ePenguin
/
ePenguin-Software-Framework
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
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.
stable
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'a2aee475ea'
${ noResults }
ePenguin-Software-Framework/toolchains/avr/share/doc/gccint/Regs-and-Memory.html

438 lines
24 KiB
HTML
Raw Normal View History Unescape Escape

the monkey dances and laughs and sings and dances
4 years ago
<html lang="en">
<head>
<title>Regs and Memory - GNU Compiler Collection (GCC) Internals</title>
<meta http-equiv="Content-Type" content="text/html">
<meta name="description" content="GNU Compiler Collection (GCC) Internals">
<meta name="generator" content="makeinfo 4.13">
<link title="Top" rel="start" href="index.html#Top">
<link rel="up" href="RTL.html#RTL" title="RTL">
<link rel="prev" href="Constants.html#Constants" title="Constants">
<link rel="next" href="Arithmetic.html#Arithmetic" title="Arithmetic">
<link href="http://www.gnu.org/software/texinfo/" rel="generator-home" title="Texinfo Homepage">
<!--
Copyright (C) 1988-2015 Free Software Foundation, Inc.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with the
Invariant Sections being ``Funding Free Software'', the Front-Cover
Texts being (a) (see below), and with the Back-Cover Texts being (b)
(see below). A copy of the license is included in the section entitled
``GNU Free Documentation License''.
(a) The FSF's Front-Cover Text is:
A GNU Manual
(b) The FSF's Back-Cover Text is:
You have freedom to copy and modify this GNU Manual, like GNU
software. Copies published by the Free Software Foundation raise
funds for GNU development.-->
<meta http-equiv="Content-Style-Type" content="text/css">
<style type="text/css"><!--
pre.display { font-family:inherit }
pre.format { font-family:inherit }
pre.smalldisplay { font-family:inherit; font-size:smaller }
pre.smallformat { font-family:inherit; font-size:smaller }
pre.smallexample { font-size:smaller }
pre.smalllisp { font-size:smaller }
span.sc { font-variant:small-caps }
span.roman { font-family:serif; font-weight:normal; }
span.sansserif { font-family:sans-serif; font-weight:normal; }
--></style>
</head>
<body>
<div class="node">
<a name="Regs-and-Memory"></a>
<p>
Next:&nbsp;<a rel="next" accesskey="n" href="Arithmetic.html#Arithmetic">Arithmetic</a>,
Previous:&nbsp;<a rel="previous" accesskey="p" href="Constants.html#Constants">Constants</a>,
Up:&nbsp;<a rel="up" accesskey="u" href="RTL.html#RTL">RTL</a>
<hr>
</div>
<h3 class="section">13.8 Registers and Memory</h3>
<p><a name="index-RTL-register-expressions-2844"></a><a name="index-RTL-memory-expressions-2845"></a>
Here are the RTL expression types for describing access to machine
registers and to main memory.
<a name="index-reg-2846"></a>
<a name="index-hard-registers-2847"></a>
<a name="index-pseudo-registers-2848"></a>
<dl><dt><code>(reg:</code><var>m</var> <var>n</var><code>)</code><dd>For small values of the integer <var>n</var> (those that are less than
<code>FIRST_PSEUDO_REGISTER</code>), this stands for a reference to machine
register number <var>n</var>: a <dfn>hard register</dfn>. For larger values of
<var>n</var>, it stands for a temporary value or <dfn>pseudo register</dfn>.
The compiler's strategy is to generate code assuming an unlimited
number of such pseudo registers, and later convert them into hard
registers or into memory references.
<p><var>m</var> is the machine mode of the reference. It is necessary because
machines can generally refer to each register in more than one mode.
For example, a register may contain a full word but there may be
instructions to refer to it as a half word or as a single byte, as
well as instructions to refer to it as a floating point number of
various precisions.
<p>Even for a register that the machine can access in only one mode,
the mode must always be specified.
<p>The symbol <code>FIRST_PSEUDO_REGISTER</code> is defined by the machine
description, since the number of hard registers on the machine is an
invariant characteristic of the machine. Note, however, that not
all of the machine registers must be general registers. All the
machine registers that can be used for storage of data are given
hard register numbers, even those that can be used only in certain
instructions or can hold only certain types of data.
<p>A hard register may be accessed in various modes throughout one
function, but each pseudo register is given a natural mode
and is accessed only in that mode. When it is necessary to describe
an access to a pseudo register using a nonnatural mode, a <code>subreg</code>
expression is used.
<p>A <code>reg</code> expression with a machine mode that specifies more than
one word of data may actually stand for several consecutive registers.
If in addition the register number specifies a hardware register, then
it actually represents several consecutive hardware registers starting
with the specified one.
<p>Each pseudo register number used in a function's RTL code is
represented by a unique <code>reg</code> expression.
<p><a name="index-FIRST_005fVIRTUAL_005fREGISTER-2849"></a><a name="index-LAST_005fVIRTUAL_005fREGISTER-2850"></a>Some pseudo register numbers, those within the range of
<code>FIRST_VIRTUAL_REGISTER</code> to <code>LAST_VIRTUAL_REGISTER</code> only
appear during the RTL generation phase and are eliminated before the
optimization phases. These represent locations in the stack frame that
cannot be determined until RTL generation for the function has been
completed. The following virtual register numbers are defined:
<a name="index-VIRTUAL_005fINCOMING_005fARGS_005fREGNUM-2851"></a>
<dl><dt><code>VIRTUAL_INCOMING_ARGS_REGNUM</code><dd>This points to the first word of the incoming arguments passed on the
stack. Normally these arguments are placed there by the caller, but the
callee may have pushed some arguments that were previously passed in
registers.
<p><a name="index-g_t_0040code_007bFIRST_005fPARM_005fOFFSET_007d-and-virtual-registers-2852"></a><a name="index-g_t_0040code_007bARG_005fPOINTER_005fREGNUM_007d-and-virtual-registers-2853"></a>When RTL generation is complete, this virtual register is replaced
by the sum of the register given by <code>ARG_POINTER_REGNUM</code> and the
value of <code>FIRST_PARM_OFFSET</code>.
<p><a name="index-VIRTUAL_005fSTACK_005fVARS_005fREGNUM-2854"></a><a name="index-g_t_0040code_007bFRAME_005fGROWS_005fDOWNWARD_007d-and-virtual-registers-2855"></a><br><dt><code>VIRTUAL_STACK_VARS_REGNUM</code><dd>If <code>FRAME_GROWS_DOWNWARD</code> is defined to a nonzero value, this points
to immediately above the first variable on the stack. Otherwise, it points
to the first variable on the stack.
<p><a name="index-g_t_0040code_007bSTARTING_005fFRAME_005fOFFSET_007d-and-virtual-registers-2856"></a><a name="index-g_t_0040code_007bFRAME_005fPOINTER_005fREGNUM_007d-and-virtual-registers-2857"></a><code>VIRTUAL_STACK_VARS_REGNUM</code> is replaced with the sum of the
register given by <code>FRAME_POINTER_REGNUM</code> and the value
<code>STARTING_FRAME_OFFSET</code>.
<p><a name="index-VIRTUAL_005fSTACK_005fDYNAMIC_005fREGNUM-2858"></a><br><dt><code>VIRTUAL_STACK_DYNAMIC_REGNUM</code><dd>This points to the location of dynamically allocated memory on the stack
immediately after the stack pointer has been adjusted by the amount of
memory desired.
<p><a name="index-g_t_0040code_007bSTACK_005fDYNAMIC_005fOFFSET_007d-and-virtual-registers-2859"></a><a name="index-g_t_0040code_007bSTACK_005fPOINTER_005fREGNUM_007d-and-virtual-registers-2860"></a>This virtual register is replaced by the sum of the register given by
<code>STACK_POINTER_REGNUM</code> and the value <code>STACK_DYNAMIC_OFFSET</code>.
<p><a name="index-VIRTUAL_005fOUTGOING_005fARGS_005fREGNUM-2861"></a><br><dt><code>VIRTUAL_OUTGOING_ARGS_REGNUM</code><dd>This points to the location in the stack at which outgoing arguments
should be written when the stack is pre-pushed (arguments pushed using
push insns should always use <code>STACK_POINTER_REGNUM</code>).
<p><a name="index-g_t_0040code_007bSTACK_005fPOINTER_005fOFFSET_007d-and-virtual-registers-2862"></a>This virtual register is replaced by the sum of the register given by
<code>STACK_POINTER_REGNUM</code> and the value <code>STACK_POINTER_OFFSET</code>.
</dl>
<p><a name="index-subreg-2863"></a><br><dt><code>(subreg:</code><var>m1</var> <var>reg:m2</var> <var>bytenum</var><code>)</code><dd>
<code>subreg</code> expressions are used to refer to a register in a machine
mode other than its natural one, or to refer to one register of
a multi-part <code>reg</code> that actually refers to several registers.
<p>Each pseudo register has a natural mode. If it is necessary to
operate on it in a different mode, the register must be
enclosed in a <code>subreg</code>.
<p>There are currently three supported types for the first operand of a
<code>subreg</code>:
<ul>
<li>pseudo registers
This is the most common case. Most <code>subreg</code>s have pseudo
<code>reg</code>s as their first operand.
<li>mem
<code>subreg</code>s of <code>mem</code> were common in earlier versions of GCC and
are still supported. During the reload pass these are replaced by plain
<code>mem</code>s. On machines that do not do instruction scheduling, use of
<code>subreg</code>s of <code>mem</code> are still used, but this is no longer
recommended. Such <code>subreg</code>s are considered to be
<code>register_operand</code>s rather than <code>memory_operand</code>s before and
during reload. Because of this, the scheduling passes cannot properly
schedule instructions with <code>subreg</code>s of <code>mem</code>, so for machines
that do scheduling, <code>subreg</code>s of <code>mem</code> should never be used.
To support this, the combine and recog passes have explicit code to
inhibit the creation of <code>subreg</code>s of <code>mem</code> when
<code>INSN_SCHEDULING</code> is defined.
<p>The use of <code>subreg</code>s of <code>mem</code> after the reload pass is an area
that is not well understood and should be avoided. There is still some
code in the compiler to support this, but this code has possibly rotted.
This use of <code>subreg</code>s is discouraged and will most likely not be
supported in the future.
<li>hard registers
It is seldom necessary to wrap hard registers in <code>subreg</code>s; such
registers would normally reduce to a single <code>reg</code> rtx. This use of
<code>subreg</code>s is discouraged and may not be supported in the future.
</ul>
<p><code>subreg</code>s of <code>subreg</code>s are not supported. Using
<code>simplify_gen_subreg</code> is the recommended way to avoid this problem.
<p><code>subreg</code>s come in two distinct flavors, each having its own
usage and rules:
<dl>
<dt>Paradoxical subregs<dd>When <var>m1</var> is strictly wider than <var>m2</var>, the <code>subreg</code>
expression is called <dfn>paradoxical</dfn>. The canonical test for this
class of <code>subreg</code> is:
<pre class="smallexample"> GET_MODE_SIZE (<var>m1</var>) &gt; GET_MODE_SIZE (<var>m2</var>)
</pre>
<p>Paradoxical <code>subreg</code>s can be used as both lvalues and rvalues.
When used as an lvalue, the low-order bits of the source value
are stored in <var>reg</var> and the high-order bits are discarded.
When used as an rvalue, the low-order bits of the <code>subreg</code> are
taken from <var>reg</var> while the high-order bits may or may not be
defined.
<p>The high-order bits of rvalues are in the following circumstances:
<ul>
<li><code>subreg</code>s of <code>mem</code>
When <var>m2</var> is smaller than a word, the macro <code>LOAD_EXTEND_OP</code>,
can control how the high-order bits are defined.
<li><code>subreg</code> of <code>reg</code>s
The upper bits are defined when <code>SUBREG_PROMOTED_VAR_P</code> is true.
<code>SUBREG_PROMOTED_UNSIGNED_P</code> describes what the upper bits hold.
Such subregs usually represent local variables, register variables
and parameter pseudo variables that have been promoted to a wider mode.
</ul>
<p><var>bytenum</var> is always zero for a paradoxical <code>subreg</code>, even on
big-endian targets.
<p>For example, the paradoxical <code>subreg</code>:
<pre class="smallexample"> (set (subreg:SI (reg:HI <var>x</var>) 0) <var>y</var>)
</pre>
<p>stores the lower 2 bytes of <var>y</var> in <var>x</var> and discards the upper
2 bytes. A subsequent:
<pre class="smallexample"> (set <var>z</var> (subreg:SI (reg:HI <var>x</var>) 0))
</pre>
<p>would set the lower two bytes of <var>z</var> to <var>y</var> and set the upper
two bytes to an unknown value assuming <code>SUBREG_PROMOTED_VAR_P</code> is
false.
<br><dt>Normal subregs<dd>When <var>m1</var> is at least as narrow as <var>m2</var> the <code>subreg</code>
expression is called <dfn>normal</dfn>.
<p>Normal <code>subreg</code>s restrict consideration to certain bits of
<var>reg</var>. There are two cases. If <var>m1</var> is smaller than a word,
the <code>subreg</code> refers to the least-significant part (or
<dfn>lowpart</dfn>) of one word of <var>reg</var>. If <var>m1</var> is word-sized or
greater, the <code>subreg</code> refers to one or more complete words.
<p>When used as an lvalue, <code>subreg</code> is a word-based accessor.
Storing to a <code>subreg</code> modifies all the words of <var>reg</var> that
overlap the <code>subreg</code>, but it leaves the other words of <var>reg</var>
alone.
<p>When storing to a normal <code>subreg</code> that is smaller than a word,
the other bits of the referenced word are usually left in an undefined
state. This laxity makes it easier to generate efficient code for
such instructions. To represent an instruction that preserves all the
bits outside of those in the <code>subreg</code>, use <code>strict_low_part</code>
or <code>zero_extract</code> around the <code>subreg</code>.
<p><var>bytenum</var> must identify the offset of the first byte of the
<code>subreg</code> from the start of <var>reg</var>, assuming that <var>reg</var> is
laid out in memory order. The memory order of bytes is defined by
two target macros, <code>WORDS_BIG_ENDIAN</code> and <code>BYTES_BIG_ENDIAN</code>:
<ul>
<li><a name="index-g_t_0040code_007bWORDS_005fBIG_005fENDIAN_007d_002c-effect-on-_0040code_007bsubreg_007d-2864"></a><code>WORDS_BIG_ENDIAN</code>, if set to 1, says that byte number zero is
part of the most significant word; otherwise, it is part of the least
significant word.
<li><a name="index-g_t_0040code_007bBYTES_005fBIG_005fENDIAN_007d_002c-effect-on-_0040code_007bsubreg_007d-2865"></a><code>BYTES_BIG_ENDIAN</code>, if set to 1, says that byte number zero is
the most significant byte within a word; otherwise, it is the least
significant byte within a word.
</ul>
<p><a name="index-g_t_0040code_007bFLOAT_005fWORDS_005fBIG_005fENDIAN_007d_002c-_0028lack-of_0029-effect-on-_0040code_007bsubreg_007d-2866"></a>On a few targets, <code>FLOAT_WORDS_BIG_ENDIAN</code> disagrees with
<code>WORDS_BIG_ENDIAN</code>. However, most parts of the compiler treat
floating point values as if they had the same endianness as integer
values. This works because they handle them solely as a collection of
integer values, with no particular numerical value. Only real.c and
the runtime libraries care about <code>FLOAT_WORDS_BIG_ENDIAN</code>.
<p>Thus,
<pre class="smallexample"> (subreg:HI (reg:SI <var>x</var>) 2)
</pre>
<p>on a <code>BYTES_BIG_ENDIAN</code>, &lsquo;<samp><span class="samp">UNITS_PER_WORD == 4</span></samp>&rsquo; target is the same as
<pre class="smallexample"> (subreg:HI (reg:SI <var>x</var>) 0)
</pre>
<p>on a little-endian, &lsquo;<samp><span class="samp">UNITS_PER_WORD == 4</span></samp>&rsquo; target. Both
<code>subreg</code>s access the lower two bytes of register <var>x</var>.
</dl>
<p>A <code>MODE_PARTIAL_INT</code> mode behaves as if it were as wide as the
corresponding <code>MODE_INT</code> mode, except that it has an unknown
number of undefined bits. For example:
<pre class="smallexample"> (subreg:PSI (reg:SI 0) 0)
</pre>
<p>accesses the whole of &lsquo;<samp><span class="samp">(reg:SI 0)</span></samp>&rsquo;, but the exact relationship
between the <code>PSImode</code> value and the <code>SImode</code> value is not
defined. If we assume &lsquo;<samp><span class="samp">UNITS_PER_WORD &lt;= 4</span></samp>&rsquo;, then the following
two <code>subreg</code>s:
<pre class="smallexample"> (subreg:PSI (reg:DI 0) 0)
(subreg:PSI (reg:DI 0) 4)
</pre>
<p>represent independent 4-byte accesses to the two halves of
&lsquo;<samp><span class="samp">(reg:DI 0)</span></samp>&rsquo;. Both <code>subreg</code>s have an unknown number
of undefined bits.
<p>If &lsquo;<samp><span class="samp">UNITS_PER_WORD &lt;= 2</span></samp>&rsquo; then these two <code>subreg</code>s:
<pre class="smallexample"> (subreg:HI (reg:PSI 0) 0)
(subreg:HI (reg:PSI 0) 2)
</pre>
<p>represent independent 2-byte accesses that together span the whole
of &lsquo;<samp><span class="samp">(reg:PSI 0)</span></samp>&rsquo;. Storing to the first <code>subreg</code> does not
affect the value of the second, and vice versa. &lsquo;<samp><span class="samp">(reg:PSI 0)</span></samp>&rsquo;
has an unknown number of undefined bits, so the assignment:
<pre class="smallexample"> (set (subreg:HI (reg:PSI 0) 0) (reg:HI 4))
</pre>
<p>does not guarantee that &lsquo;<samp><span class="samp">(subreg:HI (reg:PSI 0) 0)</span></samp>&rsquo; has the
value &lsquo;<samp><span class="samp">(reg:HI 4)</span></samp>&rsquo;.
<p><a name="index-g_t_0040code_007bCANNOT_005fCHANGE_005fMODE_005fCLASS_007d-and-subreg-semantics-2867"></a>The rules above apply to both pseudo <var>reg</var>s and hard <var>reg</var>s.
If the semantics are not correct for particular combinations of
<var>m1</var>, <var>m2</var> and hard <var>reg</var>, the target-specific code
must ensure that those combinations are never used. For example:
<pre class="smallexample"> CANNOT_CHANGE_MODE_CLASS (<var>m2</var>, <var>m1</var>, <var>class</var>)
</pre>
<p>must be true for every class <var>class</var> that includes <var>reg</var>.
<p><a name="index-SUBREG_005fREG-2868"></a><a name="index-SUBREG_005fBYTE-2869"></a>The first operand of a <code>subreg</code> expression is customarily accessed
with the <code>SUBREG_REG</code> macro and the second operand is customarily
accessed with the <code>SUBREG_BYTE</code> macro.
<p>It has been several years since a platform in which
<code>BYTES_BIG_ENDIAN</code> not equal to <code>WORDS_BIG_ENDIAN</code> has
been tested. Anyone wishing to support such a platform in the future
may be confronted with code rot.
<p><a name="index-scratch-2870"></a><a name="index-scratch-operands-2871"></a><br><dt><code>(scratch:</code><var>m</var><code>)</code><dd>This represents a scratch register that will be required for the
execution of a single instruction and not used subsequently. It is
converted into a <code>reg</code> by either the local register allocator or
the reload pass.
<p><code>scratch</code> is usually present inside a <code>clobber</code> operation
(see <a href="Side-Effects.html#Side-Effects">Side Effects</a>).
<p><a name="index-cc0-2872"></a><a name="index-condition-code-register-2873"></a><br><dt><code>(cc0)</code><dd>This refers to the machine's condition code register. It has no
operands and may not have a machine mode. There are two ways to use it:
<ul>
<li>To stand for a complete set of condition code flags. This is best on
most machines, where each comparison sets the entire series of flags.
<p>With this technique, <code>(cc0)</code> may be validly used in only two
contexts: as the destination of an assignment (in test and compare
instructions) and in comparison operators comparing against zero
(<code>const_int</code> with value zero; that is to say, <code>const0_rtx</code>).
<li>To stand for a single flag that is the result of a single condition.
This is useful on machines that have only a single flag bit, and in
which comparison instructions must specify the condition to test.
<p>With this technique, <code>(cc0)</code> may be validly used in only two
contexts: as the destination of an assignment (in test and compare
instructions) where the source is a comparison operator, and as the
first operand of <code>if_then_else</code> (in a conditional branch).
</ul>
<p><a name="index-cc0_005frtx-2874"></a>There is only one expression object of code <code>cc0</code>; it is the
value of the variable <code>cc0_rtx</code>. Any attempt to create an
expression of code <code>cc0</code> will return <code>cc0_rtx</code>.
<p>Instructions can set the condition code implicitly. On many machines,
nearly all instructions set the condition code based on the value that
they compute or store. It is not necessary to record these actions
explicitly in the RTL because the machine description includes a
prescription for recognizing the instructions that do so (by means of
the macro <code>NOTICE_UPDATE_CC</code>). See <a href="Condition-Code.html#Condition-Code">Condition Code</a>. Only
instructions whose sole purpose is to set the condition code, and
instructions that use the condition code, need mention <code>(cc0)</code>.
<p>On some machines, the condition code register is given a register number
and a <code>reg</code> is used instead of <code>(cc0)</code>. This is usually the
preferable approach if only a small subset of instructions modify the
condition code. Other machines store condition codes in general
registers; in such cases a pseudo register should be used.
<p>Some machines, such as the SPARC and RS/6000, have two sets of
arithmetic instructions, one that sets and one that does not set the
condition code. This is best handled by normally generating the
instruction that does not set the condition code, and making a pattern
that both performs the arithmetic and sets the condition code register
(which would not be <code>(cc0)</code> in this case). For examples, search
for &lsquo;<samp><span class="samp">addcc</span></samp>&rsquo; and &lsquo;<samp><span class="samp">andcc</span></samp>&rsquo; in <samp><span class="file">sparc.md</span></samp>.
<p><a name="index-pc-2875"></a><br><dt><code>(pc)</code><dd><a name="index-program-counter-2876"></a>This represents the machine's program counter. It has no operands and
may not have a machine mode. <code>(pc)</code> may be validly used only in
certain specific contexts in jump instructions.
<p><a name="index-pc_005frtx-2877"></a>There is only one expression object of code <code>pc</code>; it is the value
of the variable <code>pc_rtx</code>. Any attempt to create an expression of
code <code>pc</code> will return <code>pc_rtx</code>.
<p>All instructions that do not jump alter the program counter implicitly
by incrementing it, but there is no need to mention this in the RTL.
<p><a name="index-mem-2878"></a><br><dt><code>(mem:</code><var>m</var> <var>addr</var> <var>alias</var><code>)</code><dd>This RTX represents a reference to main memory at an address
represented by the expression <var>addr</var>. <var>m</var> specifies how large
a unit of memory is accessed. <var>alias</var> specifies an alias set for the
reference. In general two items are in different alias sets if they cannot
reference the same memory address.
<p>The construct <code>(mem:BLK (scratch))</code> is considered to alias all
other memories. Thus it may be used as a memory barrier in epilogue
stack deallocation patterns.
<p><a name="index-concat-2879"></a><br><dt><code>(concat</code><var>m</var> <var>rtx</var> <var>rtx</var><code>)</code><dd>This RTX represents the concatenation of two other RTXs. This is used
for complex values. It should only appear in the RTL attached to
declarations and during RTL generation. It should not appear in the
ordinary insn chain.
<p><a name="index-concatn-2880"></a><br><dt><code>(concatn</code><var>m</var><code> [</code><var>rtx</var><code> ...])</code><dd>This RTX represents the concatenation of all the <var>rtx</var> to make a
single value. Like <code>concat</code>, this should only appear in
declarations, and not in the insn chain.
</dl>
</body></html>