toolchain/gcc-linaro-6.3.1-2017.02-x8.../share/doc/gcc/MIPS-Options.html

1215 lines
52 KiB
HTML

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
<html>
<!-- Copyright (C) 1988-2016 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 5.2, http://www.gnu.org/software/texinfo/ -->
<head>
<title>Using the GNU Compiler Collection (GCC): MIPS Options</title>
<meta name="description" content="Using the GNU Compiler Collection (GCC): MIPS Options">
<meta name="keywords" content="Using the GNU Compiler Collection (GCC): MIPS Options">
<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="Submodel-Options.html#Submodel-Options" rel="up" title="Submodel Options">
<link href="MMIX-Options.html#MMIX-Options" rel="next" title="MMIX Options">
<link href="MicroBlaze-Options.html#MicroBlaze-Options" rel="prev" title="MicroBlaze Options">
<style type="text/css">
<!--
a.summary-letter {text-decoration: none}
blockquote.smallquotation {font-size: smaller}
div.display {margin-left: 3.2em}
div.example {margin-left: 3.2em}
div.indentedblock {margin-left: 3.2em}
div.lisp {margin-left: 3.2em}
div.smalldisplay {margin-left: 3.2em}
div.smallexample {margin-left: 3.2em}
div.smallindentedblock {margin-left: 3.2em; font-size: smaller}
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.nocodebreak {white-space:nowrap}
span.nolinebreak {white-space:nowrap}
span.roman {font-family:serif; font-weight:normal}
span.sansserif {font-family:sans-serif; font-weight:normal}
ul.no-bullet {list-style: none}
-->
</style>
</head>
<body lang="en" bgcolor="#FFFFFF" text="#000000" link="#0000FF" vlink="#800080" alink="#FF0000">
<a name="MIPS-Options"></a>
<div class="header">
<p>
Next: <a href="MMIX-Options.html#MMIX-Options" accesskey="n" rel="next">MMIX Options</a>, Previous: <a href="MicroBlaze-Options.html#MicroBlaze-Options" accesskey="p" rel="prev">MicroBlaze Options</a>, Up: <a href="Submodel-Options.html#Submodel-Options" accesskey="u" rel="up">Submodel Options</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="MIPS-Options-1"></a>
<h4 class="subsection">3.18.26 MIPS Options</h4>
<a name="index-MIPS-options"></a>
<dl compact="compact">
<dt><code>-EB</code></dt>
<dd><a name="index-EB-1"></a>
<p>Generate big-endian code.
</p>
</dd>
<dt><code>-EL</code></dt>
<dd><a name="index-EL-1"></a>
<p>Generate little-endian code. This is the default for &lsquo;<samp>mips*el-*-*</samp>&rsquo;
configurations.
</p>
</dd>
<dt><code>-march=<var>arch</var></code></dt>
<dd><a name="index-march-7"></a>
<p>Generate code that runs on <var>arch</var>, which can be the name of a
generic MIPS ISA, or the name of a particular processor.
The ISA names are:
&lsquo;<samp>mips1</samp>&rsquo;, &lsquo;<samp>mips2</samp>&rsquo;, &lsquo;<samp>mips3</samp>&rsquo;, &lsquo;<samp>mips4</samp>&rsquo;,
&lsquo;<samp>mips32</samp>&rsquo;, &lsquo;<samp>mips32r2</samp>&rsquo;, &lsquo;<samp>mips32r3</samp>&rsquo;, &lsquo;<samp>mips32r5</samp>&rsquo;,
&lsquo;<samp>mips32r6</samp>&rsquo;, &lsquo;<samp>mips64</samp>&rsquo;, &lsquo;<samp>mips64r2</samp>&rsquo;, &lsquo;<samp>mips64r3</samp>&rsquo;,
&lsquo;<samp>mips64r5</samp>&rsquo; and &lsquo;<samp>mips64r6</samp>&rsquo;.
The processor names are:
&lsquo;<samp>4kc</samp>&rsquo;, &lsquo;<samp>4km</samp>&rsquo;, &lsquo;<samp>4kp</samp>&rsquo;, &lsquo;<samp>4ksc</samp>&rsquo;,
&lsquo;<samp>4kec</samp>&rsquo;, &lsquo;<samp>4kem</samp>&rsquo;, &lsquo;<samp>4kep</samp>&rsquo;, &lsquo;<samp>4ksd</samp>&rsquo;,
&lsquo;<samp>5kc</samp>&rsquo;, &lsquo;<samp>5kf</samp>&rsquo;,
&lsquo;<samp>20kc</samp>&rsquo;,
&lsquo;<samp>24kc</samp>&rsquo;, &lsquo;<samp>24kf2_1</samp>&rsquo;, &lsquo;<samp>24kf1_1</samp>&rsquo;,
&lsquo;<samp>24kec</samp>&rsquo;, &lsquo;<samp>24kef2_1</samp>&rsquo;, &lsquo;<samp>24kef1_1</samp>&rsquo;,
&lsquo;<samp>34kc</samp>&rsquo;, &lsquo;<samp>34kf2_1</samp>&rsquo;, &lsquo;<samp>34kf1_1</samp>&rsquo;, &lsquo;<samp>34kn</samp>&rsquo;,
&lsquo;<samp>74kc</samp>&rsquo;, &lsquo;<samp>74kf2_1</samp>&rsquo;, &lsquo;<samp>74kf1_1</samp>&rsquo;, &lsquo;<samp>74kf3_2</samp>&rsquo;,
&lsquo;<samp>1004kc</samp>&rsquo;, &lsquo;<samp>1004kf2_1</samp>&rsquo;, &lsquo;<samp>1004kf1_1</samp>&rsquo;,
&lsquo;<samp>i6400</samp>&rsquo;,
&lsquo;<samp>interaptiv</samp>&rsquo;,
&lsquo;<samp>loongson2e</samp>&rsquo;, &lsquo;<samp>loongson2f</samp>&rsquo;, &lsquo;<samp>loongson3a</samp>&rsquo;,
&lsquo;<samp>m4k</samp>&rsquo;,
&lsquo;<samp>m14k</samp>&rsquo;, &lsquo;<samp>m14kc</samp>&rsquo;, &lsquo;<samp>m14ke</samp>&rsquo;, &lsquo;<samp>m14kec</samp>&rsquo;,
&lsquo;<samp>m5100</samp>&rsquo;, &lsquo;<samp>m5101</samp>&rsquo;,
&lsquo;<samp>octeon</samp>&rsquo;, &lsquo;<samp>octeon+</samp>&rsquo;, &lsquo;<samp>octeon2</samp>&rsquo;, &lsquo;<samp>octeon3</samp>&rsquo;,
&lsquo;<samp>orion</samp>&rsquo;,
&lsquo;<samp>p5600</samp>&rsquo;,
&lsquo;<samp>r2000</samp>&rsquo;, &lsquo;<samp>r3000</samp>&rsquo;, &lsquo;<samp>r3900</samp>&rsquo;, &lsquo;<samp>r4000</samp>&rsquo;, &lsquo;<samp>r4400</samp>&rsquo;,
&lsquo;<samp>r4600</samp>&rsquo;, &lsquo;<samp>r4650</samp>&rsquo;, &lsquo;<samp>r4700</samp>&rsquo;, &lsquo;<samp>r6000</samp>&rsquo;, &lsquo;<samp>r8000</samp>&rsquo;,
&lsquo;<samp>rm7000</samp>&rsquo;, &lsquo;<samp>rm9000</samp>&rsquo;,
&lsquo;<samp>r10000</samp>&rsquo;, &lsquo;<samp>r12000</samp>&rsquo;, &lsquo;<samp>r14000</samp>&rsquo;, &lsquo;<samp>r16000</samp>&rsquo;,
&lsquo;<samp>sb1</samp>&rsquo;,
&lsquo;<samp>sr71000</samp>&rsquo;,
&lsquo;<samp>vr4100</samp>&rsquo;, &lsquo;<samp>vr4111</samp>&rsquo;, &lsquo;<samp>vr4120</samp>&rsquo;, &lsquo;<samp>vr4130</samp>&rsquo;, &lsquo;<samp>vr4300</samp>&rsquo;,
&lsquo;<samp>vr5000</samp>&rsquo;, &lsquo;<samp>vr5400</samp>&rsquo;, &lsquo;<samp>vr5500</samp>&rsquo;,
&lsquo;<samp>xlr</samp>&rsquo; and &lsquo;<samp>xlp</samp>&rsquo;.
The special value &lsquo;<samp>from-abi</samp>&rsquo; selects the
most compatible architecture for the selected ABI (that is,
&lsquo;<samp>mips1</samp>&rsquo; for 32-bit ABIs and &lsquo;<samp>mips3</samp>&rsquo; for 64-bit ABIs).
</p>
<p>The native Linux/GNU toolchain also supports the value &lsquo;<samp>native</samp>&rsquo;,
which selects the best architecture option for the host processor.
<samp>-march=native</samp> has no effect if GCC does not recognize
the processor.
</p>
<p>In processor names, a final &lsquo;<samp>000</samp>&rsquo; can be abbreviated as &lsquo;<samp>k</samp>&rsquo;
(for example, <samp>-march=r2k</samp>). Prefixes are optional, and
&lsquo;<samp>vr</samp>&rsquo; may be written &lsquo;<samp>r</samp>&rsquo;.
</p>
<p>Names of the form &lsquo;<samp><var>n</var>f2_1</samp>&rsquo; refer to processors with
FPUs clocked at half the rate of the core, names of the form
&lsquo;<samp><var>n</var>f1_1</samp>&rsquo; refer to processors with FPUs clocked at the same
rate as the core, and names of the form &lsquo;<samp><var>n</var>f3_2</samp>&rsquo; refer to
processors with FPUs clocked a ratio of 3:2 with respect to the core.
For compatibility reasons, &lsquo;<samp><var>n</var>f</samp>&rsquo; is accepted as a synonym
for &lsquo;<samp><var>n</var>f2_1</samp>&rsquo; while &lsquo;<samp><var>n</var>x</samp>&rsquo; and &lsquo;<samp><var>b</var>fx</samp>&rsquo; are
accepted as synonyms for &lsquo;<samp><var>n</var>f1_1</samp>&rsquo;.
</p>
<p>GCC defines two macros based on the value of this option. The first
is <code>_MIPS_ARCH</code>, which gives the name of target architecture, as
a string. The second has the form <code>_MIPS_ARCH_<var>foo</var></code>,
where <var>foo</var> is the capitalized value of <code>_MIPS_ARCH</code>.
For example, <samp>-march=r2000</samp> sets <code>_MIPS_ARCH</code>
to <code>&quot;r2000&quot;</code> and defines the macro <code>_MIPS_ARCH_R2000</code>.
</p>
<p>Note that the <code>_MIPS_ARCH</code> macro uses the processor names given
above. In other words, it has the full prefix and does not
abbreviate &lsquo;<samp>000</samp>&rsquo; as &lsquo;<samp>k</samp>&rsquo;. In the case of &lsquo;<samp>from-abi</samp>&rsquo;,
the macro names the resolved architecture (either <code>&quot;mips1&quot;</code> or
<code>&quot;mips3&quot;</code>). It names the default architecture when no
<samp>-march</samp> option is given.
</p>
</dd>
<dt><code>-mtune=<var>arch</var></code></dt>
<dd><a name="index-mtune-8"></a>
<p>Optimize for <var>arch</var>. Among other things, this option controls
the way instructions are scheduled, and the perceived cost of arithmetic
operations. The list of <var>arch</var> values is the same as for
<samp>-march</samp>.
</p>
<p>When this option is not used, GCC optimizes for the processor
specified by <samp>-march</samp>. By using <samp>-march</samp> and
<samp>-mtune</samp> together, it is possible to generate code that
runs on a family of processors, but optimize the code for one
particular member of that family.
</p>
<p><samp>-mtune</samp> defines the macros <code>_MIPS_TUNE</code> and
<code>_MIPS_TUNE_<var>foo</var></code>, which work in the same way as the
<samp>-march</samp> ones described above.
</p>
</dd>
<dt><code>-mips1</code></dt>
<dd><a name="index-mips1"></a>
<p>Equivalent to <samp>-march=mips1</samp>.
</p>
</dd>
<dt><code>-mips2</code></dt>
<dd><a name="index-mips2"></a>
<p>Equivalent to <samp>-march=mips2</samp>.
</p>
</dd>
<dt><code>-mips3</code></dt>
<dd><a name="index-mips3"></a>
<p>Equivalent to <samp>-march=mips3</samp>.
</p>
</dd>
<dt><code>-mips4</code></dt>
<dd><a name="index-mips4"></a>
<p>Equivalent to <samp>-march=mips4</samp>.
</p>
</dd>
<dt><code>-mips32</code></dt>
<dd><a name="index-mips32"></a>
<p>Equivalent to <samp>-march=mips32</samp>.
</p>
</dd>
<dt><code>-mips32r3</code></dt>
<dd><a name="index-mips32r3"></a>
<p>Equivalent to <samp>-march=mips32r3</samp>.
</p>
</dd>
<dt><code>-mips32r5</code></dt>
<dd><a name="index-mips32r5"></a>
<p>Equivalent to <samp>-march=mips32r5</samp>.
</p>
</dd>
<dt><code>-mips32r6</code></dt>
<dd><a name="index-mips32r6"></a>
<p>Equivalent to <samp>-march=mips32r6</samp>.
</p>
</dd>
<dt><code>-mips64</code></dt>
<dd><a name="index-mips64"></a>
<p>Equivalent to <samp>-march=mips64</samp>.
</p>
</dd>
<dt><code>-mips64r2</code></dt>
<dd><a name="index-mips64r2"></a>
<p>Equivalent to <samp>-march=mips64r2</samp>.
</p>
</dd>
<dt><code>-mips64r3</code></dt>
<dd><a name="index-mips64r3"></a>
<p>Equivalent to <samp>-march=mips64r3</samp>.
</p>
</dd>
<dt><code>-mips64r5</code></dt>
<dd><a name="index-mips64r5"></a>
<p>Equivalent to <samp>-march=mips64r5</samp>.
</p>
</dd>
<dt><code>-mips64r6</code></dt>
<dd><a name="index-mips64r6"></a>
<p>Equivalent to <samp>-march=mips64r6</samp>.
</p>
</dd>
<dt><code>-mips16</code></dt>
<dt><code>-mno-mips16</code></dt>
<dd><a name="index-mips16"></a>
<a name="index-mno_002dmips16"></a>
<p>Generate (do not generate) MIPS16 code. If GCC is targeting a
MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE.
</p>
<p>MIPS16 code generation can also be controlled on a per-function basis
by means of <code>mips16</code> and <code>nomips16</code> attributes.
See <a href="Function-Attributes.html#Function-Attributes">Function Attributes</a>, for more information.
</p>
</dd>
<dt><code>-mflip-mips16</code></dt>
<dd><a name="index-mflip_002dmips16"></a>
<p>Generate MIPS16 code on alternating functions. This option is provided
for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
not intended for ordinary use in compiling user code.
</p>
</dd>
<dt><code>-minterlink-compressed</code></dt>
<dt><code>-mno-interlink-compressed</code></dt>
<dd><a name="index-minterlink_002dcompressed"></a>
<a name="index-mno_002dinterlink_002dcompressed"></a>
<p>Require (do not require) that code using the standard (uncompressed) MIPS ISA
be link-compatible with MIPS16 and microMIPS code, and vice versa.
</p>
<p>For example, code using the standard ISA encoding cannot jump directly
to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
<samp>-minterlink-compressed</samp> therefore disables direct jumps unless GCC
knows that the target of the jump is not compressed.
</p>
</dd>
<dt><code>-minterlink-mips16</code></dt>
<dt><code>-mno-interlink-mips16</code></dt>
<dd><a name="index-minterlink_002dmips16"></a>
<a name="index-mno_002dinterlink_002dmips16"></a>
<p>Aliases of <samp>-minterlink-compressed</samp> and
<samp>-mno-interlink-compressed</samp>. These options predate the microMIPS ASE
and are retained for backwards compatibility.
</p>
</dd>
<dt><code>-mabi=32</code></dt>
<dt><code>-mabi=o64</code></dt>
<dt><code>-mabi=n32</code></dt>
<dt><code>-mabi=64</code></dt>
<dt><code>-mabi=eabi</code></dt>
<dd><a name="index-mabi_003d32"></a>
<a name="index-mabi_003do64"></a>
<a name="index-mabi_003dn32"></a>
<a name="index-mabi_003d64"></a>
<a name="index-mabi_003deabi"></a>
<p>Generate code for the given ABI.
</p>
<p>Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
generates 64-bit code when you select a 64-bit architecture, but you
can use <samp>-mgp32</samp> to get 32-bit code instead.
</p>
<p>For information about the O64 ABI, see
<a href="http://gcc.gnu.org/projects/mipso64-abi.html">http://gcc.gnu.org/projects/mipso64-abi.html</a>.
</p>
<p>GCC supports a variant of the o32 ABI in which floating-point registers
are 64 rather than 32 bits wide. You can select this combination with
<samp>-mabi=32</samp> <samp>-mfp64</samp>. This ABI relies on the <code>mthc1</code>
and <code>mfhc1</code> instructions and is therefore only supported for
MIPS32R2, MIPS32R3 and MIPS32R5 processors.
</p>
<p>The register assignments for arguments and return values remain the
same, but each scalar value is passed in a single 64-bit register
rather than a pair of 32-bit registers. For example, scalar
floating-point values are returned in &lsquo;<samp>$f0</samp>&rsquo; only, not a
&lsquo;<samp>$f0</samp>&rsquo;/&lsquo;<samp>$f1</samp>&rsquo; pair. The set of call-saved registers also
remains the same in that the even-numbered double-precision registers
are saved.
</p>
<p>Two additional variants of the o32 ABI are supported to enable
a transition from 32-bit to 64-bit registers. These are FPXX
(<samp>-mfpxx</samp>) and FP64A (<samp>-mfp64</samp> <samp>-mno-odd-spreg</samp>).
The FPXX extension mandates that all code must execute correctly
when run using 32-bit or 64-bit registers. The code can be interlinked
with either FP32 or FP64, but not both.
The FP64A extension is similar to the FP64 extension but forbids the
use of odd-numbered single-precision registers. This can be used
in conjunction with the <code>FRE</code> mode of FPUs in MIPS32R5
processors and allows both FP32 and FP64A code to interlink and
run in the same process without changing FPU modes.
</p>
</dd>
<dt><code>-mabicalls</code></dt>
<dt><code>-mno-abicalls</code></dt>
<dd><a name="index-mabicalls"></a>
<a name="index-mno_002dabicalls"></a>
<p>Generate (do not generate) code that is suitable for SVR4-style
dynamic objects. <samp>-mabicalls</samp> is the default for SVR4-based
systems.
</p>
</dd>
<dt><code>-mshared</code></dt>
<dt><code>-mno-shared</code></dt>
<dd><p>Generate (do not generate) code that is fully position-independent,
and that can therefore be linked into shared libraries. This option
only affects <samp>-mabicalls</samp>.
</p>
<p>All <samp>-mabicalls</samp> code has traditionally been position-independent,
regardless of options like <samp>-fPIC</samp> and <samp>-fpic</samp>. However,
as an extension, the GNU toolchain allows executables to use absolute
accesses for locally-binding symbols. It can also use shorter GP
initialization sequences and generate direct calls to locally-defined
functions. This mode is selected by <samp>-mno-shared</samp>.
</p>
<p><samp>-mno-shared</samp> depends on binutils 2.16 or higher and generates
objects that can only be linked by the GNU linker. However, the option
does not affect the ABI of the final executable; it only affects the ABI
of relocatable objects. Using <samp>-mno-shared</samp> generally makes
executables both smaller and quicker.
</p>
<p><samp>-mshared</samp> is the default.
</p>
</dd>
<dt><code>-mplt</code></dt>
<dt><code>-mno-plt</code></dt>
<dd><a name="index-mplt"></a>
<a name="index-mno_002dplt"></a>
<p>Assume (do not assume) that the static and dynamic linkers
support PLTs and copy relocations. This option only affects
<samp>-mno-shared -mabicalls</samp>. For the n64 ABI, this option
has no effect without <samp>-msym32</samp>.
</p>
<p>You can make <samp>-mplt</samp> the default by configuring
GCC with <samp>--with-mips-plt</samp>. The default is
<samp>-mno-plt</samp> otherwise.
</p>
</dd>
<dt><code>-mxgot</code></dt>
<dt><code>-mno-xgot</code></dt>
<dd><a name="index-mxgot-1"></a>
<a name="index-mno_002dxgot-1"></a>
<p>Lift (do not lift) the usual restrictions on the size of the global
offset table.
</p>
<p>GCC normally uses a single instruction to load values from the GOT.
While this is relatively efficient, it only works if the GOT
is smaller than about 64k. Anything larger causes the linker
to report an error such as:
</p>
<a name="index-relocation-truncated-to-fit-_0028MIPS_0029"></a>
<div class="smallexample">
<pre class="smallexample">relocation truncated to fit: R_MIPS_GOT16 foobar
</pre></div>
<p>If this happens, you should recompile your code with <samp>-mxgot</samp>.
This works with very large GOTs, although the code is also
less efficient, since it takes three instructions to fetch the
value of a global symbol.
</p>
<p>Note that some linkers can create multiple GOTs. If you have such a
linker, you should only need to use <samp>-mxgot</samp> when a single object
file accesses more than 64k&rsquo;s worth of GOT entries. Very few do.
</p>
<p>These options have no effect unless GCC is generating position
independent code.
</p>
</dd>
<dt><code>-mgp32</code></dt>
<dd><a name="index-mgp32"></a>
<p>Assume that general-purpose registers are 32 bits wide.
</p>
</dd>
<dt><code>-mgp64</code></dt>
<dd><a name="index-mgp64"></a>
<p>Assume that general-purpose registers are 64 bits wide.
</p>
</dd>
<dt><code>-mfp32</code></dt>
<dd><a name="index-mfp32"></a>
<p>Assume that floating-point registers are 32 bits wide.
</p>
</dd>
<dt><code>-mfp64</code></dt>
<dd><a name="index-mfp64"></a>
<p>Assume that floating-point registers are 64 bits wide.
</p>
</dd>
<dt><code>-mfpxx</code></dt>
<dd><a name="index-mfpxx"></a>
<p>Do not assume the width of floating-point registers.
</p>
</dd>
<dt><code>-mhard-float</code></dt>
<dd><a name="index-mhard_002dfloat-3"></a>
<p>Use floating-point coprocessor instructions.
</p>
</dd>
<dt><code>-msoft-float</code></dt>
<dd><a name="index-msoft_002dfloat-6"></a>
<p>Do not use floating-point coprocessor instructions. Implement
floating-point calculations using library calls instead.
</p>
</dd>
<dt><code>-mno-float</code></dt>
<dd><a name="index-mno_002dfloat"></a>
<p>Equivalent to <samp>-msoft-float</samp>, but additionally asserts that the
program being compiled does not perform any floating-point operations.
This option is presently supported only by some bare-metal MIPS
configurations, where it may select a special set of libraries
that lack all floating-point support (including, for example, the
floating-point <code>printf</code> formats).
If code compiled with <samp>-mno-float</samp> accidentally contains
floating-point operations, it is likely to suffer a link-time
or run-time failure.
</p>
</dd>
<dt><code>-msingle-float</code></dt>
<dd><a name="index-msingle_002dfloat"></a>
<p>Assume that the floating-point coprocessor only supports single-precision
operations.
</p>
</dd>
<dt><code>-mdouble-float</code></dt>
<dd><a name="index-mdouble_002dfloat"></a>
<p>Assume that the floating-point coprocessor supports double-precision
operations. This is the default.
</p>
</dd>
<dt><code>-modd-spreg</code></dt>
<dt><code>-mno-odd-spreg</code></dt>
<dd><a name="index-modd_002dspreg"></a>
<a name="index-mno_002dodd_002dspreg"></a>
<p>Enable the use of odd-numbered single-precision floating-point registers
for the o32 ABI. This is the default for processors that are known to
support these registers. When using the o32 FPXX ABI, <samp>-mno-odd-spreg</samp>
is set by default.
</p>
</dd>
<dt><code>-mabs=2008</code></dt>
<dt><code>-mabs=legacy</code></dt>
<dd><a name="index-mabs_003d2008"></a>
<a name="index-mabs_003dlegacy"></a>
<p>These options control the treatment of the special not-a-number (NaN)
IEEE 754 floating-point data with the <code>abs.<i>fmt</i></code> and
<code>neg.<i>fmt</i></code> machine instructions.
</p>
<p>By default or when <samp>-mabs=legacy</samp> is used the legacy
treatment is selected. In this case these instructions are considered
arithmetic and avoided where correct operation is required and the
input operand might be a NaN. A longer sequence of instructions that
manipulate the sign bit of floating-point datum manually is used
instead unless the <samp>-ffinite-math-only</samp> option has also been
specified.
</p>
<p>The <samp>-mabs=2008</samp> option selects the IEEE 754-2008 treatment. In
this case these instructions are considered non-arithmetic and therefore
operating correctly in all cases, including in particular where the
input operand is a NaN. These instructions are therefore always used
for the respective operations.
</p>
</dd>
<dt><code>-mnan=2008</code></dt>
<dt><code>-mnan=legacy</code></dt>
<dd><a name="index-mnan_003d2008"></a>
<a name="index-mnan_003dlegacy"></a>
<p>These options control the encoding of the special not-a-number (NaN)
IEEE 754 floating-point data.
</p>
<p>The <samp>-mnan=legacy</samp> option selects the legacy encoding. In this
case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
significand field being 0, whereas signaling NaNs (sNaNs) are denoted
by the first bit of their trailing significand field being 1.
</p>
<p>The <samp>-mnan=2008</samp> option selects the IEEE 754-2008 encoding. In
this case qNaNs are denoted by the first bit of their trailing
significand field being 1, whereas sNaNs are denoted by the first bit of
their trailing significand field being 0.
</p>
<p>The default is <samp>-mnan=legacy</samp> unless GCC has been configured with
<samp>--with-nan=2008</samp>.
</p>
</dd>
<dt><code>-mllsc</code></dt>
<dt><code>-mno-llsc</code></dt>
<dd><a name="index-mllsc"></a>
<a name="index-mno_002dllsc"></a>
<p>Use (do not use) &lsquo;<samp>ll</samp>&rsquo;, &lsquo;<samp>sc</samp>&rsquo;, and &lsquo;<samp>sync</samp>&rsquo; instructions to
implement atomic memory built-in functions. When neither option is
specified, GCC uses the instructions if the target architecture
supports them.
</p>
<p><samp>-mllsc</samp> is useful if the runtime environment can emulate the
instructions and <samp>-mno-llsc</samp> can be useful when compiling for
nonstandard ISAs. You can make either option the default by
configuring GCC with <samp>--with-llsc</samp> and <samp>--without-llsc</samp>
respectively. <samp>--with-llsc</samp> is the default for some
configurations; see the installation documentation for details.
</p>
</dd>
<dt><code>-mdsp</code></dt>
<dt><code>-mno-dsp</code></dt>
<dd><a name="index-mdsp"></a>
<a name="index-mno_002ddsp"></a>
<p>Use (do not use) revision 1 of the MIPS DSP ASE.
See <a href="MIPS-DSP-Built_002din-Functions.html#MIPS-DSP-Built_002din-Functions">MIPS DSP Built-in Functions</a>. This option defines the
preprocessor macro <code>__mips_dsp</code>. It also defines
<code>__mips_dsp_rev</code> to 1.
</p>
</dd>
<dt><code>-mdspr2</code></dt>
<dt><code>-mno-dspr2</code></dt>
<dd><a name="index-mdspr2"></a>
<a name="index-mno_002ddspr2"></a>
<p>Use (do not use) revision 2 of the MIPS DSP ASE.
See <a href="MIPS-DSP-Built_002din-Functions.html#MIPS-DSP-Built_002din-Functions">MIPS DSP Built-in Functions</a>. This option defines the
preprocessor macros <code>__mips_dsp</code> and <code>__mips_dspr2</code>.
It also defines <code>__mips_dsp_rev</code> to 2.
</p>
</dd>
<dt><code>-msmartmips</code></dt>
<dt><code>-mno-smartmips</code></dt>
<dd><a name="index-msmartmips"></a>
<a name="index-mno_002dsmartmips"></a>
<p>Use (do not use) the MIPS SmartMIPS ASE.
</p>
</dd>
<dt><code>-mpaired-single</code></dt>
<dt><code>-mno-paired-single</code></dt>
<dd><a name="index-mpaired_002dsingle"></a>
<a name="index-mno_002dpaired_002dsingle"></a>
<p>Use (do not use) paired-single floating-point instructions.
See <a href="MIPS-Paired_002dSingle-Support.html#MIPS-Paired_002dSingle-Support">MIPS Paired-Single Support</a>. This option requires
hardware floating-point support to be enabled.
</p>
</dd>
<dt><code>-mdmx</code></dt>
<dt><code>-mno-mdmx</code></dt>
<dd><a name="index-mdmx"></a>
<a name="index-mno_002dmdmx"></a>
<p>Use (do not use) MIPS Digital Media Extension instructions.
This option can only be used when generating 64-bit code and requires
hardware floating-point support to be enabled.
</p>
</dd>
<dt><code>-mips3d</code></dt>
<dt><code>-mno-mips3d</code></dt>
<dd><a name="index-mips3d"></a>
<a name="index-mno_002dmips3d"></a>
<p>Use (do not use) the MIPS-3D ASE. See <a href="MIPS_002d3D-Built_002din-Functions.html#MIPS_002d3D-Built_002din-Functions">MIPS-3D Built-in Functions</a>.
The option <samp>-mips3d</samp> implies <samp>-mpaired-single</samp>.
</p>
</dd>
<dt><code>-mmicromips</code></dt>
<dt><code>-mno-micromips</code></dt>
<dd><a name="index-mmicromips"></a>
<a name="index-mno_002dmmicromips"></a>
<p>Generate (do not generate) microMIPS code.
</p>
<p>MicroMIPS code generation can also be controlled on a per-function basis
by means of <code>micromips</code> and <code>nomicromips</code> attributes.
See <a href="Function-Attributes.html#Function-Attributes">Function Attributes</a>, for more information.
</p>
</dd>
<dt><code>-mmt</code></dt>
<dt><code>-mno-mt</code></dt>
<dd><a name="index-mmt"></a>
<a name="index-mno_002dmt"></a>
<p>Use (do not use) MT Multithreading instructions.
</p>
</dd>
<dt><code>-mmcu</code></dt>
<dt><code>-mno-mcu</code></dt>
<dd><a name="index-mmcu-1"></a>
<a name="index-mno_002dmcu"></a>
<p>Use (do not use) the MIPS MCU ASE instructions.
</p>
</dd>
<dt><code>-meva</code></dt>
<dt><code>-mno-eva</code></dt>
<dd><a name="index-meva"></a>
<a name="index-mno_002deva"></a>
<p>Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
</p>
</dd>
<dt><code>-mvirt</code></dt>
<dt><code>-mno-virt</code></dt>
<dd><a name="index-mvirt"></a>
<a name="index-mno_002dvirt"></a>
<p>Use (do not use) the MIPS Virtualization Application Specific instructions.
</p>
</dd>
<dt><code>-mxpa</code></dt>
<dt><code>-mno-xpa</code></dt>
<dd><a name="index-mxpa"></a>
<a name="index-mno_002dxpa"></a>
<p>Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
</p>
</dd>
<dt><code>-mlong64</code></dt>
<dd><a name="index-mlong64"></a>
<p>Force <code>long</code> types to be 64 bits wide. See <samp>-mlong32</samp> for
an explanation of the default and the way that the pointer size is
determined.
</p>
</dd>
<dt><code>-mlong32</code></dt>
<dd><a name="index-mlong32"></a>
<p>Force <code>long</code>, <code>int</code>, and pointer types to be 32 bits wide.
</p>
<p>The default size of <code>int</code>s, <code>long</code>s and pointers depends on
the ABI. All the supported ABIs use 32-bit <code>int</code>s. The n64 ABI
uses 64-bit <code>long</code>s, as does the 64-bit EABI; the others use
32-bit <code>long</code>s. Pointers are the same size as <code>long</code>s,
or the same size as integer registers, whichever is smaller.
</p>
</dd>
<dt><code>-msym32</code></dt>
<dt><code>-mno-sym32</code></dt>
<dd><a name="index-msym32"></a>
<a name="index-mno_002dsym32"></a>
<p>Assume (do not assume) that all symbols have 32-bit values, regardless
of the selected ABI. This option is useful in combination with
<samp>-mabi=64</samp> and <samp>-mno-abicalls</samp> because it allows GCC
to generate shorter and faster references to symbolic addresses.
</p>
</dd>
<dt><code>-G <var>num</var></code></dt>
<dd><a name="index-G-1"></a>
<p>Put definitions of externally-visible data in a small data section
if that data is no bigger than <var>num</var> bytes. GCC can then generate
more efficient accesses to the data; see <samp>-mgpopt</samp> for details.
</p>
<p>The default <samp>-G</samp> option depends on the configuration.
</p>
</dd>
<dt><code>-mlocal-sdata</code></dt>
<dt><code>-mno-local-sdata</code></dt>
<dd><a name="index-mlocal_002dsdata"></a>
<a name="index-mno_002dlocal_002dsdata"></a>
<p>Extend (do not extend) the <samp>-G</samp> behavior to local data too,
such as to static variables in C. <samp>-mlocal-sdata</samp> is the
default for all configurations.
</p>
<p>If the linker complains that an application is using too much small data,
you might want to try rebuilding the less performance-critical parts with
<samp>-mno-local-sdata</samp>. You might also want to build large
libraries with <samp>-mno-local-sdata</samp>, so that the libraries leave
more room for the main program.
</p>
</dd>
<dt><code>-mextern-sdata</code></dt>
<dt><code>-mno-extern-sdata</code></dt>
<dd><a name="index-mextern_002dsdata"></a>
<a name="index-mno_002dextern_002dsdata"></a>
<p>Assume (do not assume) that externally-defined data is in
a small data section if the size of that data is within the <samp>-G</samp> limit.
<samp>-mextern-sdata</samp> is the default for all configurations.
</p>
<p>If you compile a module <var>Mod</var> with <samp>-mextern-sdata</samp> <samp>-G
<var>num</var></samp> <samp>-mgpopt</samp>, and <var>Mod</var> references a variable <var>Var</var>
that is no bigger than <var>num</var> bytes, you must make sure that <var>Var</var>
is placed in a small data section. If <var>Var</var> is defined by another
module, you must either compile that module with a high-enough
<samp>-G</samp> setting or attach a <code>section</code> attribute to <var>Var</var>&rsquo;s
definition. If <var>Var</var> is common, you must link the application
with a high-enough <samp>-G</samp> setting.
</p>
<p>The easiest way of satisfying these restrictions is to compile
and link every module with the same <samp>-G</samp> option. However,
you may wish to build a library that supports several different
small data limits. You can do this by compiling the library with
the highest supported <samp>-G</samp> setting and additionally using
<samp>-mno-extern-sdata</samp> to stop the library from making assumptions
about externally-defined data.
</p>
</dd>
<dt><code>-mgpopt</code></dt>
<dt><code>-mno-gpopt</code></dt>
<dd><a name="index-mgpopt"></a>
<a name="index-mno_002dgpopt"></a>
<p>Use (do not use) GP-relative accesses for symbols that are known to be
in a small data section; see <samp>-G</samp>, <samp>-mlocal-sdata</samp> and
<samp>-mextern-sdata</samp>. <samp>-mgpopt</samp> is the default for all
configurations.
</p>
<p><samp>-mno-gpopt</samp> is useful for cases where the <code>$gp</code> register
might not hold the value of <code>_gp</code>. For example, if the code is
part of a library that might be used in a boot monitor, programs that
call boot monitor routines pass an unknown value in <code>$gp</code>.
(In such situations, the boot monitor itself is usually compiled
with <samp>-G0</samp>.)
</p>
<p><samp>-mno-gpopt</samp> implies <samp>-mno-local-sdata</samp> and
<samp>-mno-extern-sdata</samp>.
</p>
</dd>
<dt><code>-membedded-data</code></dt>
<dt><code>-mno-embedded-data</code></dt>
<dd><a name="index-membedded_002ddata"></a>
<a name="index-mno_002dembedded_002ddata"></a>
<p>Allocate variables to the read-only data section first if possible, then
next in the small data section if possible, otherwise in data. This gives
slightly slower code than the default, but reduces the amount of RAM required
when executing, and thus may be preferred for some embedded systems.
</p>
</dd>
<dt><code>-muninit-const-in-rodata</code></dt>
<dt><code>-mno-uninit-const-in-rodata</code></dt>
<dd><a name="index-muninit_002dconst_002din_002drodata"></a>
<a name="index-mno_002duninit_002dconst_002din_002drodata"></a>
<p>Put uninitialized <code>const</code> variables in the read-only data section.
This option is only meaningful in conjunction with <samp>-membedded-data</samp>.
</p>
</dd>
<dt><code>-mcode-readable=<var>setting</var></code></dt>
<dd><a name="index-mcode_002dreadable"></a>
<p>Specify whether GCC may generate code that reads from executable sections.
There are three possible settings:
</p>
<dl compact="compact">
<dt><code>-mcode-readable=yes</code></dt>
<dd><p>Instructions may freely access executable sections. This is the
default setting.
</p>
</dd>
<dt><code>-mcode-readable=pcrel</code></dt>
<dd><p>MIPS16 PC-relative load instructions can access executable sections,
but other instructions must not do so. This option is useful on 4KSc
and 4KSd processors when the code TLBs have the Read Inhibit bit set.
It is also useful on processors that can be configured to have a dual
instruction/data SRAM interface and that, like the M4K, automatically
redirect PC-relative loads to the instruction RAM.
</p>
</dd>
<dt><code>-mcode-readable=no</code></dt>
<dd><p>Instructions must not access executable sections. This option can be
useful on targets that are configured to have a dual instruction/data
SRAM interface but that (unlike the M4K) do not automatically redirect
PC-relative loads to the instruction RAM.
</p></dd>
</dl>
</dd>
<dt><code>-msplit-addresses</code></dt>
<dt><code>-mno-split-addresses</code></dt>
<dd><a name="index-msplit_002daddresses"></a>
<a name="index-mno_002dsplit_002daddresses"></a>
<p>Enable (disable) use of the <code>%hi()</code> and <code>%lo()</code> assembler
relocation operators. This option has been superseded by
<samp>-mexplicit-relocs</samp> but is retained for backwards compatibility.
</p>
</dd>
<dt><code>-mexplicit-relocs</code></dt>
<dt><code>-mno-explicit-relocs</code></dt>
<dd><a name="index-mexplicit_002drelocs-1"></a>
<a name="index-mno_002dexplicit_002drelocs-1"></a>
<p>Use (do not use) assembler relocation operators when dealing with symbolic
addresses. The alternative, selected by <samp>-mno-explicit-relocs</samp>,
is to use assembler macros instead.
</p>
<p><samp>-mexplicit-relocs</samp> is the default if GCC was configured
to use an assembler that supports relocation operators.
</p>
</dd>
<dt><code>-mcheck-zero-division</code></dt>
<dt><code>-mno-check-zero-division</code></dt>
<dd><a name="index-mcheck_002dzero_002ddivision"></a>
<a name="index-mno_002dcheck_002dzero_002ddivision"></a>
<p>Trap (do not trap) on integer division by zero.
</p>
<p>The default is <samp>-mcheck-zero-division</samp>.
</p>
</dd>
<dt><code>-mdivide-traps</code></dt>
<dt><code>-mdivide-breaks</code></dt>
<dd><a name="index-mdivide_002dtraps"></a>
<a name="index-mdivide_002dbreaks"></a>
<p>MIPS systems check for division by zero by generating either a
conditional trap or a break instruction. Using traps results in
smaller code, but is only supported on MIPS II and later. Also, some
versions of the Linux kernel have a bug that prevents trap from
generating the proper signal (<code>SIGFPE</code>). Use <samp>-mdivide-traps</samp> to
allow conditional traps on architectures that support them and
<samp>-mdivide-breaks</samp> to force the use of breaks.
</p>
<p>The default is usually <samp>-mdivide-traps</samp>, but this can be
overridden at configure time using <samp>--with-divide=breaks</samp>.
Divide-by-zero checks can be completely disabled using
<samp>-mno-check-zero-division</samp>.
</p>
</dd>
<dt><code>-mmemcpy</code></dt>
<dt><code>-mno-memcpy</code></dt>
<dd><a name="index-mmemcpy-1"></a>
<a name="index-mno_002dmemcpy"></a>
<p>Force (do not force) the use of <code>memcpy</code> for non-trivial block
moves. The default is <samp>-mno-memcpy</samp>, which allows GCC to inline
most constant-sized copies.
</p>
</dd>
<dt><code>-mlong-calls</code></dt>
<dt><code>-mno-long-calls</code></dt>
<dd><a name="index-mlong_002dcalls-5"></a>
<a name="index-mno_002dlong_002dcalls-3"></a>
<p>Disable (do not disable) use of the <code>jal</code> instruction. Calling
functions using <code>jal</code> is more efficient but requires the caller
and callee to be in the same 256 megabyte segment.
</p>
<p>This option has no effect on abicalls code. The default is
<samp>-mno-long-calls</samp>.
</p>
</dd>
<dt><code>-mmad</code></dt>
<dt><code>-mno-mad</code></dt>
<dd><a name="index-mmad"></a>
<a name="index-mno_002dmad"></a>
<p>Enable (disable) use of the <code>mad</code>, <code>madu</code> and <code>mul</code>
instructions, as provided by the R4650 ISA.
</p>
</dd>
<dt><code>-mimadd</code></dt>
<dt><code>-mno-imadd</code></dt>
<dd><a name="index-mimadd"></a>
<a name="index-mno_002dimadd"></a>
<p>Enable (disable) use of the <code>madd</code> and <code>msub</code> integer
instructions. The default is <samp>-mimadd</samp> on architectures
that support <code>madd</code> and <code>msub</code> except for the 74k
architecture where it was found to generate slower code.
</p>
</dd>
<dt><code>-mfused-madd</code></dt>
<dt><code>-mno-fused-madd</code></dt>
<dd><a name="index-mfused_002dmadd-1"></a>
<a name="index-mno_002dfused_002dmadd-1"></a>
<p>Enable (disable) use of the floating-point multiply-accumulate
instructions, when they are available. The default is
<samp>-mfused-madd</samp>.
</p>
<p>On the R8000 CPU when multiply-accumulate instructions are used,
the intermediate product is calculated to infinite precision
and is not subject to the FCSR Flush to Zero bit. This may be
undesirable in some circumstances. On other processors the result
is numerically identical to the equivalent computation using
separate multiply, add, subtract and negate instructions.
</p>
</dd>
<dt><code>-nocpp</code></dt>
<dd><a name="index-nocpp"></a>
<p>Tell the MIPS assembler to not run its preprocessor over user
assembler files (with a &lsquo;<samp>.s</samp>&rsquo; suffix) when assembling them.
</p>
</dd>
<dt><code>-mfix-24k</code></dt>
<dt><code>-mno-fix-24k</code></dt>
<dd><a name="index-mfix_002d24k"></a>
<a name="index-mno_002dfix_002d24k"></a>
<p>Work around the 24K E48 (lost data on stores during refill) errata.
The workarounds are implemented by the assembler rather than by GCC.
</p>
</dd>
<dt><code>-mfix-r4000</code></dt>
<dt><code>-mno-fix-r4000</code></dt>
<dd><a name="index-mfix_002dr4000"></a>
<a name="index-mno_002dfix_002dr4000"></a>
<p>Work around certain R4000 CPU errata:
</p><ul class="no-bullet">
<li>- A double-word or a variable shift may give an incorrect result if executed
immediately after starting an integer division.
</li><li>- A double-word or a variable shift may give an incorrect result if executed
while an integer multiplication is in progress.
</li><li>- An integer division may give an incorrect result if started in a delay slot
of a taken branch or a jump.
</li></ul>
</dd>
<dt><code>-mfix-r4400</code></dt>
<dt><code>-mno-fix-r4400</code></dt>
<dd><a name="index-mfix_002dr4400"></a>
<a name="index-mno_002dfix_002dr4400"></a>
<p>Work around certain R4400 CPU errata:
</p><ul class="no-bullet">
<li>- A double-word or a variable shift may give an incorrect result if executed
immediately after starting an integer division.
</li></ul>
</dd>
<dt><code>-mfix-r10000</code></dt>
<dt><code>-mno-fix-r10000</code></dt>
<dd><a name="index-mfix_002dr10000"></a>
<a name="index-mno_002dfix_002dr10000"></a>
<p>Work around certain R10000 errata:
</p><ul class="no-bullet">
<li>- <code>ll</code>/<code>sc</code> sequences may not behave atomically on revisions
prior to 3.0. They may deadlock on revisions 2.6 and earlier.
</li></ul>
<p>This option can only be used if the target architecture supports
branch-likely instructions. <samp>-mfix-r10000</samp> is the default when
<samp>-march=r10000</samp> is used; <samp>-mno-fix-r10000</samp> is the default
otherwise.
</p>
</dd>
<dt><code>-mfix-rm7000</code></dt>
<dt><code>-mno-fix-rm7000</code></dt>
<dd><a name="index-mfix_002drm7000"></a>
<p>Work around the RM7000 <code>dmult</code>/<code>dmultu</code> errata. The
workarounds are implemented by the assembler rather than by GCC.
</p>
</dd>
<dt><code>-mfix-vr4120</code></dt>
<dt><code>-mno-fix-vr4120</code></dt>
<dd><a name="index-mfix_002dvr4120"></a>
<p>Work around certain VR4120 errata:
</p><ul class="no-bullet">
<li>- <code>dmultu</code> does not always produce the correct result.
</li><li>- <code>div</code> and <code>ddiv</code> do not always produce the correct result if one
of the operands is negative.
</li></ul>
<p>The workarounds for the division errata rely on special functions in
<samp>libgcc.a</samp>. At present, these functions are only provided by
the <code>mips64vr*-elf</code> configurations.
</p>
<p>Other VR4120 errata require a NOP to be inserted between certain pairs of
instructions. These errata are handled by the assembler, not by GCC itself.
</p>
</dd>
<dt><code>-mfix-vr4130</code></dt>
<dd><a name="index-mfix_002dvr4130"></a>
<p>Work around the VR4130 <code>mflo</code>/<code>mfhi</code> errata. The
workarounds are implemented by the assembler rather than by GCC,
although GCC avoids using <code>mflo</code> and <code>mfhi</code> if the
VR4130 <code>macc</code>, <code>macchi</code>, <code>dmacc</code> and <code>dmacchi</code>
instructions are available instead.
</p>
</dd>
<dt><code>-mfix-sb1</code></dt>
<dt><code>-mno-fix-sb1</code></dt>
<dd><a name="index-mfix_002dsb1"></a>
<p>Work around certain SB-1 CPU core errata.
(This flag currently works around the SB-1 revision 2
&ldquo;F1&rdquo; and &ldquo;F2&rdquo; floating-point errata.)
</p>
</dd>
<dt><code>-mr10k-cache-barrier=<var>setting</var></code></dt>
<dd><a name="index-mr10k_002dcache_002dbarrier"></a>
<p>Specify whether GCC should insert cache barriers to avoid the
side-effects of speculation on R10K processors.
</p>
<p>In common with many processors, the R10K tries to predict the outcome
of a conditional branch and speculatively executes instructions from
the &ldquo;taken&rdquo; branch. It later aborts these instructions if the
predicted outcome is wrong. However, on the R10K, even aborted
instructions can have side effects.
</p>
<p>This problem only affects kernel stores and, depending on the system,
kernel loads. As an example, a speculatively-executed store may load
the target memory into cache and mark the cache line as dirty, even if
the store itself is later aborted. If a DMA operation writes to the
same area of memory before the &ldquo;dirty&rdquo; line is flushed, the cached
data overwrites the DMA-ed data. See the R10K processor manual
for a full description, including other potential problems.
</p>
<p>One workaround is to insert cache barrier instructions before every memory
access that might be speculatively executed and that might have side
effects even if aborted. <samp>-mr10k-cache-barrier=<var>setting</var></samp>
controls GCC&rsquo;s implementation of this workaround. It assumes that
aborted accesses to any byte in the following regions does not have
side effects:
</p>
<ol>
<li> the memory occupied by the current function&rsquo;s stack frame;
</li><li> the memory occupied by an incoming stack argument;
</li><li> the memory occupied by an object with a link-time-constant address.
</li></ol>
<p>It is the kernel&rsquo;s responsibility to ensure that speculative
accesses to these regions are indeed safe.
</p>
<p>If the input program contains a function declaration such as:
</p>
<div class="smallexample">
<pre class="smallexample">void foo (void);
</pre></div>
<p>then the implementation of <code>foo</code> must allow <code>j foo</code> and
<code>jal foo</code> to be executed speculatively. GCC honors this
restriction for functions it compiles itself. It expects non-GCC
functions (such as hand-written assembly code) to do the same.
</p>
<p>The option has three forms:
</p>
<dl compact="compact">
<dt><code>-mr10k-cache-barrier=load-store</code></dt>
<dd><p>Insert a cache barrier before a load or store that might be
speculatively executed and that might have side effects even
if aborted.
</p>
</dd>
<dt><code>-mr10k-cache-barrier=store</code></dt>
<dd><p>Insert a cache barrier before a store that might be speculatively
executed and that might have side effects even if aborted.
</p>
</dd>
<dt><code>-mr10k-cache-barrier=none</code></dt>
<dd><p>Disable the insertion of cache barriers. This is the default setting.
</p></dd>
</dl>
</dd>
<dt><code>-mflush-func=<var>func</var></code></dt>
<dt><code>-mno-flush-func</code></dt>
<dd><a name="index-mflush_002dfunc"></a>
<p>Specifies the function to call to flush the I and D caches, or to not
call any such function. If called, the function must take the same
arguments as the common <code>_flush_func</code>, that is, the address of the
memory range for which the cache is being flushed, the size of the
memory range, and the number 3 (to flush both caches). The default
depends on the target GCC was configured for, but commonly is either
<code>_flush_func</code> or <code>__cpu_flush</code>.
</p>
</dd>
<dt><code>mbranch-cost=<var>num</var></code></dt>
<dd><a name="index-mbranch_002dcost-2"></a>
<p>Set the cost of branches to roughly <var>num</var> &ldquo;simple&rdquo; instructions.
This cost is only a heuristic and is not guaranteed to produce
consistent results across releases. A zero cost redundantly selects
the default, which is based on the <samp>-mtune</samp> setting.
</p>
</dd>
<dt><code>-mbranch-likely</code></dt>
<dt><code>-mno-branch-likely</code></dt>
<dd><a name="index-mbranch_002dlikely"></a>
<a name="index-mno_002dbranch_002dlikely"></a>
<p>Enable or disable use of Branch Likely instructions, regardless of the
default for the selected architecture. By default, Branch Likely
instructions may be generated if they are supported by the selected
architecture. An exception is for the MIPS32 and MIPS64 architectures
and processors that implement those architectures; for those, Branch
Likely instructions are not be generated by default because the MIPS32
and MIPS64 architectures specifically deprecate their use.
</p>
</dd>
<dt><code>-mcompact-branches=never</code></dt>
<dt><code>-mcompact-branches=optimal</code></dt>
<dt><code>-mcompact-branches=always</code></dt>
<dd><a name="index-mcompact_002dbranches_003dnever"></a>
<a name="index-mcompact_002dbranches_003doptimal"></a>
<a name="index-mcompact_002dbranches_003dalways"></a>
<p>These options control which form of branches will be generated. The
default is <samp>-mcompact-branches=optimal</samp>.
</p>
<p>The <samp>-mcompact-branches=never</samp> option ensures that compact branch
instructions will never be generated.
</p>
<p>The <samp>-mcompact-branches=always</samp> option ensures that a compact
branch instruction will be generated if available. If a compact branch
instruction is not available, a delay slot form of the branch will be
used instead.
</p>
<p>This option is supported from MIPS Release 6 onwards.
</p>
<p>The <samp>-mcompact-branches=optimal</samp> option will cause a delay slot
branch to be used if one is available in the current ISA and the delay
slot is successfully filled. If the delay slot is not filled, a compact
branch will be chosen if one is available.
</p>
</dd>
<dt><code>-mfp-exceptions</code></dt>
<dt><code>-mno-fp-exceptions</code></dt>
<dd><a name="index-mfp_002dexceptions"></a>
<p>Specifies whether FP exceptions are enabled. This affects how
FP instructions are scheduled for some processors.
The default is that FP exceptions are
enabled.
</p>
<p>For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
64-bit code, then we can use both FP pipes. Otherwise, we can only use one
FP pipe.
</p>
</dd>
<dt><code>-mvr4130-align</code></dt>
<dt><code>-mno-vr4130-align</code></dt>
<dd><a name="index-mvr4130_002dalign"></a>
<p>The VR4130 pipeline is two-way superscalar, but can only issue two
instructions together if the first one is 8-byte aligned. When this
option is enabled, GCC aligns pairs of instructions that it
thinks should execute in parallel.
</p>
<p>This option only has an effect when optimizing for the VR4130.
It normally makes code faster, but at the expense of making it bigger.
It is enabled by default at optimization level <samp>-O3</samp>.
</p>
</dd>
<dt><code>-msynci</code></dt>
<dt><code>-mno-synci</code></dt>
<dd><a name="index-msynci"></a>
<p>Enable (disable) generation of <code>synci</code> instructions on
architectures that support it. The <code>synci</code> instructions (if
enabled) are generated when <code>__builtin___clear_cache</code> is
compiled.
</p>
<p>This option defaults to <samp>-mno-synci</samp>, but the default can be
overridden by configuring GCC with <samp>--with-synci</samp>.
</p>
<p>When compiling code for single processor systems, it is generally safe
to use <code>synci</code>. However, on many multi-core (SMP) systems, it
does not invalidate the instruction caches on all cores and may lead
to undefined behavior.
</p>
</dd>
<dt><code>-mrelax-pic-calls</code></dt>
<dt><code>-mno-relax-pic-calls</code></dt>
<dd><a name="index-mrelax_002dpic_002dcalls"></a>
<p>Try to turn PIC calls that are normally dispatched via register
<code>$25</code> into direct calls. This is only possible if the linker can
resolve the destination at link time and if the destination is within
range for a direct call.
</p>
<p><samp>-mrelax-pic-calls</samp> is the default if GCC was configured to use
an assembler and a linker that support the <code>.reloc</code> assembly
directive and <samp>-mexplicit-relocs</samp> is in effect. With
<samp>-mno-explicit-relocs</samp>, this optimization can be performed by the
assembler and the linker alone without help from the compiler.
</p>
</dd>
<dt><code>-mmcount-ra-address</code></dt>
<dt><code>-mno-mcount-ra-address</code></dt>
<dd><a name="index-mmcount_002dra_002daddress"></a>
<a name="index-mno_002dmcount_002dra_002daddress"></a>
<p>Emit (do not emit) code that allows <code>_mcount</code> to modify the
calling function&rsquo;s return address. When enabled, this option extends
the usual <code>_mcount</code> interface with a new <var>ra-address</var>
parameter, which has type <code>intptr_t *</code> and is passed in register
<code>$12</code>. <code>_mcount</code> can then modify the return address by
doing both of the following:
</p><ul>
<li> Returning the new address in register <code>$31</code>.
</li><li> Storing the new address in <code>*<var>ra-address</var></code>,
if <var>ra-address</var> is nonnull.
</li></ul>
<p>The default is <samp>-mno-mcount-ra-address</samp>.
</p>
</dd>
<dt><code>-mframe-header-opt</code></dt>
<dt><code>-mno-frame-header-opt</code></dt>
<dd><a name="index-mframe_002dheader_002dopt"></a>
<p>Enable (disable) frame header optimization in the o32 ABI. When using the
o32 ABI, calling functions will allocate 16 bytes on the stack for the called
function to write out register arguments. When enabled, this optimization
will suppress the allocation of the frame header if it can be determined that
it is unused.
</p>
<p>This optimization is off by default at all optimization levels.
</p>
</dd>
</dl>
<hr>
<div class="header">
<p>
Next: <a href="MMIX-Options.html#MMIX-Options" accesskey="n" rel="next">MMIX Options</a>, Previous: <a href="MicroBlaze-Options.html#MicroBlaze-Options" accesskey="p" rel="prev">MicroBlaze Options</a>, Up: <a href="Submodel-Options.html#Submodel-Options" accesskey="u" rel="up">Submodel Options</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>