterns. First, only the predicate for the reload
operand is examined, i.e., <code>reload_in</code> examines operand 1, but not
the predicates for operand 0 or 2. Second, there may be only one
alternative in the constraints. Third, only a single register class
letter may be used for the constraint; subsequent constraint letters
are ignored. As a special exception, an empty constraint string
matches the <code>ALL_REGS</code> register class. This may relieve ports
of the burden of defining an <code>ALL_REGS</code> constraint letter just
for these patterns.
</p>
<a name="index-movstrictm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>movstrict<var>m</var></samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>mov<var>m</var></samp>&rsquo; except that if operand 0 is a <code>subreg</code>
with mode <var>m</var> of a register whose natural mode is wider,
the &lsquo;<samp>movstrict<var>m</var></samp>&rsquo; instruction is guaranteed not to alter
any of the register except the part which belongs to mode <var>m</var>.
</p>
<a name="index-movmisalignm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>movmisalign<var>m</var></samp>&rsquo;</dt>
<dd><p>This variant of a move pattern is designed to load or store a value
from a memory address that is not naturally aligned for its mode.
For a store, the memory will be in operand 0; for a load, the memory
will be in operand 1. The other operand is guaranteed not to be a
memory, so that it&rsquo;s easy to tell whether this is a load or store.
</p>
<p>This pattern is used by the autovectorizer, and when expanding a
<code>MISALIGNED_INDIRECT_REF</code> expression.
</p>
<a name="index-load_005fmultiple-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>load_multiple</samp>&rsquo;</dt>
<dd><p>Load several consecutive memory locations into consecutive registers.
Operand 0 is the first of the consecutive registers, operand 1
is the first memory location, and operand 2 is a constant: the
number of consecutive registers.
</p>
<p>Define this only if the target machine really has such an instruction;
do not define this if the most efficient way of loading consecutive
registers from memory is to do them one at a time.
</p>
<p>On some machines, there are restrictions as to which consecutive
registers can be stored into memory, such as particular starting or
ending register numbers or only a range of valid counts. For those
machines, use a <code>define_expand</code> (see <a href="Expander-Definitions.html#Expander-Definitions">Expander Definitions</a>)
and make the pattern fail if the restrictions are not met.
</p>
<p>Write the generated insn as a <code>parallel</code> with elements being a
<code>set</code> of one register from the appropriate memory location (you may
also need <code>use</code> or <code>clobber</code> elements). Use a
<code>match_parallel</code> (see <a href="RTL-Template.html#RTL-Template">RTL Template</a>) to recognize the insn. See
<samp>rs6000.md</samp> for examples of the use of this insn pattern.
</p>
<a name="index-store_005fmultiple-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>store_multiple</samp>&rsquo;</dt>
<dd><p>Similar to &lsquo;<samp>load_multiple</samp>&rsquo;, but store several consecutive registers
into consecutive memory locations. Operand 0 is the first of the
consecutive memory locations, operand 1 is the first register, and
operand 2 is a constant: the number of consecutive registers.
</p>
<a name="index-vec_005fload_005flanesmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_load_lanes<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Perform an interleaved load of several vectors from memory operand 1
into register operand 0. Both operands have mode <var>m</var>. The register
operand is viewed as holding consecutive vectors of mode <var>n</var>,
while the memory operand is a flat array that contains the same number
of elements. The operation is equivalent to:
</p>
<div class="smallexample">
<pre class="smallexample">int c = GET_MODE_SIZE (<var>m</var>) / GET_MODE_SIZE (<var>n</var>);
for (j = 0; j &lt; GET_MODE_NUNITS (<var>n</var>); j++)
for (i = 0; i &lt; c; i++)
operand0[i][j] = operand1[j * c + i];
</pre></div>
<p>For example, &lsquo;<samp>vec_load_lanestiv4hi</samp>&rsquo; loads 8 16-bit values
from memory into a register of mode &lsquo;<samp>TI</samp>&rsquo;. The register
contains two consecutive vectors of mode &lsquo;<samp>V4HI</samp>&rsquo;.
</p>
<p>This pattern can only be used if:
</p><div class="smallexample">
<pre class="smallexample">TARGET_ARRAY_MODE_SUPPORTED_P (<var>n</var>, <var>c</var>)
</pre></div>
<p>is true. GCC assumes that, if a target supports this kind of
instruction for some mode <var>n</var>, it also supports unaligned
loads for vectors of mode <var>n</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-vec_005fmask_005fload_005flanesmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_mask_load_lanes<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>vec_load_lanes<var>m</var><var>n</var></samp>&rsquo;, but takes an additional
mask operand (operand 2) that specifies which elements of the destination
vectors should be loaded. Other elements of the destination
vectors are set to zero. The operation is equivalent to:
</p>
<div class="smallexample">
<pre class="smallexample">int c = GET_MODE_SIZE (<var>m</var>) / GET_MODE_SIZE (<var>n</var>);
for (j = 0; j &lt; GET_MODE_NUNITS (<var>n</var>); j++)
if (operand2[j])
for (i = 0; i &lt; c; i++)
operand0[i][j] = operand1[j * c + i];
else
for (i = 0; i &lt; c; i++)
operand0[i][j] = 0;
</pre></div>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-vec_005fstore_005flanesmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_store_lanes<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Equivalent to &lsquo;<samp>vec_load_lanes<var>m</var><var>n</var></samp>&rsquo;, with the memory
and register operands reversed. That is, the instruction is
equivalent to:
</p>
<div class="smallexample">
<pre class="smallexample">int c = GET_MODE_SIZE (<var>m</var>) / GET_MODE_SIZE (<var>n</var>);
for (j = 0; j &lt; GET_MODE_NUNITS (<var>n</var>); j++)
for (i = 0; i &lt; c; i++)
operand0[j * c + i] = operand1[i][j];
</pre></div>
<p>for a memory operand 0 and register operand 1.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-vec_005fmask_005fstore_005flanesmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_mask_store_lanes<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>vec_store_lanes<var>m</var><var>n</var></samp>&rsquo;, but takes an additional
mask operand (operand 2) that specifies which elements of the source
vectors should be stored. The operation is equivalent to:
</p>
<div class="smallexample">
<pre class="smallexample">int c = GET_MODE_SIZE (<var>m</var>) / GET_MODE_SIZE (<var>n</var>);
for (j = 0; j &lt; GET_MODE_NUNITS (<var>n</var>); j++)
if (operand2[j])
for (i = 0; i &lt; c; i++)
operand0[j * c + i] = operand1[i][j];
</pre></div>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-gather_005floadm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>gather_load<var>m</var></samp>&rsquo;</dt>
<dd><p>Load several separate memory locations into a vector of mode <var>m</var>.
Operand 1 is a scalar base address and operand 2 is a vector of
offsets from that base. Operand 0 is a destination vector with the
same number of elements as the offset. For each element index <var>i</var>:
</p>
<ul>
<li> extend the offset element <var>i</var> to address width, using zero
extension if operand 3 is 1 and sign extension if operand 3 is zero;
</li><li> multiply the extended offset by operand 4;
</li><li> add the result to the base; and
</li><li> load the value at that address into element <var>i</var> of operand 0.
</li></ul>
<p>The value of operand 3 does not matter if the offsets are already
address width.
</p>
<a name="index-mask_005fgather_005floadm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>mask_gather_load<var>m</var></samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>gather_load<var>m</var></samp>&rsquo;, but takes an extra mask operand as
operand 5. Bit <var>i</var> of the mask is set if element <var>i</var>
of the result should be loaded from memory and clear if element <var>i</var>
of the result should be set to zero.
</p>
<a name="index-scatter_005fstorem-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>scatter_store<var>m</var></samp>&rsquo;</dt>
<dd><p>Store a vector of mode <var>m</var> into several distinct memory locations.
Operand 0 is a scalar base address and operand 1 is a vector of offsets
from that base. Operand 4 is the vector of values that should be stored,
which has the same number of elements as the offset. For each element
index <var>i</var>:
</p>
<ul>
<li> extend the offset element <var>i</var> to address width, using zero
extension if operand 2 is 1 and sign extension if operand 2 is zero;
</li><li> multiply the extended offset by operand 3;
</li><li> add the result to the base; and
</li><li> store element <var>i</var> of operand 4 to that address.
</li></ul>
<p>The value of operand 2 does not matter if the offsets are already
address width.
</p>
<a name="index-mask_005fscatter_005fstorem-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>mask_scatter_store<var>m</var></samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>scatter_store<var>m</var></samp>&rsquo;, but takes an extra mask operand as
operand 5. Bit <var>i</var> of the mask is set if element <var>i</var>
of the result should be stored to memory.
</p>
<a name="index-vec_005fsetm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_set<var>m</var></samp>&rsquo;</dt>
<dd><p>Set given field in the vector value. Operand 0 is the vector to modify,
operand 1 is new value of field and operand 2 specify the field index.
</p>
<a name="index-vec_005fextractmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_extract<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Extract given field from the vector value. Operand 1 is the vector, operand 2
specify field index and operand 0 place to store value into. The
<var>n</var> mode is the mode of the field or vector of fields that should be
extracted, should be either element mode of the vector mode <var>m</var>, or
a vector mode with the same element mode and smaller number of elements.
If <var>n</var> is a vector mode, the index is counted in units of that mode.
</p>
<a name="index-vec_005finitmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_init<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Initialize the vector to given values. Operand 0 is the vector to initialize
and operand 1 is parallel containing values for individual fields. The
<var>n</var> mode is the mode of the elements, should be either element mode of
the vector mode <var>m</var>, or a vector mode with the same element mode and
smaller number of elements.
</p>
<a name="index-vec_005fduplicatem-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_duplicate<var>m</var></samp>&rsquo;</dt>
<dd><p>Initialize vector output operand 0 so that each element has the value given
by scalar input operand 1. The vector has mode <var>m</var> and the scalar has
the mode appropriate for one element of <var>m</var>.
</p>
<p>This pattern only handles duplicates of non-constant inputs. Constant
vectors go through the <code>mov<var>m</var></code> pattern instead.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-vec_005fseriesm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_series<var>m</var></samp>&rsquo;</dt>
<dd><p>Initialize vector output operand 0 so that element <var>i</var> is equal to
operand 1 plus <var>i</var> times operand 2. In other words, create a linear
series whose base value is operand 1 and whose step is operand 2.
</p>
<p>The vector output has mode <var>m</var> and the scalar inputs have the mode
appropriate for one element of <var>m</var>. This pattern is not used for
floating-point vectors, in order to avoid having to specify the
rounding behavior for <var>i</var> &gt; 1.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-while_005fultmn-instruction-pattern"></a>
</dd>
<dt><code>while_ult<var>m</var><var>n</var></code></dt>
<dd><p>Set operand 0 to a mask that is true while incrementing operand 1
gives a value that is less than operand 2. Operand 0 has mode <var>n</var>
and operands 1 and 2 are scalar integers of mode <var>m</var>.
The operation is equivalent to:
</p>
<div class="smallexample">
<pre class="smallexample">operand0[0] = operand1 &lt; operand2;
for (i = 1; i &lt; GET_MODE_NUNITS (<var>n</var>); i++)
operand0[i] = operand0[i - 1] &amp;&amp; (operand1 + i &lt; operand2);
</pre></div>
<a name="index-vec_005fcmpmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_cmp<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Output a vector comparison. Operand 0 of mode <var>n</var> is the destination for
predicate in operand 1 which is a signed vector comparison with operands of
mode <var>m</var> in operands 2 and 3. Predicate is computed by element-wise
evaluation of the vector comparison with a truth value of all-ones and a false
value of all-zeros.
</p>
<a name="index-vec_005fcmpumn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_cmpu<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Similar to <code>vec_cmp<var>m</var><var>n</var></code> but perform unsigned vector comparison.
</p>
<a name="index-vec_005fcmpeqmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_cmpeq<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Similar to <code>vec_cmp<var>m</var><var>n</var></code> but perform equality or non-equality
vector comparison only. If <code>vec_cmp<var>m</var><var>n</var></code>
or <code>vec_cmpu<var>m</var><var>n</var></code> instruction pattern is supported,
it will be preferred over <code>vec_cmpeq<var>m</var><var>n</var></code>, so there is
no need to define this instruction pattern if the others are supported.
</p>
<a name="index-vcondmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vcond<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Output a conditional vector move. Operand 0 is the destination to
receive a combination of operand 1 and operand 2, which are of mode <var>m</var>,
dependent on the outcome of the predicate in operand 3 which is a signed
vector comparison with operands of mode <var>n</var> in operands 4 and 5. The
modes <var>m</var> and <var>n</var> should have the same size. Operand 0
will be set to the value <var>op1</var> &amp; <var>msk</var> | <var>op2</var> &amp; ~<var>msk</var>
where <var>msk</var> is computed by element-wise evaluation of the vector
comparison with a truth value of all-ones and a false value of all-zeros.
</p>
<a name="index-vcondumn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vcondu<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Similar to <code>vcond<var>m</var><var>n</var></code> but performs unsigned vector
comparison.
</p>
<a name="index-vcondeqmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vcondeq<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Similar to <code>vcond<var>m</var><var>n</var></code> but performs equality or
non-equality vector comparison only. If <code>vcond<var>m</var><var>n</var></code>
or <code>vcondu<var>m</var><var>n</var></code> instruction pattern is supported,
it will be preferred over <code>vcondeq<var>m</var><var>n</var></code>, so there is
no need to define this instruction pattern if the others are supported.
</p>
<a name="index-vcond_005fmask_005fmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vcond_mask_<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Similar to <code>vcond<var>m</var><var>n</var></code> but operand 3 holds a pre-computed
result of vector comparison.
</p>
<a name="index-maskloadmn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>maskload<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Perform a masked load of vector from memory operand 1 of mode <var>m</var>
into register operand 0. Mask is provided in register operand 2 of
mode <var>n</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-maskstoremn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>maskstore<var>m</var><var>n</var></samp>&rsquo;</dt>
<dd><p>Perform a masked store of vector from register operand 1 of mode <var>m</var>
into memory operand 0. Mask is provided in register operand 2 of
mode <var>n</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-vec_005fpermm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_perm<var>m</var></samp>&rsquo;</dt>
<dd><p>Output a (variable) vector permutation. Operand 0 is the destination
to receive elements from operand 1 and operand 2, which are of mode
<var>m</var>. Operand 3 is the <em>selector</em>. It is an integral mode
vector of the same width and number of elements as mode <var>m</var>.
</p>
<p>The input elements are numbered from 0 in operand 1 through
<em>2*<var>N</var>-1</em> in operand 2. The elements of the selector must
be computed modulo <em>2*<var>N</var></em>. Note that if
<code>rtx_equal_p(operand1, operand2)</code>, this can be implemented
with just operand 1 and selector elements modulo <var>N</var>.
</p>
<p>In order to make things easy for a number of targets, if there is no
&lsquo;<samp>vec_perm</samp>&rsquo; pattern for mode <var>m</var>, but there is for mode <var>q</var>
where <var>q</var> is a vector of <code>QImode</code> of the same width as <var>m</var>,
the middle-end will lower the mode <var>m</var> <code>VEC_PERM_EXPR</code> to
mode <var>q</var>.
</p>
<p>See also <code>TARGET_VECTORIZER_VEC_PERM_CONST</code>, which performs
the analogous operation for constant selectors.
</p>
<a name="index-pushm1-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>push<var>m</var>1</samp>&rsquo;</dt>
<dd><p>Output a push instruction. Operand 0 is value to push. Used only when
<code>PUSH_ROUNDING</code> is defined. For historical reason, this pattern may be
missing and in such case an <code>mov</code> expander is used instead, with a
<code>MEM</code> expression forming the push operation. The <code>mov</code> expander
method is deprecated.
</p>
<a name="index-addm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>add<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Add operand 2 and operand 1, storing the result in operand 0. All operands
must have mode <var>m</var>. This can be used even on two-address machines, by
means of constraints requiring operands 1 and 0 to be the same location.
</p>
<a name="index-ssaddm3-instruction-pattern"></a>
<a name="index-usaddm3-instruction-pattern"></a>
<a name="index-subm3-instruction-pattern"></a>
<a name="index-sssubm3-instruction-pattern"></a>
<a name="index-ussubm3-instruction-pattern"></a>
<a name="index-mulm3-instruction-pattern"></a>
<a name="index-ssmulm3-instruction-pattern"></a>
<a name="index-usmulm3-instruction-pattern"></a>
<a name="index-divm3-instruction-pattern"></a>
<a name="index-ssdivm3-instruction-pattern"></a>
<a name="index-udivm3-instruction-pattern"></a>
<a name="index-usdivm3-instruction-pattern"></a>
<a name="index-modm3-instruction-pattern"></a>
<a name="index-umodm3-instruction-pattern"></a>
<a name="index-uminm3-instruction-pattern"></a>
<a name="index-umaxm3-instruction-pattern"></a>
<a name="index-andm3-instruction-pattern"></a>
<a name="index-iorm3-instruction-pattern"></a>
<a name="index-xorm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ssadd<var>m</var>3</samp>&rsquo;, &lsquo;<samp>usadd<var>m</var>3</samp>&rsquo;</dt>
<dt>&lsquo;<samp>sub<var>m</var>3</samp>&rsquo;, &lsquo;<samp>sssub<var>m</var>3</samp>&rsquo;, &lsquo;<samp>ussub<var>m</var>3</samp>&rsquo;</dt>
<dt>&lsquo;<samp>mul<var>m</var>3</samp>&rsquo;, &lsquo;<samp>ssmul<var>m</var>3</samp>&rsquo;, &lsquo;<samp>usmul<var>m</var>3</samp>&rsquo;</dt>
<dt>&lsquo;<samp>div<var>m</var>3</samp>&rsquo;, &lsquo;<samp>ssdiv<var>m</var>3</samp>&rsquo;</dt>
<dt>&lsquo;<samp>udiv<var>m</var>3</samp>&rsquo;, &lsquo;<samp>usdiv<var>m</var>3</samp>&rsquo;</dt>
<dt>&lsquo;<samp>mod<var>m</var>3</samp>&rsquo;, &lsquo;<samp>umod<var>m</var>3</samp>&rsquo;</dt>
<dt>&lsquo;<samp>umin<var>m</var>3</samp>&rsquo;, &lsquo;<samp>umax<var>m</var>3</samp>&rsquo;</dt>
<dt>&lsquo;<samp>and<var>m</var>3</samp>&rsquo;, &lsquo;<samp>ior<var>m</var>3</samp>&rsquo;, &lsquo;<samp>xor<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Similar, for other arithmetic operations.
</p>
<a name="index-addvm4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>addv<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>add<var>m</var>3</code> but takes a <code>code_label</code> as operand 3 and
emits code to jump to it if signed overflow occurs during the addition.
This pattern is used to implement the built-in functions performing
signed integer addition with overflow checking.
</p>
<a name="index-subvm4-instruction-pattern"></a>
<a name="index-mulvm4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>subv<var>m</var>4</samp>&rsquo;, &lsquo;<samp>mulv<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Similar, for other signed arithmetic operations.
</p>
<a name="index-uaddvm4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>uaddv<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>addv<var>m</var>4</code> but for unsigned addition. That is to
say, the operation is the same as signed addition but the jump
is taken only on unsigned overflow.
</p>
<a name="index-usubvm4-instruction-pattern"></a>
<a name="index-umulvm4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>usubv<var>m</var>4</samp>&rsquo;, &lsquo;<samp>umulv<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Similar, for other unsigned arithmetic operations.
</p>
<a name="index-addptrm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>addptr<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Like <code>add<var>m</var>3</code> but is guaranteed to only be used for address
calculations. The expanded code is not allowed to clobber the
condition code. It only needs to be defined if <code>add<var>m</var>3</code>
sets the condition code. If adds used for address calculations and
normal adds are not compatible it is required to expand a distinct
pattern (e.g. using an unspec). The pattern is used by LRA to emit
address calculations. <code>add<var>m</var>3</code> is used if
<code>addptr<var>m</var>3</code> is not defined.
</p>
<a name="index-fmam4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fma<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Multiply operand 2 and operand 1, then add operand 3, storing the
result in operand 0 without doing an intermediate rounding step. All
operands must have mode <var>m</var>. This pattern is used to implement
the <code>fma</code>, <code>fmaf</code>, and <code>fmal</code> builtin functions from
the ISO C99 standard.
</p>
<a name="index-fmsm4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fms<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>fma<var>m</var>4</code>, except operand 3 subtracted from the
product instead of added to the product. This is represented
in the rtl as
</p>
<div class="smallexample">
<pre class="smallexample">(fma:<var>m</var> <var>op1</var> <var>op2</var> (neg:<var>m</var> <var>op3</var>))
</pre></div>
<a name="index-fnmam4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fnma<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>fma<var>m</var>4</code> except that the intermediate product
is negated before being added to operand 3. This is represented
in the rtl as
</p>
<div class="smallexample">
<pre class="smallexample">(fma:<var>m</var> (neg:<var>m</var> <var>op1</var>) <var>op2</var> <var>op3</var>)
</pre></div>
<a name="index-fnmsm4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fnms<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>fms<var>m</var>4</code> except that the intermediate product
is negated before subtracting operand 3. This is represented
in the rtl as
</p>
<div class="smallexample">
<pre class="smallexample">(fma:<var>m</var> (neg:<var>m</var> <var>op1</var>) <var>op2</var> (neg:<var>m</var> <var>op3</var>))
</pre></div>
<a name="index-minm3-instruction-pattern"></a>
<a name="index-maxm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>smin<var>m</var>3</samp>&rsquo;, &lsquo;<samp>smax<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Signed minimum and maximum operations. When used with floating point,
if both operands are zeros, or if either operand is <code>NaN</code>, then
it is unspecified which of the two operands is returned as the result.
</p>
<a name="index-fminm3-instruction-pattern"></a>
<a name="index-fmaxm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fmin<var>m</var>3</samp>&rsquo;, &lsquo;<samp>fmax<var>m</var>3</samp>&rsquo;</dt>
<dd><p>IEEE-conformant minimum and maximum operations. If one operand is a quiet
<code>NaN</code>, then the other operand is returned. If both operands are quiet
<code>NaN</code>, then a quiet <code>NaN</code> is returned. In the case when gcc supports
signaling <code>NaN</code> (-fsignaling-nans) an invalid floating point exception is
raised and a quiet <code>NaN</code> is returned.
</p>
<p>All operands have mode <var>m</var>, which is a scalar or vector
floating-point mode. These patterns are not allowed to <code>FAIL</code>.
</p>
<a name="index-reduc_005fsmin_005fscal_005fm-instruction-pattern"></a>
<a name="index-reduc_005fsmax_005fscal_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>reduc_smin_scal_<var>m</var></samp>&rsquo;, &lsquo;<samp>reduc_smax_scal_<var>m</var></samp>&rsquo;</dt>
<dd><p>Find the signed minimum/maximum of the elements of a vector. The vector is
operand 1, and operand 0 is the scalar result, with mode equal to the mode of
the elements of the input vector.
</p>
<a name="index-reduc_005fumin_005fscal_005fm-instruction-pattern"></a>
<a name="index-reduc_005fumax_005fscal_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>reduc_umin_scal_<var>m</var></samp>&rsquo;, &lsquo;<samp>reduc_umax_scal_<var>m</var></samp>&rsquo;</dt>
<dd><p>Find the unsigned minimum/maximum of the elements of a vector. The vector is
operand 1, and operand 0 is the scalar result, with mode equal to the mode of
the elements of the input vector.
</p>
<a name="index-reduc_005fplus_005fscal_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>reduc_plus_scal_<var>m</var></samp>&rsquo;</dt>
<dd><p>Compute the sum of the elements of a vector. The vector is operand 1, and
operand 0 is the scalar result, with mode equal to the mode of the elements of
the input vector.
</p>
<a name="index-reduc_005fand_005fscal_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>reduc_and_scal_<var>m</var></samp>&rsquo;</dt>
<dd><a name="index-reduc_005fior_005fscal_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>reduc_ior_scal_<var>m</var></samp>&rsquo;</dt>
<dd><a name="index-reduc_005fxor_005fscal_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>reduc_xor_scal_<var>m</var></samp>&rsquo;</dt>
<dd><p>Compute the bitwise <code>AND</code>/<code>IOR</code>/<code>XOR</code> reduction of the elements
of a vector of mode <var>m</var>. Operand 1 is the vector input and operand 0
is the scalar result. The mode of the scalar result is the same as one
element of <var>m</var>.
</p>
<a name="index-extract_005flast_005fm-instruction-pattern"></a>
</dd>
<dt><code>extract_last_<var>m</var></code></dt>
<dd><p>Find the last set bit in mask operand 1 and extract the associated element
of vector operand 2. Store the result in scalar operand 0. Operand 2
has vector mode <var>m</var> while operand 0 has the mode appropriate for one
element of <var>m</var>. Operand 1 has the usual mask mode for vectors of mode
<var>m</var>; see <code>TARGET_VECTORIZE_GET_MASK_MODE</code>.
</p>
<a name="index-fold_005fextract_005flast_005fm-instruction-pattern"></a>
</dd>
<dt><code>fold_extract_last_<var>m</var></code></dt>
<dd><p>If any bits of mask operand 2 are set, find the last set bit, extract
the associated element from vector operand 3, and store the result
in operand 0. Store operand 1 in operand 0 otherwise. Operand 3
has mode <var>m</var> and operands 0 and 1 have the mode appropriate for
one element of <var>m</var>. Operand 2 has the usual mask mode for vectors
of mode <var>m</var>; see <code>TARGET_VECTORIZE_GET_MASK_MODE</code>.
</p>
<a name="index-fold_005fleft_005fplus_005fm-instruction-pattern"></a>
</dd>
<dt><code>fold_left_plus_<var>m</var></code></dt>
<dd><p>Take scalar operand 1 and successively add each element from vector
operand 2. Store the result in scalar operand 0. The vector has
mode <var>m</var> and the scalars have the mode appropriate for one
element of <var>m</var>. The operation is strictly in-order: there is
no reassociation.
</p>
<a name="index-sdot_005fprodm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sdot_prod<var>m</var></samp>&rsquo;</dt>
<dd><a name="index-udot_005fprodm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>udot_prod<var>m</var></samp>&rsquo;</dt>
<dd><p>Compute the sum of the products of two signed/unsigned elements.
Operand 1 and operand 2 are of the same mode. Their product, which is of a
wider mode, is computed and added to operand 3. Operand 3 is of a mode equal or
wider than the mode of the product. The result is placed in operand 0, which
is of the same mode as operand 3.
</p>
<a name="index-ssadm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ssad<var>m</var></samp>&rsquo;</dt>
<dd><a name="index-usadm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>usad<var>m</var></samp>&rsquo;</dt>
<dd><p>Compute the sum of absolute differences of two signed/unsigned elements.
Operand 1 and operand 2 are of the same mode. Their absolute difference, which
is of a wider mode, is computed and added to operand 3. Operand 3 is of a mode
equal or wider than the mode of the absolute difference. The result is placed
in operand 0, which is of the same mode as operand 3.
</p>
<a name="index-widen_005fssumm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>widen_ssum<var>m3</var></samp>&rsquo;</dt>
<dd><a name="index-widen_005fusumm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>widen_usum<var>m3</var></samp>&rsquo;</dt>
<dd><p>Operands 0 and 2 are of the same mode, which is wider than the mode of
operand 1. Add operand 1 to operand 2 and place the widened result in
operand 0. (This is used express accumulation of elements into an accumulator
of a wider mode.)
</p>
<a name="index-vec_005fshl_005finsert_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_shl_insert_<var>m</var></samp>&rsquo;</dt>
<dd><p>Shift the elements in vector input operand 1 left one element (i.e.
away from element 0) and fill the vacated element 0 with the scalar
in operand 2. Store the result in vector output operand 0. Operands
0 and 1 have mode <var>m</var> and operand 2 has the mode appropriate for
one element of <var>m</var>.
</p>
<a name="index-vec_005fshr_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_shr_<var>m</var></samp>&rsquo;</dt>
<dd><p>Whole vector right shift in bits, i.e. towards element 0.
Operand 1 is a vector to be shifted.
Operand 2 is an integer shift amount in bits.
Operand 0 is where the resulting shifted vector is stored.
The output and input vectors should have the same modes.
</p>
<a name="index-vec_005fpack_005ftrunc_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_pack_trunc_<var>m</var></samp>&rsquo;</dt>
<dd><p>Narrow (demote) and merge the elements of two vectors. Operands 1 and 2
are vectors of the same mode having N integral or floating point elements
of size S. Operand 0 is the resulting vector in which 2*N elements of
size N/2 are concatenated after narrowing them down using truncation.
</p>
<a name="index-vec_005fpack_005fssat_005fm-instruction-pattern"></a>
<a name="index-vec_005fpack_005fusat_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_pack_ssat_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_pack_usat_<var>m</var></samp>&rsquo;</dt>
<dd><p>Narrow (demote) and merge the elements of two vectors. Operands 1 and 2
are vectors of the same mode having N integral elements of size S.
Operand 0 is the resulting vector in which the elements of the two input
vectors are concatenated after narrowing them down using signed/unsigned
saturating arithmetic.
</p>
<a name="index-vec_005fpack_005fsfix_005ftrunc_005fm-instruction-pattern"></a>
<a name="index-vec_005fpack_005fufix_005ftrunc_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_pack_sfix_trunc_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_pack_ufix_trunc_<var>m</var></samp>&rsquo;</dt>
<dd><p>Narrow, convert to signed/unsigned integral type and merge the elements
of two vectors. Operands 1 and 2 are vectors of the same mode having N
floating point elements of size S. Operand 0 is the resulting vector
in which 2*N elements of size N/2 are concatenated.
</p>
<a name="index-vec_005funpacks_005fhi_005fm-instruction-pattern"></a>
<a name="index-vec_005funpacks_005flo_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_unpacks_hi_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_unpacks_lo_<var>m</var></samp>&rsquo;</dt>
<dd><p>Extract and widen (promote) the high/low part of a vector of signed
integral or floating point elements. The input vector (operand 1) has N
elements of size S. Widen (promote) the high/low elements of the vector
using signed or floating point extension and place the resulting N/2
values of size 2*S in the output vector (operand 0).
</p>
<a name="index-vec_005funpacku_005fhi_005fm-instruction-pattern"></a>
<a name="index-vec_005funpacku_005flo_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_unpacku_hi_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_unpacku_lo_<var>m</var></samp>&rsquo;</dt>
<dd><p>Extract and widen (promote) the high/low part of a vector of unsigned
integral elements. The input vector (operand 1) has N elements of size S.
Widen (promote) the high/low elements of the vector using zero extension and
place the resulting N/2 values of size 2*S in the output vector (operand 0).
</p>
<a name="index-vec_005funpacks_005ffloat_005fhi_005fm-instruction-pattern"></a>
<a name="index-vec_005funpacks_005ffloat_005flo_005fm-instruction-pattern"></a>
<a name="index-vec_005funpacku_005ffloat_005fhi_005fm-instruction-pattern"></a>
<a name="index-vec_005funpacku_005ffloat_005flo_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_unpacks_float_hi_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_unpacks_float_lo_<var>m</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>vec_unpacku_float_hi_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_unpacku_float_lo_<var>m</var></samp>&rsquo;</dt>
<dd><p>Extract, convert to floating point type and widen the high/low part of a
vector of signed/unsigned integral elements. The input vector (operand 1)
has N elements of size S. Convert the high/low elements of the vector using
floating point conversion and place the resulting N/2 values of size 2*S in
the output vector (operand 0).
</p>
<a name="index-vec_005fwiden_005fumult_005fhi_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fumult_005flo_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fsmult_005fhi_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fsmult_005flo_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fumult_005feven_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fumult_005fodd_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fsmult_005feven_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fsmult_005fodd_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_widen_umult_hi_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_widen_umult_lo_<var>m</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>vec_widen_smult_hi_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_widen_smult_lo_<var>m</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>vec_widen_umult_even_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_widen_umult_odd_<var>m</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>vec_widen_smult_even_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_widen_smult_odd_<var>m</var></samp>&rsquo;</dt>
<dd><p>Signed/Unsigned widening multiplication. The two inputs (operands 1 and 2)
are vectors with N signed/unsigned elements of size S. Multiply the high/low
or even/odd elements of the two vectors, and put the N/2 products of size 2*S
in the output vector (operand 0). A target shouldn&rsquo;t implement even/odd pattern
pair if it is less efficient than lo/hi one.
</p>
<a name="index-vec_005fwiden_005fushiftl_005fhi_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fushiftl_005flo_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fsshiftl_005fhi_005fm-instruction-pattern"></a>
<a name="index-vec_005fwiden_005fsshiftl_005flo_005fm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vec_widen_ushiftl_hi_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_widen_ushiftl_lo_<var>m</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>vec_widen_sshiftl_hi_<var>m</var></samp>&rsquo;, &lsquo;<samp>vec_widen_sshiftl_lo_<var>m</var></samp>&rsquo;</dt>
<dd><p>Signed/Unsigned widening shift left. The first input (operand 1) is a vector
with N signed/unsigned elements of size S. Operand 2 is a constant. Shift
the high/low elements of operand 1, and put the N/2 results of size 2*S in the
output vector (operand 0).
</p>
<a name="index-mulhisi3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>mulhisi3</samp>&rsquo;</dt>
<dd><p>Multiply operands 1 and 2, which have mode <code>HImode</code>, and store
a <code>SImode</code> product in operand 0.
</p>
<a name="index-mulqihi3-instruction-pattern"></a>
<a name="index-mulsidi3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>mulqihi3</samp>&rsquo;, &lsquo;<samp>mulsidi3</samp>&rsquo;</dt>
<dd><p>Similar widening-multiplication instructions of other widths.
</p>
<a name="index-umulqihi3-instruction-pattern"></a>
<a name="index-umulhisi3-instruction-pattern"></a>
<a name="index-umulsidi3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>umulqihi3</samp>&rsquo;, &lsquo;<samp>umulhisi3</samp>&rsquo;, &lsquo;<samp>umulsidi3</samp>&rsquo;</dt>
<dd><p>Similar widening-multiplication instructions that do unsigned
multiplication.
</p>
<a name="index-usmulqihi3-instruction-pattern"></a>
<a name="index-usmulhisi3-instruction-pattern"></a>
<a name="index-usmulsidi3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>usmulqihi3</samp>&rsquo;, &lsquo;<samp>usmulhisi3</samp>&rsquo;, &lsquo;<samp>usmulsidi3</samp>&rsquo;</dt>
<dd><p>Similar widening-multiplication instructions that interpret the first
operand as unsigned and the second operand as signed, then do a signed
multiplication.
</p>
<a name="index-smulm3_005fhighpart-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>smul<var>m</var>3_highpart</samp>&rsquo;</dt>
<dd><p>Perform a signed multiplication of operands 1 and 2, which have mode
<var>m</var>, and store the most significant half of the product in operand 0.
The least significant half of the product is discarded.
</p>
<a name="index-umulm3_005fhighpart-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>umul<var>m</var>3_highpart</samp>&rsquo;</dt>
<dd><p>Similar, but the multiplication is unsigned.
</p>
<a name="index-maddmn4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>madd<var>m</var><var>n</var>4</samp>&rsquo;</dt>
<dd><p>Multiply operands 1 and 2, sign-extend them to mode <var>n</var>, add
operand 3, and store the result in operand 0. Operands 1 and 2
have mode <var>m</var> and operands 0 and 3 have mode <var>n</var>.
Both modes must be integer or fixed-point modes and <var>n</var> must be twice
the size of <var>m</var>.
</p>
<p>In other words, <code>madd<var>m</var><var>n</var>4</code> is like
<code>mul<var>m</var><var>n</var>3</code> except that it also adds operand 3.
</p>
<p>These instructions are not allowed to <code>FAIL</code>.
</p>
<a name="index-umaddmn4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>umadd<var>m</var><var>n</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>madd<var>m</var><var>n</var>4</code>, but zero-extend the multiplication
operands instead of sign-extending them.
</p>
<a name="index-ssmaddmn4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ssmadd<var>m</var><var>n</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>madd<var>m</var><var>n</var>4</code>, but all involved operations must be
signed-saturating.
</p>
<a name="index-usmaddmn4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>usmadd<var>m</var><var>n</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>umadd<var>m</var><var>n</var>4</code>, but all involved operations must be
unsigned-saturating.
</p>
<a name="index-msubmn4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>msub<var>m</var><var>n</var>4</samp>&rsquo;</dt>
<dd><p>Multiply operands 1 and 2, sign-extend them to mode <var>n</var>, subtract the
result from operand 3, and store the result in operand 0. Operands 1 and 2
have mode <var>m</var> and operands 0 and 3 have mode <var>n</var>.
Both modes must be integer or fixed-point modes and <var>n</var> must be twice
the size of <var>m</var>.
</p>
<p>In other words, <code>msub<var>m</var><var>n</var>4</code> is like
<code>mul<var>m</var><var>n</var>3</code> except that it also subtracts the result
from operand 3.
</p>
<p>These instructions are not allowed to <code>FAIL</code>.
</p>
<a name="index-umsubmn4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>umsub<var>m</var><var>n</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>msub<var>m</var><var>n</var>4</code>, but zero-extend the multiplication
operands instead of sign-extending them.
</p>
<a name="index-ssmsubmn4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ssmsub<var>m</var><var>n</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>msub<var>m</var><var>n</var>4</code>, but all involved operations must be
signed-saturating.
</p>
<a name="index-usmsubmn4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>usmsub<var>m</var><var>n</var>4</samp>&rsquo;</dt>
<dd><p>Like <code>umsub<var>m</var><var>n</var>4</code>, but all involved operations must be
unsigned-saturating.
</p>
<a name="index-divmodm4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>divmod<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Signed division that produces both a quotient and a remainder.
Operand 1 is divided by operand 2 to produce a quotient stored
in operand 0 and a remainder stored in operand 3.
</p>
<p>For machines with an instruction that produces both a quotient and a
remainder, provide a pattern for &lsquo;<samp>divmod<var>m</var>4</samp>&rsquo; but do not
provide patterns for &lsquo;<samp>div<var>m</var>3</samp>&rsquo; and &lsquo;<samp>mod<var>m</var>3</samp>&rsquo;. This
allows optimization in the relatively common case when both the quotient
and remainder are computed.
</p>
<p>If an instruction that just produces a quotient or just a remainder
exists and is more efficient than the instruction that produces both,
write the output routine of &lsquo;<samp>divmod<var>m</var>4</samp>&rsquo; to call
<code>find_reg_note</code> and look for a <code>REG_UNUSED</code> note on the
quotient or remainder and generate the appropriate instruction.
</p>
<a name="index-udivmodm4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>udivmod<var>m</var>4</samp>&rsquo;</dt>
<dd><p>Similar, but does unsigned division.
</p>
<a name="shift-patterns"></a><a name="index-ashlm3-instruction-pattern"></a>
<a name="index-ssashlm3-instruction-pattern"></a>
<a name="index-usashlm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ashl<var>m</var>3</samp>&rsquo;, &lsquo;<samp>ssashl<var>m</var>3</samp>&rsquo;, &lsquo;<samp>usashl<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Arithmetic-shift operand 1 left by a number of bits specified by operand
2, and store the result in operand 0. Here <var>m</var> is the mode of
operand 0 and operand 1; operand 2&rsquo;s mode is specified by the
instruction pattern, and the compiler will convert the operand to that
mode before generating the instruction. The shift or rotate expander
or instruction pattern should explicitly specify the mode of the operand 2,
it should never be <code>VOIDmode</code>. The meaning of out-of-range shift
counts can optionally be specified by <code>TARGET_SHIFT_TRUNCATION_MASK</code>.
See <a href="Misc.html#TARGET_005fSHIFT_005fTRUNCATION_005fMASK">TARGET_SHIFT_TRUNCATION_MASK</a>. Operand 2 is always a scalar type.
</p>
<a name="index-ashrm3-instruction-pattern"></a>
<a name="index-lshrm3-instruction-pattern"></a>
<a name="index-rotlm3-instruction-pattern"></a>
<a name="index-rotrm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ashr<var>m</var>3</samp>&rsquo;, &lsquo;<samp>lshr<var>m</var>3</samp>&rsquo;, &lsquo;<samp>rotl<var>m</var>3</samp>&rsquo;, &lsquo;<samp>rotr<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Other shift and rotate instructions, analogous to the
<code>ashl<var>m</var>3</code> instructions. Operand 2 is always a scalar type.
</p>
<a name="index-vashlm3-instruction-pattern"></a>
<a name="index-vashrm3-instruction-pattern"></a>
<a name="index-vlshrm3-instruction-pattern"></a>
<a name="index-vrotlm3-instruction-pattern"></a>
<a name="index-vrotrm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>vashl<var>m</var>3</samp>&rsquo;, &lsquo;<samp>vashr<var>m</var>3</samp>&rsquo;, &lsquo;<samp>vlshr<var>m</var>3</samp>&rsquo;, &lsquo;<samp>vrotl<var>m</var>3</samp>&rsquo;, &lsquo;<samp>vrotr<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Vector shift and rotate instructions that take vectors as operand 2
instead of a scalar type.
</p>
<a name="index-bswapm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>bswap<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Reverse the order of bytes of operand 1 and store the result in operand 0.
</p>
<a name="index-negm2-instruction-pattern"></a>
<a name="index-ssnegm2-instruction-pattern"></a>
<a name="index-usnegm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>neg<var>m</var>2</samp>&rsquo;, &lsquo;<samp>ssneg<var>m</var>2</samp>&rsquo;, &lsquo;<samp>usneg<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Negate operand 1 and store the result in operand 0.
</p>
<a name="index-negvm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>negv<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Like <code>neg<var>m</var>2</code> but takes a <code>code_label</code> as operand 2 and
emits code to jump to it if signed overflow occurs during the negation.
</p>
<a name="index-absm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>abs<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the absolute value of operand 1 into operand 0.
</p>
<a name="index-sqrtm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sqrt<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the square root of operand 1 into operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-rsqrtm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>rsqrt<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the reciprocal of the square root of operand 1 into operand 0.
Both operands have mode <var>m</var>, which is a scalar or vector
floating-point mode.
</p>
<p>On most architectures this pattern is only approximate, so either
its C condition or the <code>TARGET_OPTAB_SUPPORTED_P</code> hook should
check for the appropriate math flags. (Using the C condition is
more direct, but using <code>TARGET_OPTAB_SUPPORTED_P</code> can be useful
if a target-specific built-in also uses the &lsquo;<samp>rsqrt<var>m</var>2</samp>&rsquo;
pattern.)
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-fmodm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fmod<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Store the remainder of dividing operand 1 by operand 2 into
operand 0, rounded towards zero to an integer. All operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-remainderm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>remainder<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Store the remainder of dividing operand 1 by operand 2 into
operand 0, rounded to the nearest integer. All operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-scalbm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>scalb<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Raise <code>FLT_RADIX</code> to the power of operand 2, multiply it by
operand 1, and store the result in operand 0. All operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-ldexpm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ldexp<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Raise 2 to the power of operand 2, multiply it by operand 1, and store
the result in operand 0. Operands 0 and 1 have mode <var>m</var>, which is
a scalar or vector floating-point mode. Operand 2&rsquo;s mode has
the same number of elements as <var>m</var> and each element is wide
enough to store an <code>int</code>. The integers are signed.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-cosm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>cos<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the cosine of operand 1 into operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-sinm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sin<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the sine of operand 1 into operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-sincosm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sincos<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Store the cosine of operand 2 into operand 0 and the sine of
operand 2 into operand 1. All operands have mode <var>m</var>,
which is a scalar or vector floating-point mode.
</p>
<p>Targets that can calculate the sine and cosine simultaneously can
implement this pattern as opposed to implementing individual
<code>sin<var>m</var>2</code> and <code>cos<var>m</var>2</code> patterns. The <code>sin</code>
and <code>cos</code> built-in functions will then be expanded to the
<code>sincos<var>m</var>3</code> pattern, with one of the output values
left unused.
</p>
<a name="index-tanm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>tan<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the tangent of operand 1 into operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-asinm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>asin<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the arc sine of operand 1 into operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-acosm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>acos<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the arc cosine of operand 1 into operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-atanm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atan<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the arc tangent of operand 1 into operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-expm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>exp<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Raise e (the base of natural logarithms) to the power of operand 1
and store the result in operand 0. Both operands have mode <var>m</var>,
which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-expm1m2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>expm1<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Raise e (the base of natural logarithms) to the power of operand 1,
subtract 1, and store the result in operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>For inputs close to zero, the pattern is expected to be more
accurate than a separate <code>exp<var>m</var>2</code> and <code>sub<var>m</var>3</code>
would be.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-exp10m2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>exp10<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Raise 10 to the power of operand 1 and store the result in operand 0.
Both operands have mode <var>m</var>, which is a scalar or vector
floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-exp2m2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>exp2<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Raise 2 to the power of operand 1 and store the result in operand 0.
Both operands have mode <var>m</var>, which is a scalar or vector
floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-logm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>log<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the natural logarithm of operand 1 into operand 0. Both operands
have mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-log1pm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>log1p<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Add 1 to operand 1, compute the natural logarithm, and store
the result in operand 0. Both operands have mode <var>m</var>, which is
a scalar or vector floating-point mode.
</p>
<p>For inputs close to zero, the pattern is expected to be more
accurate than a separate <code>add<var>m</var>3</code> and <code>log<var>m</var>2</code>
would be.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-log10m2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>log10<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the base-10 logarithm of operand 1 into operand 0. Both operands
have mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-log2m2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>log2<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the base-2 logarithm of operand 1 into operand 0. Both operands
have mode <var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-logbm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>logb<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the base-<code>FLT_RADIX</code> logarithm of operand 1 into operand 0.
Both operands have mode <var>m</var>, which is a scalar or vector
floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-significandm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>significand<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the significand of floating-point operand 1 in operand 0.
Both operands have mode <var>m</var>, which is a scalar or vector
floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-powm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>pow<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Store the value of operand 1 raised to the exponent operand 2
into operand 0. All operands have mode <var>m</var>, which is a scalar
or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-atan2m3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atan2<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Store the arc tangent (inverse tangent) of operand 1 divided by
operand 2 into operand 0, using the signs of both arguments to
determine the quadrant of the result. All operands have mode
<var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-floorm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>floor<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the largest integral value not greater than operand 1 in operand 0.
Both operands have mode <var>m</var>, which is a scalar or vector
floating-point mode. If <samp>-ffp-int-builtin-inexact</samp> is in
effect, the &ldquo;inexact&rdquo; exception may be raised for noninteger
operands; otherwise, it may not.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-btruncm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>btrunc<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Round operand 1 to an integer, towards zero, and store the result in
operand 0. Both operands have mode <var>m</var>, which is a scalar or
vector floating-point mode. If <samp>-ffp-int-builtin-inexact</samp> is
in effect, the &ldquo;inexact&rdquo; exception may be raised for noninteger
operands; otherwise, it may not.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-roundm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>round<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Round operand 1 to the nearest integer, rounding away from zero in the
event of a tie, and store the result in operand 0. Both operands have
mode <var>m</var>, which is a scalar or vector floating-point mode. If
<samp>-ffp-int-builtin-inexact</samp> is in effect, the &ldquo;inexact&rdquo;
exception may be raised for noninteger operands; otherwise, it may
not.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-ceilm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ceil<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the smallest integral value not less than operand 1 in operand 0.
Both operands have mode <var>m</var>, which is a scalar or vector
floating-point mode. If <samp>-ffp-int-builtin-inexact</samp> is in
effect, the &ldquo;inexact&rdquo; exception may be raised for noninteger
operands; otherwise, it may not.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-nearbyintm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>nearbyint<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Round operand 1 to an integer, using the current rounding mode, and
store the result in operand 0. Do not raise an inexact condition when
the result is different from the argument. Both operands have mode
<var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-rintm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>rint<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Round operand 1 to an integer, using the current rounding mode, and
store the result in operand 0. Raise an inexact condition when
the result is different from the argument. Both operands have mode
<var>m</var>, which is a scalar or vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-lrintmn2"></a>
</dd>
<dt>&lsquo;<samp>lrint<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 (valid for floating point mode <var>m</var>) to fixed
point mode <var>n</var> as a signed number according to the current
rounding mode and store in operand 0 (which has mode <var>n</var>).
</p>
<a name="index-lroundmn2"></a>
</dd>
<dt>&lsquo;<samp>lround<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 (valid for floating point mode <var>m</var>) to fixed
point mode <var>n</var> as a signed number rounding to nearest and away
from zero and store in operand 0 (which has mode <var>n</var>).
</p>
<a name="index-lfloormn2"></a>
</dd>
<dt>&lsquo;<samp>lfloor<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 (valid for floating point mode <var>m</var>) to fixed
point mode <var>n</var> as a signed number rounding down and store in
operand 0 (which has mode <var>n</var>).
</p>
<a name="index-lceilmn2"></a>
</dd>
<dt>&lsquo;<samp>lceil<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 (valid for floating point mode <var>m</var>) to fixed
point mode <var>n</var> as a signed number rounding up and store in
operand 0 (which has mode <var>n</var>).
</p>
<a name="index-copysignm3-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>copysign<var>m</var>3</samp>&rsquo;</dt>
<dd><p>Store a value with the magnitude of operand 1 and the sign of operand
2 into operand 0. All operands have mode <var>m</var>, which is a scalar or
vector floating-point mode.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-ffsm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ffs<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store into operand 0 one plus the index of the least significant 1-bit
of operand 1. If operand 1 is zero, store zero.
</p>
<p><var>m</var> is either a scalar or vector integer mode. When it is a scalar,
operand 1 has mode <var>m</var> but operand 0 can have whatever scalar
integer mode is suitable for the target. The compiler will insert
conversion instructions as necessary (typically to convert the result
to the same width as <code>int</code>). When <var>m</var> is a vector, both
operands must have mode <var>m</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-clrsbm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>clrsb<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Count leading redundant sign bits.
Store into operand 0 the number of redundant sign bits in operand 1, starting
at the most significant bit position.
A redundant sign bit is defined as any sign bit after the first. As such,
this count will be one less than the count of leading sign bits.
</p>
<p><var>m</var> is either a scalar or vector integer mode. When it is a scalar,
operand 1 has mode <var>m</var> but operand 0 can have whatever scalar
integer mode is suitable for the target. The compiler will insert
conversion instructions as necessary (typically to convert the result
to the same width as <code>int</code>). When <var>m</var> is a vector, both
operands must have mode <var>m</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-clzm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>clz<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store into operand 0 the number of leading 0-bits in operand 1, starting
at the most significant bit position. If operand 1 is 0, the
<code>CLZ_DEFINED_VALUE_AT_ZERO</code> (see <a href="Misc.html#Misc">Misc</a>) macro defines if
the result is undefined or has a useful value.
</p>
<p><var>m</var> is either a scalar or vector integer mode. When it is a scalar,
operand 1 has mode <var>m</var> but operand 0 can have whatever scalar
integer mode is suitable for the target. The compiler will insert
conversion instructions as necessary (typically to convert the result
to the same width as <code>int</code>). When <var>m</var> is a vector, both
operands must have mode <var>m</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-ctzm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ctz<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store into operand 0 the number of trailing 0-bits in operand 1, starting
at the least significant bit position. If operand 1 is 0, the
<code>CTZ_DEFINED_VALUE_AT_ZERO</code> (see <a href="Misc.html#Misc">Misc</a>) macro defines if
the result is undefined or has a useful value.
</p>
<p><var>m</var> is either a scalar or vector integer mode. When it is a scalar,
operand 1 has mode <var>m</var> but operand 0 can have whatever scalar
integer mode is suitable for the target. The compiler will insert
conversion instructions as necessary (typically to convert the result
to the same width as <code>int</code>). When <var>m</var> is a vector, both
operands must have mode <var>m</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-popcountm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>popcount<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store into operand 0 the number of 1-bits in operand 1.
</p>
<p><var>m</var> is either a scalar or vector integer mode. When it is a scalar,
operand 1 has mode <var>m</var> but operand 0 can have whatever scalar
integer mode is suitable for the target. The compiler will insert
conversion instructions as necessary (typically to convert the result
to the same width as <code>int</code>). When <var>m</var> is a vector, both
operands must have mode <var>m</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-paritym2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>parity<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store into operand 0 the parity of operand 1, i.e. the number of 1-bits
in operand 1 modulo 2.
</p>
<p><var>m</var> is either a scalar or vector integer mode. When it is a scalar,
operand 1 has mode <var>m</var> but operand 0 can have whatever scalar
integer mode is suitable for the target. The compiler will insert
conversion instructions as necessary (typically to convert the result
to the same width as <code>int</code>). When <var>m</var> is a vector, both
operands must have mode <var>m</var>.
</p>
<p>This pattern is not allowed to <code>FAIL</code>.
</p>
<a name="index-one_005fcmplm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>one_cmpl<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Store the bitwise-complement of operand 1 into operand 0.
</p>
<a name="index-movmemm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>movmem<var>m</var></samp>&rsquo;</dt>
<dd><p>Block move instruction. The destination and source blocks of memory
are the first two operands, and both are <code>mem:BLK</code>s with an
address in mode <code>Pmode</code>.
</p>
<p>The number of bytes to move is the third operand, in mode <var>m</var>.
Usually, you specify <code>Pmode</code> for <var>m</var>. However, if you can
generate better code knowing the range of valid lengths is smaller than
those representable in a full Pmode pointer, you should provide
a pattern with a
mode corresponding to the range of values you can handle efficiently
(e.g., <code>QImode</code> for values in the range 0&ndash;127; note we avoid numbers
that appear negative) and also a pattern with <code>Pmode</code>.
</p>
<p>The fourth operand is the known shared alignment of the source and
destination, in the form of a <code>const_int</code> rtx. Thus, if the
compiler knows that both source and destination are word-aligned,
it may provide the value 4 for this operand.
</p>
<p>Optional operands 5 and 6 specify expected alignment and size of block
respectively. The expected alignment differs from alignment in operand 4
in a way that the blocks are not required to be aligned according to it in
all cases. This expected alignment is also in bytes, just like operand 4.
Expected size, when unknown, is set to <code>(const_int -1)</code>.
</p>
<p>Descriptions of multiple <code>movmem<var>m</var></code> patterns can only be
beneficial if the patterns for smaller modes have fewer restrictions
on their first, second and fourth operands. Note that the mode <var>m</var>
in <code>movmem<var>m</var></code> does not impose any restriction on the mode of
individually moved data units in the block.
</p>
<p>These patterns need not give special consideration to the possibility
that the source and destination strings might overlap.
</p>
<a name="index-movstr-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>movstr</samp>&rsquo;</dt>
<dd><p>String copy instruction, with <code>stpcpy</code> semantics. Operand 0 is
an output operand in mode <code>Pmode</code>. The addresses of the
destination and source strings are operands 1 and 2, and both are
<code>mem:BLK</code>s with addresses in mode <code>Pmode</code>. The execution of
the expansion of this pattern should store in operand 0 the address in
which the <code>NUL</code> terminator was stored in the destination string.
</p>
<p>This patern has also several optional operands that are same as in
<code>setmem</code>.
</p>
<a name="index-setmemm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>setmem<var>m</var></samp>&rsquo;</dt>
<dd><p>Block set instruction. The destination string is the first operand,
given as a <code>mem:BLK</code> whose address is in mode <code>Pmode</code>. The
number of bytes to set is the second operand, in mode <var>m</var>. The value to
initialize the memory with is the third operand. Targets that only support the
clearing of memory should reject any value that is not the constant 0. See
&lsquo;<samp>movmem<var>m</var></samp>&rsquo; for a discussion of the choice of mode.
</p>
<p>The fourth operand is the known alignment of the destination, in the form
of a <code>const_int</code> rtx. Thus, if the compiler knows that the
destination is word-aligned, it may provide the value 4 for this
operand.
</p>
<p>Optional operands 5 and 6 specify expected alignment and size of block
respectively. The expected alignment differs from alignment in operand 4
in a way that the blocks are not required to be aligned according to it in
all cases. This expected alignment is also in bytes, just like operand 4.
Expected size, when unknown, is set to <code>(const_int -1)</code>.
Operand 7 is the minimal size of the block and operand 8 is the
maximal size of the block (NULL if it can not be represented as CONST_INT).
Operand 9 is the probable maximal size (i.e. we can not rely on it for correctness,
but it can be used for choosing proper code sequence for a given size).
</p>
<p>The use for multiple <code>setmem<var>m</var></code> is as for <code>movmem<var>m</var></code>.
</p>
<a name="index-cmpstrnm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>cmpstrn<var>m</var></samp>&rsquo;</dt>
<dd><p>String compare instruction, with five operands. Operand 0 is the output;
it has mode <var>m</var>. The remaining four operands are like the operands
of &lsquo;<samp>movmem<var>m</var></samp>&rsquo;. The two memory blocks specified are compared
byte by byte in lexicographic order starting at the beginning of each
string. The instruction is not allowed to prefetch more than one byte
at a time since either string may end in the first byte and reading past
that may access an invalid page or segment and cause a fault. The
comparison terminates early if the fetched bytes are different or if
they are equal to zero. The effect of the instruction is to store a
value in operand 0 whose sign indicates the result of the comparison.
</p>
<a name="index-cmpstrm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>cmpstr<var>m</var></samp>&rsquo;</dt>
<dd><p>String compare instruction, without known maximum length. Operand 0 is the
output; it has mode <var>m</var>. The second and third operand are the blocks of
memory to be compared; both are <code>mem:BLK</code> with an address in mode
<code>Pmode</code>.
</p>
<p>The fourth operand is the known shared alignment of the source and
destination, in the form of a <code>const_int</code> rtx. Thus, if the
compiler knows that both source and destination are word-aligned,
it may provide the value 4 for this operand.
</p>
<p>The two memory blocks specified are compared byte by byte in lexicographic
order starting at the beginning of each string. The instruction is not allowed
to prefetch more than one byte at a time since either string may end in the
first byte and reading past that may access an invalid page or segment and
cause a fault. The comparison will terminate when the fetched bytes
are different or if they are equal to zero. The effect of the
instruction is to store a value in operand 0 whose sign indicates the
result of the comparison.
</p>
<a name="index-cmpmemm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>cmpmem<var>m</var></samp>&rsquo;</dt>
<dd><p>Block compare instruction, with five operands like the operands
of &lsquo;<samp>cmpstr<var>m</var></samp>&rsquo;. The two memory blocks specified are compared
byte by byte in lexicographic order starting at the beginning of each
block. Unlike &lsquo;<samp>cmpstr<var>m</var></samp>&rsquo; the instruction can prefetch
any bytes in the two memory blocks. Also unlike &lsquo;<samp>cmpstr<var>m</var></samp>&rsquo;
the comparison will not stop if both bytes are zero. The effect of
the instruction is to store a value in operand 0 whose sign indicates
the result of the comparison.
</p>
<a name="index-strlenm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>strlen<var>m</var></samp>&rsquo;</dt>
<dd><p>Compute the length of a string, with three operands.
Operand 0 is the result (of mode <var>m</var>), operand 1 is
a <code>mem</code> referring to the first character of the string,
operand 2 is the character to search for (normally zero),
and operand 3 is a constant describing the known alignment
of the beginning of the string.
</p>
<a name="index-floatmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>float<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert signed integer operand 1 (valid for fixed point mode <var>m</var>) to
floating point mode <var>n</var> and store in operand 0 (which has mode
<var>n</var>).
</p>
<a name="index-floatunsmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>floatuns<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert unsigned integer operand 1 (valid for fixed point mode <var>m</var>)
to floating point mode <var>n</var> and store in operand 0 (which has mode
<var>n</var>).
</p>
<a name="index-fixmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fix<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 (valid for floating point mode <var>m</var>) to fixed
point mode <var>n</var> as a signed number and store in operand 0 (which
has mode <var>n</var>). This instruction&rsquo;s result is defined only when
the value of operand 1 is an integer.
</p>
<p>If the machine description defines this pattern, it also needs to
define the <code>ftrunc</code> pattern.
</p>
<a name="index-fixunsmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fixuns<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 (valid for floating point mode <var>m</var>) to fixed
point mode <var>n</var> as an unsigned number and store in operand 0 (which
has mode <var>n</var>). This instruction&rsquo;s result is defined only when the
value of operand 1 is an integer.
</p>
<a name="index-ftruncm2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ftrunc<var>m</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 (valid for floating point mode <var>m</var>) to an
integer value, still represented in floating point mode <var>m</var>, and
store it in operand 0 (valid for floating point mode <var>m</var>).
</p>
<a name="index-fix_005ftruncmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fix_trunc<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>fix<var>m</var><var>n</var>2</samp>&rsquo; but works for any floating point value
of mode <var>m</var> by converting the value to an integer.
</p>
<a name="index-fixuns_005ftruncmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fixuns_trunc<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>fixuns<var>m</var><var>n</var>2</samp>&rsquo; but works for any floating point
value of mode <var>m</var> by converting the value to an integer.
</p>
<a name="index-truncmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>trunc<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Truncate operand 1 (valid for mode <var>m</var>) to mode <var>n</var> and
store in operand 0 (which has mode <var>n</var>). Both modes must be fixed
point or both floating point.
</p>
<a name="index-extendmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>extend<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Sign-extend operand 1 (valid for mode <var>m</var>) to mode <var>n</var> and
store in operand 0 (which has mode <var>n</var>). Both modes must be fixed
point or both floating point.
</p>
<a name="index-zero_005fextendmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>zero_extend<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Zero-extend operand 1 (valid for mode <var>m</var>) to mode <var>n</var> and
store in operand 0 (which has mode <var>n</var>). Both modes must be fixed
point.
</p>
<a name="index-fractmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fract<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 of mode <var>m</var> to mode <var>n</var> and store in
operand 0 (which has mode <var>n</var>). Mode <var>m</var> and mode <var>n</var>
could be fixed-point to fixed-point, signed integer to fixed-point,
fixed-point to signed integer, floating-point to fixed-point,
or fixed-point to floating-point.
When overflows or underflows happen, the results are undefined.
</p>
<a name="index-satfractmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>satfract<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 of mode <var>m</var> to mode <var>n</var> and store in
operand 0 (which has mode <var>n</var>). Mode <var>m</var> and mode <var>n</var>
could be fixed-point to fixed-point, signed integer to fixed-point,
or floating-point to fixed-point.
When overflows or underflows happen, the instruction saturates the
results to the maximum or the minimum.
</p>
<a name="index-fractunsmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>fractuns<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert operand 1 of mode <var>m</var> to mode <var>n</var> and store in
operand 0 (which has mode <var>n</var>). Mode <var>m</var> and mode <var>n</var>
could be unsigned integer to fixed-point, or
fixed-point to unsigned integer.
When overflows or underflows happen, the results are undefined.
</p>
<a name="index-satfractunsmn2-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>satfractuns<var>m</var><var>n</var>2</samp>&rsquo;</dt>
<dd><p>Convert unsigned integer operand 1 of mode <var>m</var> to fixed-point mode
<var>n</var> and store in operand 0 (which has mode <var>n</var>).
When overflows or underflows happen, the instruction saturates the
results to the maximum or the minimum.
</p>
<a name="index-extvm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>extv<var>m</var></samp>&rsquo;</dt>
<dd><p>Extract a bit-field from register operand 1, sign-extend it, and store
it in operand 0. Operand 2 specifies the width of the field in bits
and operand 3 the starting bit, which counts from the most significant
bit if &lsquo;<samp>BITS_BIG_ENDIAN</samp>&rsquo; is true and from the least significant bit
otherwise.
</p>
<p>Operands 0 and 1 both have mode <var>m</var>. Operands 2 and 3 have a
target-specific mode.
</p>
<a name="index-extvmisalignm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>extvmisalign<var>m</var></samp>&rsquo;</dt>
<dd><p>Extract a bit-field from memory operand 1, sign extend it, and store
it in operand 0. Operand 2 specifies the width in bits and operand 3
the starting bit. The starting bit is always somewhere in the first byte of
operand 1; it counts from the most significant bit if &lsquo;<samp>BITS_BIG_ENDIAN</samp>&rsquo;
is true and from the least significant bit otherwise.
</p>
<p>Operand 0 has mode <var>m</var> while operand 1 has <code>BLK</code> mode.
Operands 2 and 3 have a target-specific mode.
</p>
<p>The instruction must not read beyond the last byte of the bit-field.
</p>
<a name="index-extzvm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>extzv<var>m</var></samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>extv<var>m</var></samp>&rsquo; except that the bit-field value is zero-extended.
</p>
<a name="index-extzvmisalignm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>extzvmisalign<var>m</var></samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>extvmisalign<var>m</var></samp>&rsquo; except that the bit-field value is
zero-extended.
</p>
<a name="index-insvm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>insv<var>m</var></samp>&rsquo;</dt>
<dd><p>Insert operand 3 into a bit-field of register operand 0. Operand 1
specifies the width of the field in bits and operand 2 the starting bit,
which counts from the most significant bit if &lsquo;<samp>BITS_BIG_ENDIAN</samp>&rsquo;
is true and from the least significant bit otherwise.
</p>
<p>Operands 0 and 3 both have mode <var>m</var>. Operands 1 and 2 have a
target-specific mode.
</p>
<a name="index-insvmisalignm-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>insvmisalign<var>m</var></samp>&rsquo;</dt>
<dd><p>Insert operand 3 into a bit-field of memory operand 0. Operand 1
specifies the width of the field in bits and operand 2 the starting bit.
The starting bit is always somewhere in the first byte of operand 0;
it counts from the most significant bit if &lsquo;<samp>BITS_BIG_ENDIAN</samp>&rsquo;
is true and from the least significant bit otherwise.
</p>
<p>Operand 3 has mode <var>m</var> while operand 0 has <code>BLK</code> mode.
Operands 1 and 2 have a target-specific mode.
</p>
<p>The instruction must not read or write beyond the last byte of the bit-field.
</p>
<a name="index-extv-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>extv</samp>&rsquo;</dt>
<dd><p>Extract a bit-field from operand 1 (a register or memory operand), where
operand 2 specifies the width in bits and operand 3 the starting bit,
and store it in operand 0. Operand 0 must have mode <code>word_mode</code>.
Operand 1 may have mode <code>byte_mode</code> or <code>word_mode</code>; often
<code>word_mode</code> is allowed only for registers. Operands 2 and 3 must
be valid for <code>word_mode</code>.
</p>
<p>The RTL generation pass generates this instruction only with constants
for operands 2 and 3 and the constant is never zero for operand 2.
</p>
<p>The bit-field value is sign-extended to a full word integer
before it is stored in operand 0.
</p>
<p>This pattern is deprecated; please use &lsquo;<samp>extv<var>m</var></samp>&rsquo; and
<code>extvmisalign<var>m</var></code> instead.
</p>
<a name="index-extzv-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>extzv</samp>&rsquo;</dt>
<dd><p>Like &lsquo;<samp>extv</samp>&rsquo; except that the bit-field value is zero-extended.
</p>
<p>This pattern is deprecated; please use &lsquo;<samp>extzv<var>m</var></samp>&rsquo; and
<code>extzvmisalign<var>m</var></code> instead.
</p>
<a name="index-insv-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>insv</samp>&rsquo;</dt>
<dd><p>Store operand 3 (which must be valid for <code>word_mode</code>) into a
bit-field in operand 0, where operand 1 specifies the width in bits and
operand 2 the starting bit. Operand 0 may have mode <code>byte_mode</code> or
<code>word_mode</code>; often <code>word_mode</code> is allowed only for registers.
Operands 1 and 2 must be valid for <code>word_mode</code>.
</p>
<p>The RTL generation pass generates this instruction only with constants
for operands 1 and 2 and the constant is never zero for operand 1.
</p>
<p>This pattern is deprecated; please use &lsquo;<samp>insv<var>m</var></samp>&rsquo; and
<code>insvmisalign<var>m</var></code> instead.
</p>
<a name="index-movmodecc-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>mov<var>mode</var>cc</samp>&rsquo;</dt>
<dd><p>Conditionally move operand 2 or operand 3 into operand 0 according to the
comparison in operand 1. If the comparison is true, operand 2 is moved
into operand 0, otherwise operand 3 is moved.
</p>
<p>The mode of the operands being compared need not be the same as the operands
being moved. Some machines, sparc64 for example, have instructions that
conditionally move an integer value based on the floating point condition
codes and vice versa.
</p>
<p>If the machine does not have conditional move instructions, do not
define these patterns.
</p>
<a name="index-addmodecc-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>add<var>mode</var>cc</samp>&rsquo;</dt>
<dd><p>Similar to &lsquo;<samp>mov<var>mode</var>cc</samp>&rsquo; but for conditional addition. Conditionally
move operand 2 or (operands 2 + operand 3) into operand 0 according to the
comparison in operand 1. If the comparison is false, operand 2 is moved into
operand 0, otherwise (operand 2 + operand 3) is moved.
</p>
<a name="index-cond_005faddmode-instruction-pattern"></a>
<a name="index-cond_005fsubmode-instruction-pattern"></a>
<a name="index-cond_005fandmode-instruction-pattern"></a>
<a name="index-cond_005fiormode-instruction-pattern"></a>
<a name="index-cond_005fxormode-instruction-pattern"></a>
<a name="index-cond_005fsminmode-instruction-pattern"></a>
<a name="index-cond_005fsmaxmode-instruction-pattern"></a>
<a name="index-cond_005fuminmode-instruction-pattern"></a>
<a name="index-cond_005fumaxmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>cond_add<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>cond_sub<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>cond_and<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>cond_ior<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>cond_xor<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>cond_smin<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>cond_smax<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>cond_umin<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>cond_umax<var>mode</var></samp>&rsquo;</dt>
<dd><p>Perform an elementwise operation on vector operands 2 and 3,
under the control of the vector mask in operand 1, and store the result
in operand 0. This is equivalent to:
</p>
<div class="smallexample">
<pre class="smallexample">for (i = 0; i &lt; GET_MODE_NUNITS (<var>n</var>); i++)
op0[i] = op1[i] ? op2[i] <var>op</var> op3[i] : op2[i];
</pre></div>
<p>where, for example, <var>op</var> is <code>+</code> for &lsquo;<samp>cond_add<var>mode</var></samp>&rsquo;.
</p>
<p>When defined for floating-point modes, the contents of &lsquo;<samp>op3[i]</samp>&rsquo;
are not interpreted if <var>op1[i]</var> is false, just like they would not
be in a normal C &lsquo;<samp>?:</samp>&rsquo; condition.
</p>
<p>Operands 0, 2 and 3 all have mode <var>m</var>, while operand 1 has the mode
returned by <code>TARGET_VECTORIZE_GET_MASK_MODE</code>.
</p>
<a name="index-negmodecc-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>neg<var>mode</var>cc</samp>&rsquo;</dt>
<dd><p>Similar to &lsquo;<samp>mov<var>mode</var>cc</samp>&rsquo; but for conditional negation. Conditionally
move the negation of operand 2 or the unchanged operand 3 into operand 0
according to the comparison in operand 1. If the comparison is true, the negation
of operand 2 is moved into operand 0, otherwise operand 3 is moved.
</p>
<a name="index-notmodecc-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>not<var>mode</var>cc</samp>&rsquo;</dt>
<dd><p>Similar to &lsquo;<samp>neg<var>mode</var>cc</samp>&rsquo; but for conditional complement.
Conditionally move the bitwise complement of operand 2 or the unchanged
operand 3 into operand 0 according to the comparison in operand 1.
If the comparison is true, the complement of operand 2 is moved into
operand 0, otherwise operand 3 is moved.
</p>
<a name="index-cstoremode4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>cstore<var>mode</var>4</samp>&rsquo;</dt>
<dd><p>Store zero or nonzero in operand 0 according to whether a comparison
is true. Operand 1 is a comparison operator. Operand 2 and operand 3
are the first and second operand of the comparison, respectively.
You specify the mode that operand 0 must have when you write the
<code>match_operand</code> expression. The compiler automatically sees which
mode you have used and supplies an operand of that mode.
</p>
<p>The value stored for a true condition must have 1 as its low bit, or
else must be negative. Otherwise the instruction is not suitable and
you should omit it from the machine description. You describe to the
compiler exactly which value is stored by defining the macro
<code>STORE_FLAG_VALUE</code> (see <a href="Misc.html#Misc">Misc</a>). If a description cannot be
found that can be used for all the possible comparison operators, you
should pick one and use a <code>define_expand</code> to map all results
onto the one you chose.
</p>
<p>These operations may <code>FAIL</code>, but should do so only in relatively
uncommon cases; if they would <code>FAIL</code> for common cases involving
integer comparisons, it is best to restrict the predicates to not
allow these operands. Likewise if a given comparison operator will
always fail, independent of the operands (for floating-point modes, the
<code>ordered_comparison_operator</code> predicate is often useful in this case).
</p>
<p>If this pattern is omitted, the compiler will generate a conditional
branch&mdash;for example, it may copy a constant one to the target and branching
around an assignment of zero to the target&mdash;or a libcall. If the predicate
for operand 1 only rejects some operators, it will also try reordering the
operands and/or inverting the result value (e.g. by an exclusive OR).
These possibilities could be cheaper or equivalent to the instructions
used for the &lsquo;<samp>cstore<var>mode</var>4</samp>&rsquo; pattern followed by those required
to convert a positive result from <code>STORE_FLAG_VALUE</code> to 1; in this
case, you can and should make operand 1&rsquo;s predicate reject some operators
in the &lsquo;<samp>cstore<var>mode</var>4</samp>&rsquo; pattern, or remove the pattern altogether
from the machine description.
</p>
<a name="index-cbranchmode4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>cbranch<var>mode</var>4</samp>&rsquo;</dt>
<dd><p>Conditional branch instruction combined with a compare instruction.
Operand 0 is a comparison operator. Operand 1 and operand 2 are the
first and second operands of the comparison, respectively. Operand 3
is the <code>code_label</code> to jump to.
</p>
<a name="index-jump-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>jump</samp>&rsquo;</dt>
<dd><p>A jump inside a function; an unconditional branch. Operand 0 is the
<code>code_label</code> to jump to. This pattern name is mandatory on all
machines.
</p>
<a name="index-call-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>call</samp>&rsquo;</dt>
<dd><p>Subroutine call instruction returning no value. Operand 0 is the
function to call; operand 1 is the number of bytes of arguments pushed
as a <code>const_int</code>; operand 2 is the number of registers used as
operands.
</p>
<p>On most machines, operand 2 is not actually stored into the RTL
pattern. It is supplied for the sake of some RISC machines which need
to put this information into the assembler code; they can put it in
the RTL instead of operand 1.
</p>
<p>Operand 0 should be a <code>mem</code> RTX whose address is the address of the
function. Note, however, that this address can be a <code>symbol_ref</code>
expression even if it would not be a legitimate memory address on the
target machine. If it is also not a valid argument for a call
instruction, the pattern for this operation should be a
<code>define_expand</code> (see <a href="Expander-Definitions.html#Expander-Definitions">Expander Definitions</a>) that places the
address into a register and uses that register in the call instruction.
</p>
<a name="index-call_005fvalue-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>call_value</samp>&rsquo;</dt>
<dd><p>Subroutine call instruction returning a value. Operand 0 is the hard
register in which the value is returned. There are three more
operands, the same as the three operands of the &lsquo;<samp>call</samp>&rsquo;
instruction (but with numbers increased by one).
</p>
<p>Subroutines that return <code>BLKmode</code> objects use the &lsquo;<samp>call</samp>&rsquo;
insn.
</p>
<a name="index-call_005fpop-instruction-pattern"></a>
<a name="index-call_005fvalue_005fpop-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>call_pop</samp>&rsquo;, &lsquo;<samp>call_value_pop</samp>&rsquo;</dt>
<dd><p>Similar to &lsquo;<samp>call</samp>&rsquo; and &lsquo;<samp>call_value</samp>&rsquo;, except used if defined and
if <code>RETURN_POPS_ARGS</code> is nonzero. They should emit a <code>parallel</code>
that contains both the function call and a <code>set</code> to indicate the
adjustment made to the frame pointer.
</p>
<p>For machines where <code>RETURN_POPS_ARGS</code> can be nonzero, the use of these
patterns increases the number of functions for which the frame pointer
can be eliminated, if desired.
</p>
<a name="index-untyped_005fcall-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>untyped_call</samp>&rsquo;</dt>
<dd><p>Subroutine call instruction returning a value of any type. Operand 0 is
the function to call; operand 1 is a memory location where the result of
calling the function is to be stored; operand 2 is a <code>parallel</code>
expression where each element is a <code>set</code> expression that indicates
the saving of a function return value into the result block.
</p>
<p>This instruction pattern should be defined to support
<code>__builtin_apply</code> on machines where special instructions are needed
to call a subroutine with arbitrary arguments or to save the value
returned. This instruction pattern is required on machines that have
multiple registers that can hold a return value
(i.e. <code>FUNCTION_VALUE_REGNO_P</code> is true for more than one register).
</p>
<a name="index-return-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>return</samp>&rsquo;</dt>
<dd><p>Subroutine return instruction. This instruction pattern name should be
defined only if a single instruction can do all the work of returning
from a function.
</p>
<p>Like the &lsquo;<samp>mov<var>m</var></samp>&rsquo; patterns, this pattern is also used after the
RTL generation phase. In this case it is to support machines where
multiple instructions are usually needed to return from a function, but
some class of functions only requires one instruction to implement a
return. Normally, the applicable functions are those which do not need
to save any registers or allocate stack space.
</p>
<p>It is valid for this pattern to expand to an instruction using
<code>simple_return</code> if no epilogue is required.
</p>
<a name="index-simple_005freturn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>simple_return</samp>&rsquo;</dt>
<dd><p>Subroutine return instruction. This instruction pattern name should be
defined only if a single instruction can do all the work of returning
from a function on a path where no epilogue is required. This pattern
is very similar to the <code>return</code> instruction pattern, but it is emitted
only by the shrink-wrapping optimization on paths where the function
prologue has not been executed, and a function return should occur without
any of the effects of the epilogue. Additional uses may be introduced on
paths where both the prologue and the epilogue have executed.
</p>
<a name="index-reload_005fcompleted"></a>
<a name="index-leaf_005ffunction_005fp"></a>
<p>For such machines, the condition specified in this pattern should only
be true when <code>reload_completed</code> is nonzero and the function&rsquo;s
epilogue would only be a single instruction. For machines with register
windows, the routine <code>leaf_function_p</code> may be used to determine if
a register window push is required.
</p>
<p>Machines that have conditional return instructions should define patterns
such as
</p>
<div class="smallexample">
<pre class="smallexample">(define_insn &quot;&quot;
[(set (pc)
(if_then_else (match_operator
0 &quot;comparison_operator&quot;
[(cc0) (const_int 0)])
(return)
(pc)))]
&quot;<var>condition</var>&quot;
&quot;&hellip;&quot;)
</pre></div>
<p>where <var>condition</var> would normally be the same condition specified on the
named &lsquo;<samp>return</samp>&rsquo; pattern.
</p>
<a name="index-untyped_005freturn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>untyped_return</samp>&rsquo;</dt>
<dd><p>Untyped subroutine return instruction. This instruction pattern should
be defined to support <code>__builtin_return</code> on machines where special
instructions are needed to return a value of any type.
</p>
<p>Operand 0 is a memory location where the result of calling a function
with <code>__builtin_apply</code> is stored; operand 1 is a <code>parallel</code>
expression where each element is a <code>set</code> expression that indicates
the restoring of a function return value from the result block.
</p>
<a name="index-nop-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>nop</samp>&rsquo;</dt>
<dd><p>No-op instruction. This instruction pattern name should always be defined
to output a no-op in assembler code. <code>(const_int 0)</code> will do as an
RTL pattern.
</p>
<a name="index-indirect_005fjump-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>indirect_jump</samp>&rsquo;</dt>
<dd><p>An instruction to jump to an address which is operand zero.
This pattern name is mandatory on all machines.
</p>
<a name="index-casesi-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>casesi</samp>&rsquo;</dt>
<dd><p>Instruction to jump through a dispatch table, including bounds checking.
This instruction takes five operands:
</p>
<ol>
<li> The index to dispatch on, which has mode <code>SImode</code>.
</li><li> The lower bound for indices in the table, an integer constant.
</li><li> The total range of indices in the table&mdash;the largest index
minus the smallest one (both inclusive).
</li><li> A label that precedes the table itself.
</li><li> A label to jump to if the index has a value outside the bounds.
</li></ol>
<p>The table is an <code>addr_vec</code> or <code>addr_diff_vec</code> inside of a
<code>jump_table_data</code>. The number of elements in the table is one plus the
difference between the upper bound and the lower bound.
</p>
<a name="index-tablejump-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>tablejump</samp>&rsquo;</dt>
<dd><p>Instruction to jump to a variable address. This is a low-level
capability which can be used to implement a dispatch table when there
is no &lsquo;<samp>casesi</samp>&rsquo; pattern.
</p>
<p>This pattern requires two operands: the address or offset, and a label
which should immediately precede the jump table. If the macro
<code>CASE_VECTOR_PC_RELATIVE</code> evaluates to a nonzero value then the first
operand is an offset which counts from the address of the table; otherwise,
it is an absolute address to jump to. In either case, the first operand has
mode <code>Pmode</code>.
</p>
<p>The &lsquo;<samp>tablejump</samp>&rsquo; insn is always the last insn before the jump
table it uses. Its assembler code normally has no need to use the
second operand, but you should incorporate it in the RTL pattern so
that the jump optimizer will not delete the table as unreachable code.
</p>
<a name="index-decrement_005fand_005fbranch_005funtil_005fzero-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>decrement_and_branch_until_zero</samp>&rsquo;</dt>
<dd><p>Conditional branch instruction that decrements a register and
jumps if the register is nonzero. Operand 0 is the register to
decrement and test; operand 1 is the label to jump to if the
register is nonzero. See <a href="Looping-Patterns.html#Looping-Patterns">Looping Patterns</a>.
</p>
<p>This optional instruction pattern is only used by the combiner,
typically for loops reversed by the loop optimizer when strength
reduction is enabled.
</p>
<a name="index-doloop_005fend-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>doloop_end</samp>&rsquo;</dt>
<dd><p>Conditional branch instruction that decrements a register and
jumps if the register is nonzero. Operand 0 is the register to
decrement and test; operand 1 is the label to jump to if the
register is nonzero.
See <a href="Looping-Patterns.html#Looping-Patterns">Looping Patterns</a>.
</p>
<p>This optional instruction pattern should be defined for machines with
low-overhead looping instructions as the loop optimizer will try to
modify suitable loops to utilize it. The target hook
<code>TARGET_CAN_USE_DOLOOP_P</code> controls the conditions under which
low-overhead loops can be used.
</p>
<a name="index-doloop_005fbegin-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>doloop_begin</samp>&rsquo;</dt>
<dd><p>Companion instruction to <code>doloop_end</code> required for machines that
need to perform some initialization, such as loading a special counter
register. Operand 1 is the associated <code>doloop_end</code> pattern and
operand 0 is the register that it decrements.
</p>
<p>If initialization insns do not always need to be emitted, use a
<code>define_expand</code> (see <a href="Expander-Definitions.html#Expander-Definitions">Expander Definitions</a>) and make it fail.
</p>
<a name="index-canonicalize_005ffuncptr_005ffor_005fcompare-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>canonicalize_funcptr_for_compare</samp>&rsquo;</dt>
<dd><p>Canonicalize the function pointer in operand 1 and store the result
into operand 0.
</p>
<p>Operand 0 is always a <code>reg</code> and has mode <code>Pmode</code>; operand 1
may be a <code>reg</code>, <code>mem</code>, <code>symbol_ref</code>, <code>const_int</code>, etc
and also has mode <code>Pmode</code>.
</p>
<p>Canonicalization of a function pointer usually involves computing
the address of the function which would be called if the function
pointer were used in an indirect call.
</p>
<p>Only define this pattern if function pointers on the target machine
can have different values but still call the same function when
used in an indirect call.
</p>
<a name="index-save_005fstack_005fblock-instruction-pattern"></a>
<a name="index-save_005fstack_005ffunction-instruction-pattern"></a>
<a name="index-save_005fstack_005fnonlocal-instruction-pattern"></a>
<a name="index-restore_005fstack_005fblock-instruction-pattern"></a>
<a name="index-restore_005fstack_005ffunction-instruction-pattern"></a>
<a name="index-restore_005fstack_005fnonlocal-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>save_stack_block</samp>&rsquo;</dt>
<dt>&lsquo;<samp>save_stack_function</samp>&rsquo;</dt>
<dt>&lsquo;<samp>save_stack_nonlocal</samp>&rsquo;</dt>
<dt>&lsquo;<samp>restore_stack_block</samp>&rsquo;</dt>
<dt>&lsquo;<samp>restore_stack_function</samp>&rsquo;</dt>
<dt>&lsquo;<samp>restore_stack_nonlocal</samp>&rsquo;</dt>
<dd><p>Most machines save and restore the stack pointer by copying it to or
from an object of mode <code>Pmode</code>. Do not define these patterns on
such machines.
</p>
<p>Some machines require special handling for stack pointer saves and
restores. On those machines, define the patterns corresponding to the
non-standard cases by using a <code>define_expand</code> (see <a href="Expander-Definitions.html#Expander-Definitions">Expander Definitions</a>) that produces the required insns. The three types of
saves and restores are:
</p>
<ol>
<li> &lsquo;<samp>save_stack_block</samp>&rsquo; saves the stack pointer at the start of a block
that allocates a variable-sized object, and &lsquo;<samp>restore_stack_block</samp>&rsquo;
restores the stack pointer when the block is exited.
</li><li> &lsquo;<samp>save_stack_function</samp>&rsquo; and &lsquo;<samp>restore_stack_function</samp>&rsquo; do a
similar job for the outermost block of a function and are used when the
function allocates variable-sized objects or calls <code>alloca</code>. Only
the epilogue uses the restored stack pointer, allowing a simpler save or
restore sequence on some machines.
</li><li> &lsquo;<samp>save_stack_nonlocal</samp>&rsquo; is used in functions that contain labels
branched to by nested functions. It saves the stack pointer in such a
way that the inner function can use &lsquo;<samp>restore_stack_nonlocal</samp>&rsquo; to
restore the stack pointer. The compiler generates code to restore the
frame and argument pointer registers, but some machines require saving
and restoring additional data such as register window information or
stack backchains. Place insns in these patterns to save and restore any
such required data.
</li></ol>
<p>When saving the stack pointer, operand 0 is the save area and operand 1
is the stack pointer. The mode used to allocate the save area defaults
to <code>Pmode</code> but you can override that choice by defining the
<code>STACK_SAVEAREA_MODE</code> macro (see <a href="Storage-Layout.html#Storage-Layout">Storage Layout</a>). You must
specify an integral mode, or <code>VOIDmode</code> if no save area is needed
for a particular type of save (either because no save is needed or
because a machine-specific save area can be used). Operand 0 is the
stack pointer and operand 1 is the save area for restore operations. If
&lsquo;<samp>save_stack_block</samp>&rsquo; is defined, operand 0 must not be
<code>VOIDmode</code> since these saves can be arbitrarily nested.
</p>
<p>A save area is a <code>mem</code> that is at a constant offset from
<code>virtual_stack_vars_rtx</code> when the stack pointer is saved for use by
nonlocal gotos and a <code>reg</code> in the other two cases.
</p>
<a name="index-allocate_005fstack-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>allocate_stack</samp>&rsquo;</dt>
<dd><p>Subtract (or add if <code>STACK_GROWS_DOWNWARD</code> is undefined) operand 1 from
the stack pointer to create space for dynamically allocated data.
</p>
<p>Store the resultant pointer to this space into operand 0. If you
are allocating space from the main stack, do this by emitting a
move insn to copy <code>virtual_stack_dynamic_rtx</code> to operand 0.
If you are allocating the space elsewhere, generate code to copy the
location of the space to operand 0. In the latter case, you must
ensure this space gets freed when the corresponding space on the main
stack is free.
</p>
<p>Do not define this pattern if all that must be done is the subtraction.
Some machines require other operations such as stack probes or
maintaining the back chain. Define this pattern to emit those
operations in addition to updating the stack pointer.
</p>
<a name="index-check_005fstack-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>check_stack</samp>&rsquo;</dt>
<dd><p>If stack checking (see <a href="Stack-Checking.html#Stack-Checking">Stack Checking</a>) cannot be done on your system by
probing the stack, define this pattern to perform the needed check and signal
an error if the stack has overflowed. The single operand is the address in
the stack farthest from the current stack pointer that you need to validate.
Normally, on platforms where this pattern is needed, you would obtain the
stack limit from a global or thread-specific variable or register.
</p>
<a name="index-probe_005fstack_005faddress-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>probe_stack_address</samp>&rsquo;</dt>
<dd><p>If stack checking (see <a href="Stack-Checking.html#Stack-Checking">Stack Checking</a>) can be done on your system by
probing the stack but without the need to actually access it, define this
pattern and signal an error if the stack has overflowed. The single operand
is the memory address in the stack that needs to be probed.
</p>
<a name="index-probe_005fstack-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>probe_stack</samp>&rsquo;</dt>
<dd><p>If stack checking (see <a href="Stack-Checking.html#Stack-Checking">Stack Checking</a>) can be done on your system by
probing the stack but doing it with a &ldquo;store zero&rdquo; instruction is not valid
or optimal, define this pattern to do the probing differently and signal an
error if the stack has overflowed. The single operand is the memory reference
in the stack that needs to be probed.
</p>
<a name="index-nonlocal_005fgoto-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>nonlocal_goto</samp>&rsquo;</dt>
<dd><p>Emit code to generate a non-local goto, e.g., a jump from one function
to a label in an outer function. This pattern has four arguments,
each representing a value to be used in the jump. The first
argument is to be loaded into the frame pointer, the second is
the address to branch to (code to dispatch to the actual label),
the third is the address of a location where the stack is saved,
and the last is the address of the label, to be placed in the
location for the incoming static chain.
</p>
<p>On most machines you need not define this pattern, since GCC will
already generate the correct code, which is to load the frame pointer
and static chain, restore the stack (using the
&lsquo;<samp>restore_stack_nonlocal</samp>&rsquo; pattern, if defined), and jump indirectly
to the dispatcher. You need only define this pattern if this code will
not work on your machine.
</p>
<a name="index-nonlocal_005fgoto_005freceiver-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>nonlocal_goto_receiver</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, contains code needed at the target of a
nonlocal goto after the code already generated by GCC. You will not
normally need to define this pattern. A typical reason why you might
need this pattern is if some value, such as a pointer to a global table,
must be restored when the frame pointer is restored. Note that a nonlocal
goto only occurs within a unit-of-translation, so a global table pointer
that is shared by all functions of a given module need not be restored.
There are no arguments.
</p>
<a name="index-exception_005freceiver-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>exception_receiver</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, contains code needed at the site of an
exception handler that isn&rsquo;t needed at the site of a nonlocal goto. You
will not normally need to define this pattern. A typical reason why you
might need this pattern is if some value, such as a pointer to a global
table, must be restored after control flow is branched to the handler of
an exception. There are no arguments.
</p>
<a name="index-builtin_005fsetjmp_005fsetup-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>builtin_setjmp_setup</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, contains additional code needed to initialize
the <code>jmp_buf</code>. You will not normally need to define this pattern.
A typical reason why you might need this pattern is if some value, such
as a pointer to a global table, must be restored. Though it is
preferred that the pointer value be recalculated if possible (given the
address of a label for instance). The single argument is a pointer to
the <code>jmp_buf</code>. Note that the buffer is five words long and that
the first three are normally used by the generic mechanism.
</p>
<a name="index-builtin_005fsetjmp_005freceiver-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>builtin_setjmp_receiver</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, contains code needed at the site of a
built-in setjmp that isn&rsquo;t needed at the site of a nonlocal goto. You
will not normally need to define this pattern. A typical reason why you
might need this pattern is if some value, such as a pointer to a global
table, must be restored. It takes one argument, which is the label
to which builtin_longjmp transferred control; this pattern may be emitted
at a small offset from that label.
</p>
<a name="index-builtin_005flongjmp-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>builtin_longjmp</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, performs the entire action of the longjmp.
You will not normally need to define this pattern unless you also define
<code>builtin_setjmp_setup</code>. The single argument is a pointer to the
<code>jmp_buf</code>.
</p>
<a name="index-eh_005freturn-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>eh_return</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, affects the way <code>__builtin_eh_return</code>,
and thence the call frame exception handling library routines, are
built. It is intended to handle non-trivial actions needed along
the abnormal return path.
</p>
<p>The address of the exception handler to which the function should return
is passed as operand to this pattern. It will normally need to copied by
the pattern to some special register or memory location.
If the pattern needs to determine the location of the target call
frame in order to do so, it may use <code>EH_RETURN_STACKADJ_RTX</code>,
if defined; it will have already been assigned.
</p>
<p>If this pattern is not defined, the default action will be to simply
copy the return address to <code>EH_RETURN_HANDLER_RTX</code>. Either
that macro or this pattern needs to be defined if call frame exception
handling is to be used.
</p>
<a name="index-prologue-instruction-pattern"></a>
<a name="prologue-instruction-pattern"></a></dd>
<dt>&lsquo;<samp>prologue</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, emits RTL for entry to a function. The function
entry is responsible for setting up the stack frame, initializing the frame
pointer register, saving callee saved registers, etc.
</p>
<p>Using a prologue pattern is generally preferred over defining
<code>TARGET_ASM_FUNCTION_PROLOGUE</code> to emit assembly code for the prologue.
</p>
<p>The <code>prologue</code> pattern is particularly useful for targets which perform
instruction scheduling.
</p>
<a name="index-window_005fsave-instruction-pattern"></a>
<a name="window_005fsave-instruction-pattern"></a></dd>
<dt>&lsquo;<samp>window_save</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, emits RTL for a register window save. It should
be defined if the target machine has register windows but the window events
are decoupled from calls to subroutines. The canonical example is the SPARC
architecture.
</p>
<a name="index-epilogue-instruction-pattern"></a>
<a name="epilogue-instruction-pattern"></a></dd>
<dt>&lsquo;<samp>epilogue</samp>&rsquo;</dt>
<dd><p>This pattern emits RTL for exit from a function. The function
exit is responsible for deallocating the stack frame, restoring callee saved
registers and emitting the return instruction.
</p>
<p>Using an epilogue pattern is generally preferred over defining
<code>TARGET_ASM_FUNCTION_EPILOGUE</code> to emit assembly code for the epilogue.
</p>
<p>The <code>epilogue</code> pattern is particularly useful for targets which perform
instruction scheduling or which have delay slots for their return instruction.
</p>
<a name="index-sibcall_005fepilogue-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sibcall_epilogue</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, emits RTL for exit from a function without the final
branch back to the calling function. This pattern will be emitted before any
sibling call (aka tail call) sites.
</p>
<p>The <code>sibcall_epilogue</code> pattern must not clobber any arguments used for
parameter passing or any stack slots for arguments passed to the current
function.
</p>
<a name="index-trap-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>trap</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, signals an error, typically by causing some
kind of signal to be raised.
</p>
<a name="index-ctrapMM4-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>ctrap<var>MM</var>4</samp>&rsquo;</dt>
<dd><p>Conditional trap instruction. Operand 0 is a piece of RTL which
performs a comparison, and operands 1 and 2 are the arms of the
comparison. Operand 3 is the trap code, an integer.
</p>
<p>A typical <code>ctrap</code> pattern looks like
</p>
<div class="smallexample">
<pre class="smallexample">(define_insn &quot;ctrapsi4&quot;
[(trap_if (match_operator 0 &quot;trap_operator&quot;
[(match_operand 1 &quot;register_operand&quot;)
(match_operand 2 &quot;immediate_operand&quot;)])
(match_operand 3 &quot;const_int_operand&quot; &quot;i&quot;))]
&quot;&quot;
&quot;&hellip;&quot;)
</pre></div>
<a name="index-prefetch-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>prefetch</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, emits code for a non-faulting data prefetch
instruction. Operand 0 is the address of the memory to prefetch. Operand 1
is a constant 1 if the prefetch is preparing for a write to the memory
address, or a constant 0 otherwise. Operand 2 is the expected degree of
temporal locality of the data and is a value between 0 and 3, inclusive; 0
means that the data has no temporal locality, so it need not be left in the
cache after the access; 3 means that the data has a high degree of temporal
locality and should be left in all levels of cache possible; 1 and 2 mean,
respectively, a low or moderate degree of temporal locality.
</p>
<p>Targets that do not support write prefetches or locality hints can ignore
the values of operands 1 and 2.
</p>
<a name="index-blockage-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>blockage</samp>&rsquo;</dt>
<dd><p>This pattern defines a pseudo insn that prevents the instruction
scheduler and other passes from moving instructions and using register
equivalences across the boundary defined by the blockage insn.
This needs to be an UNSPEC_VOLATILE pattern or a volatile ASM.
</p>
<a name="index-memory_005fblockage-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>memory_blockage</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, represents a compiler memory barrier, and will be
placed at points across which RTL passes may not propagate memory accesses.
This instruction needs to read and write volatile BLKmode memory. It does
not need to generate any machine instruction. If this pattern is not defined,
the compiler falls back to emitting an instruction corresponding
to <code>asm volatile (&quot;&quot; ::: &quot;memory&quot;)</code>.
</p>
<a name="index-memory_005fbarrier-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>memory_barrier</samp>&rsquo;</dt>
<dd><p>If the target memory model is not fully synchronous, then this pattern
should be defined to an instruction that orders both loads and stores
before the instruction with respect to loads and stores after the instruction.
This pattern has no operands.
</p>
<a name="index-sync_005fcompare_005fand_005fswapmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sync_compare_and_swap<var>mode</var></samp>&rsquo;</dt>
<dd><p>This pattern, if defined, emits code for an atomic compare-and-swap
operation. Operand 1 is the memory on which the atomic operation is
performed. Operand 2 is the &ldquo;old&rdquo; value to be compared against the
current contents of the memory location. Operand 3 is the &ldquo;new&rdquo; value
to store in the memory if the compare succeeds. Operand 0 is the result
of the operation; it should contain the contents of the memory
before the operation. If the compare succeeds, this should obviously be
a copy of operand 2.
</p>
<p>This pattern must show that both operand 0 and operand 1 are modified.
</p>
<p>This pattern must issue any memory barrier instructions such that all
memory operations before the atomic operation occur before the atomic
operation and all memory operations after the atomic operation occur
after the atomic operation.
</p>
<p>For targets where the success or failure of the compare-and-swap
operation is available via the status flags, it is possible to
avoid a separate compare operation and issue the subsequent
branch or store-flag operation immediately after the compare-and-swap.
To this end, GCC will look for a <code>MODE_CC</code> set in the
output of <code>sync_compare_and_swap<var>mode</var></code>; if the machine
description includes such a set, the target should also define special
<code>cbranchcc4</code> and/or <code>cstorecc4</code> instructions. GCC will then
be able to take the destination of the <code>MODE_CC</code> set and pass it
to the <code>cbranchcc4</code> or <code>cstorecc4</code> pattern as the first
operand of the comparison (the second will be <code>(const_int 0)</code>).
</p>
<p>For targets where the operating system may provide support for this
operation via library calls, the <code>sync_compare_and_swap_optab</code>
may be initialized to a function with the same interface as the
<code>__sync_val_compare_and_swap_<var>n</var></code> built-in. If the entire
set of <var>__sync</var> builtins are supported via library calls, the
target can initialize all of the optabs at once with
<code>init_sync_libfuncs</code>.
For the purposes of C++11 <code>std::atomic::is_lock_free</code>, it is
assumed that these library calls do <em>not</em> use any kind of
interruptable locking.
</p>
<a name="index-sync_005faddmode-instruction-pattern"></a>
<a name="index-sync_005fsubmode-instruction-pattern"></a>
<a name="index-sync_005fiormode-instruction-pattern"></a>
<a name="index-sync_005fandmode-instruction-pattern"></a>
<a name="index-sync_005fxormode-instruction-pattern"></a>
<a name="index-sync_005fnandmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sync_add<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_sub<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>sync_ior<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_and<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>sync_xor<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_nand<var>mode</var></samp>&rsquo;</dt>
<dd><p>These patterns emit code for an atomic operation on memory.
Operand 0 is the memory on which the atomic operation is performed.
Operand 1 is the second operand to the binary operator.
</p>
<p>This pattern must issue any memory barrier instructions such that all
memory operations before the atomic operation occur before the atomic
operation and all memory operations after the atomic operation occur
after the atomic operation.
</p>
<p>If these patterns are not defined, the operation will be constructed
from a compare-and-swap operation, if defined.
</p>
<a name="index-sync_005fold_005faddmode-instruction-pattern"></a>
<a name="index-sync_005fold_005fsubmode-instruction-pattern"></a>
<a name="index-sync_005fold_005fiormode-instruction-pattern"></a>
<a name="index-sync_005fold_005fandmode-instruction-pattern"></a>
<a name="index-sync_005fold_005fxormode-instruction-pattern"></a>
<a name="index-sync_005fold_005fnandmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sync_old_add<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_old_sub<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>sync_old_ior<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_old_and<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>sync_old_xor<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_old_nand<var>mode</var></samp>&rsquo;</dt>
<dd><p>These patterns emit code for an atomic operation on memory,
and return the value that the memory contained before the operation.
Operand 0 is the result value, operand 1 is the memory on which the
atomic operation is performed, and operand 2 is the second operand
to the binary operator.
</p>
<p>This pattern must issue any memory barrier instructions such that all
memory operations before the atomic operation occur before the atomic
operation and all memory operations after the atomic operation occur
after the atomic operation.
</p>
<p>If these patterns are not defined, the operation will be constructed
from a compare-and-swap operation, if defined.
</p>
<a name="index-sync_005fnew_005faddmode-instruction-pattern"></a>
<a name="index-sync_005fnew_005fsubmode-instruction-pattern"></a>
<a name="index-sync_005fnew_005fiormode-instruction-pattern"></a>
<a name="index-sync_005fnew_005fandmode-instruction-pattern"></a>
<a name="index-sync_005fnew_005fxormode-instruction-pattern"></a>
<a name="index-sync_005fnew_005fnandmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sync_new_add<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_new_sub<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>sync_new_ior<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_new_and<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>sync_new_xor<var>mode</var></samp>&rsquo;, &lsquo;<samp>sync_new_nand<var>mode</var></samp>&rsquo;</dt>
<dd><p>These patterns are like their <code>sync_old_<var>op</var></code> counterparts,
except that they return the value that exists in the memory location
after the operation, rather than before the operation.
</p>
<a name="index-sync_005flock_005ftest_005fand_005fsetmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sync_lock_test_and_set<var>mode</var></samp>&rsquo;</dt>
<dd><p>This pattern takes two forms, based on the capabilities of the target.
In either case, operand 0 is the result of the operand, operand 1 is
the memory on which the atomic operation is performed, and operand 2
is the value to set in the lock.
</p>
<p>In the ideal case, this operation is an atomic exchange operation, in
which the previous value in memory operand is copied into the result
operand, and the value operand is stored in the memory operand.
</p>
<p>For less capable targets, any value operand that is not the constant 1
should be rejected with <code>FAIL</code>. In this case the target may use
an atomic test-and-set bit operation. The result operand should contain
1 if the bit was previously set and 0 if the bit was previously clear.
The true contents of the memory operand are implementation defined.
</p>
<p>This pattern must issue any memory barrier instructions such that the
pattern as a whole acts as an acquire barrier, that is all memory
operations after the pattern do not occur until the lock is acquired.
</p>
<p>If this pattern is not defined, the operation will be constructed from
a compare-and-swap operation, if defined.
</p>
<a name="index-sync_005flock_005freleasemode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>sync_lock_release<var>mode</var></samp>&rsquo;</dt>
<dd><p>This pattern, if defined, releases a lock set by
<code>sync_lock_test_and_set<var>mode</var></code>. Operand 0 is the memory
that contains the lock; operand 1 is the value to store in the lock.
</p>
<p>If the target doesn&rsquo;t implement full semantics for
<code>sync_lock_test_and_set<var>mode</var></code>, any value operand which is not
the constant 0 should be rejected with <code>FAIL</code>, and the true contents
of the memory operand are implementation defined.
</p>
<p>This pattern must issue any memory barrier instructions such that the
pattern as a whole acts as a release barrier, that is the lock is
released only after all previous memory operations have completed.
</p>
<p>If this pattern is not defined, then a <code>memory_barrier</code> pattern
will be emitted, followed by a store of the value to the memory operand.
</p>
<a name="index-atomic_005fcompare_005fand_005fswapmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_compare_and_swap<var>mode</var></samp>&rsquo;</dt>
<dd><p>This pattern, if defined, emits code for an atomic compare-and-swap
operation with memory model semantics. Operand 2 is the memory on which
the atomic operation is performed. Operand 0 is an output operand which
is set to true or false based on whether the operation succeeded. Operand
1 is an output operand which is set to the contents of the memory before
the operation was attempted. Operand 3 is the value that is expected to
be in memory. Operand 4 is the value to put in memory if the expected
value is found there. Operand 5 is set to 1 if this compare and swap is to
be treated as a weak operation. Operand 6 is the memory model to be used
if the operation is a success. Operand 7 is the memory model to be used
if the operation fails.
</p>
<p>If memory referred to in operand 2 contains the value in operand 3, then
operand 4 is stored in memory pointed to by operand 2 and fencing based on
the memory model in operand 6 is issued.
</p>
<p>If memory referred to in operand 2 does not contain the value in operand 3,
then fencing based on the memory model in operand 7 is issued.
</p>
<p>If a target does not support weak compare-and-swap operations, or the port
elects not to implement weak operations, the argument in operand 5 can be
ignored. Note a strong implementation must be provided.
</p>
<p>If this pattern is not provided, the <code>__atomic_compare_exchange</code>
built-in functions will utilize the legacy <code>sync_compare_and_swap</code>
pattern with an <code>__ATOMIC_SEQ_CST</code> memory model.
</p>
<a name="index-atomic_005floadmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_load<var>mode</var></samp>&rsquo;</dt>
<dd><p>This pattern implements an atomic load operation with memory model
semantics. Operand 1 is the memory address being loaded from. Operand 0
is the result of the load. Operand 2 is the memory model to be used for
the load operation.
</p>
<p>If not present, the <code>__atomic_load</code> built-in function will either
resort to a normal load with memory barriers, or a compare-and-swap
operation if a normal load would not be atomic.
</p>
<a name="index-atomic_005fstoremode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_store<var>mode</var></samp>&rsquo;</dt>
<dd><p>This pattern implements an atomic store operation with memory model
semantics. Operand 0 is the memory address being stored to. Operand 1
is the value to be written. Operand 2 is the memory model to be used for
the operation.
</p>
<p>If not present, the <code>__atomic_store</code> built-in function will attempt to
perform a normal store and surround it with any required memory fences. If
the store would not be atomic, then an <code>__atomic_exchange</code> is
attempted with the result being ignored.
</p>
<a name="index-atomic_005fexchangemode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_exchange<var>mode</var></samp>&rsquo;</dt>
<dd><p>This pattern implements an atomic exchange operation with memory model
semantics. Operand 1 is the memory location the operation is performed on.
Operand 0 is an output operand which is set to the original value contained
in the memory pointed to by operand 1. Operand 2 is the value to be
stored. Operand 3 is the memory model to be used.
</p>
<p>If this pattern is not present, the built-in function
<code>__atomic_exchange</code> will attempt to preform the operation with a
compare and swap loop.
</p>
<a name="index-atomic_005faddmode-instruction-pattern"></a>
<a name="index-atomic_005fsubmode-instruction-pattern"></a>
<a name="index-atomic_005formode-instruction-pattern"></a>
<a name="index-atomic_005fandmode-instruction-pattern"></a>
<a name="index-atomic_005fxormode-instruction-pattern"></a>
<a name="index-atomic_005fnandmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_add<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_sub<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>atomic_or<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_and<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>atomic_xor<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_nand<var>mode</var></samp>&rsquo;</dt>
<dd><p>These patterns emit code for an atomic operation on memory with memory
model semantics. Operand 0 is the memory on which the atomic operation is
performed. Operand 1 is the second operand to the binary operator.
Operand 2 is the memory model to be used by the operation.
</p>
<p>If these patterns are not defined, attempts will be made to use legacy
<code>sync</code> patterns, or equivalent patterns which return a result. If
none of these are available a compare-and-swap loop will be used.
</p>
<a name="index-atomic_005ffetch_005faddmode-instruction-pattern"></a>
<a name="index-atomic_005ffetch_005fsubmode-instruction-pattern"></a>
<a name="index-atomic_005ffetch_005formode-instruction-pattern"></a>
<a name="index-atomic_005ffetch_005fandmode-instruction-pattern"></a>
<a name="index-atomic_005ffetch_005fxormode-instruction-pattern"></a>
<a name="index-atomic_005ffetch_005fnandmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_fetch_add<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_fetch_sub<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>atomic_fetch_or<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_fetch_and<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>atomic_fetch_xor<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_fetch_nand<var>mode</var></samp>&rsquo;</dt>
<dd><p>These patterns emit code for an atomic operation on memory with memory
model semantics, and return the original value. Operand 0 is an output
operand which contains the value of the memory location before the
operation was performed. Operand 1 is the memory on which the atomic
operation is performed. Operand 2 is the second operand to the binary
operator. Operand 3 is the memory model to be used by the operation.
</p>
<p>If these patterns are not defined, attempts will be made to use legacy
<code>sync</code> patterns. If none of these are available a compare-and-swap
loop will be used.
</p>
<a name="index-atomic_005fadd_005ffetchmode-instruction-pattern"></a>
<a name="index-atomic_005fsub_005ffetchmode-instruction-pattern"></a>
<a name="index-atomic_005for_005ffetchmode-instruction-pattern"></a>
<a name="index-atomic_005fand_005ffetchmode-instruction-pattern"></a>
<a name="index-atomic_005fxor_005ffetchmode-instruction-pattern"></a>
<a name="index-atomic_005fnand_005ffetchmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_add_fetch<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_sub_fetch<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>atomic_or_fetch<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_and_fetch<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>atomic_xor_fetch<var>mode</var></samp>&rsquo;, &lsquo;<samp>atomic_nand_fetch<var>mode</var></samp>&rsquo;</dt>
<dd><p>These patterns emit code for an atomic operation on memory with memory
model semantics and return the result after the operation is performed.
Operand 0 is an output operand which contains the value after the
operation. Operand 1 is the memory on which the atomic operation is
performed. Operand 2 is the second operand to the binary operator.
Operand 3 is the memory model to be used by the operation.
</p>
<p>If these patterns are not defined, attempts will be made to use legacy
<code>sync</code> patterns, or equivalent patterns which return the result before
the operation followed by the arithmetic operation required to produce the
result. If none of these are available a compare-and-swap loop will be
used.
</p>
<a name="index-atomic_005ftest_005fand_005fset-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_test_and_set</samp>&rsquo;</dt>
<dd><p>This pattern emits code for <code>__builtin_atomic_test_and_set</code>.
Operand 0 is an output operand which is set to true if the previous
previous contents of the byte was &quot;set&quot;, and false otherwise. Operand 1
is the <code>QImode</code> memory to be modified. Operand 2 is the memory
model to be used.
</p>
<p>The specific value that defines &quot;set&quot; is implementation defined, and
is normally based on what is performed by the native atomic test and set
instruction.
</p>
<a name="index-atomic_005fbit_005ftest_005fand_005fsetmode-instruction-pattern"></a>
<a name="index-atomic_005fbit_005ftest_005fand_005fcomplementmode-instruction-pattern"></a>
<a name="index-atomic_005fbit_005ftest_005fand_005fresetmode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>atomic_bit_test_and_set<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>atomic_bit_test_and_complement<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>atomic_bit_test_and_reset<var>mode</var></samp>&rsquo;</dt>
<dd><p>These patterns emit code for an atomic bitwise operation on memory with memory
model semantics, and return the original value of the specified bit.
Operand 0 is an output operand which contains the value of the specified bit
from the memory location before the operation was performed. Operand 1 is the
memory on which the atomic operation is performed. Operand 2 is the bit within
the operand, starting with least significant bit. Operand 3 is the memory model
to be used by the operation. Operand 4 is a flag - it is <code>const1_rtx</code>
if operand 0 should contain the original value of the specified bit in the
least significant bit of the operand, and <code>const0_rtx</code> if the bit should
be in its original position in the operand.
<code>atomic_bit_test_and_set<var>mode</var></code> atomically sets the specified bit after
remembering its original value, <code>atomic_bit_test_and_complement<var>mode</var></code>
inverts the specified bit and <code>atomic_bit_test_and_reset<var>mode</var></code> clears
the specified bit.
</p>
<p>If these patterns are not defined, attempts will be made to use
<code>atomic_fetch_or<var>mode</var></code>, <code>atomic_fetch_xor<var>mode</var></code> or
<code>atomic_fetch_and<var>mode</var></code> instruction patterns, or their <code>sync</code>
counterparts. If none of these are available a compare-and-swap
loop will be used.
</p>
<a name="index-mem_005fthread_005ffence-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>mem_thread_fence</samp>&rsquo;</dt>
<dd><p>This pattern emits code required to implement a thread fence with
memory model semantics. Operand 0 is the memory model to be used.
</p>
<p>For the <code>__ATOMIC_RELAXED</code> model no instructions need to be issued
and this expansion is not invoked.
</p>
<p>The compiler always emits a compiler memory barrier regardless of what
expanding this pattern produced.
</p>
<p>If this pattern is not defined, the compiler falls back to expanding the
<code>memory_barrier</code> pattern, then to emitting <code>__sync_synchronize</code>
library call, and finally to just placing a compiler memory barrier.
</p>
<a name="index-get_005fthread_005fpointermode-instruction-pattern"></a>
<a name="index-set_005fthread_005fpointermode-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>get_thread_pointer<var>mode</var></samp>&rsquo;</dt>
<dt>&lsquo;<samp>set_thread_pointer<var>mode</var></samp>&rsquo;</dt>
<dd><p>These patterns emit code that reads/sets the TLS thread pointer. Currently,
these are only needed if the target needs to support the
<code>__builtin_thread_pointer</code> and <code>__builtin_set_thread_pointer</code>
builtins.
</p>
<p>The get/set patterns have a single output/input operand respectively,
with <var>mode</var> intended to be <code>Pmode</code>.
</p>
<a name="index-stack_005fprotect_005fset-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>stack_protect_set</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, moves a <code>ptr_mode</code> value from the memory
in operand 1 to the memory in operand 0 without leaving the value in
a register afterward. This is to avoid leaking the value some place
that an attacker might use to rewrite the stack guard slot after
having clobbered it.
</p>
<p>If this pattern is not defined, then a plain move pattern is generated.
</p>
<a name="index-stack_005fprotect_005ftest-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>stack_protect_test</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, compares a <code>ptr_mode</code> value from the
memory in operand 1 with the memory in operand 0 without leaving the
value in a register afterward and branches to operand 2 if the values
were equal.
</p>
<p>If this pattern is not defined, then a plain compare pattern and
conditional branch pattern is used.
</p>
<a name="index-clear_005fcache-instruction-pattern"></a>
</dd>
<dt>&lsquo;<samp>clear_cache</samp>&rsquo;</dt>
<dd><p>This pattern, if defined, flushes the instruction cache for a region of
memory. The region is bounded to by the Pmode pointers in operand 0
inclusive and operand 1 exclusive.
</p>
<p>If this pattern is not defined, a call to the library function
<code>__clear_cache</code> is used.
</p>
</dd>
</dl>
<hr>
<div class="header">
<p>
Next: <a href="Pattern-Ordering.html#Pattern-Ordering" accesskey="n" rel="next">Pattern Ordering</a>, Previous: <a href="Constraints.html#Constraints" accesskey="p" rel="prev">Constraints</a>, Up: <a href="Machine-Desc.html#Machine-Desc" accesskey="u" rel="up">Machine Desc</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>