doc/gdb/Rationale.html

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd">
<html>
<!-- Copyright (C) 1988-2021 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 "Free Software" and "Free Software Needs
Free Documentation", with the Front-Cover Texts being "A GNU Manual,"
and with the Back-Cover Texts as in (a) below.

(a) The FSF's Back-Cover Text is: "You are free to copy and modify
this GNU Manual.  Buying copies from GNU Press supports the FSF in
developing GNU and promoting software freedom." -->
<!-- Created by GNU Texinfo 5.1, http://www.gnu.org/software/texinfo/ -->
<head>
<title>Debugging with GDB: Rationale</title>

<meta name="description" content="Debugging with GDB: Rationale">
<meta name="keywords" content="Debugging with GDB: Rationale">
<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="Concept-Index.html#Concept-Index" rel="index" title="Concept Index">
<link href="index.html#SEC_Contents" rel="contents" title="Table of Contents">
<link href="Agent-Expressions.html#Agent-Expressions" rel="up" title="Agent Expressions">
<link href="Target-Descriptions.html#Target-Descriptions" rel="next" title="Target Descriptions">
<link href="Varying-Target-Capabilities.html#Varying-Target-Capabilities" rel="previous" title="Varying Target Capabilities">
<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="Rationale"></a>
<div class="header">
<p>
Previous: <a href="Varying-Target-Capabilities.html#Varying-Target-Capabilities" accesskey="p" rel="previous">Varying Target Capabilities</a>, Up: <a href="Agent-Expressions.html#Agent-Expressions" accesskey="u" rel="up">Agent Expressions</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Concept-Index.html#Concept-Index" title="Index" rel="index">Index</a>]</p>
</div>
<hr>
<a name="Rationale-1"></a>
<h3 class="section">F.5 Rationale</h3>

<p>Some of the design decisions apparent above are arguable.
</p>
<dl compact="compact">
<dt><b>What about stack overflow/underflow?</b></dt>
<dd><p>GDB should be able to query the target to discover its stack size.
Given that information, GDB can determine at translation time whether a
given expression will overflow the stack.  But this spec isn&rsquo;t about
what kinds of error-checking GDB ought to do.
</p>
</dd>
<dt><b>Why are you doing everything in LONGEST?</b></dt>
<dd>
<p>Speed isn&rsquo;t important, but agent code size is; using LONGEST brings in a
bunch of support code to do things like division, etc.  So this is a
serious concern.
</p>
<p>First, note that you don&rsquo;t need different bytecodes for different
operand sizes.  You can generate code without <em>knowing</em> how big the
stack elements actually are on the target.  If the target only supports
32-bit ints, and you don&rsquo;t send any 64-bit bytecodes, everything just
works.  The observation here is that the MIPS and the Alpha have only
fixed-size registers, and you can still get C&rsquo;s semantics even though
most instructions only operate on full-sized words.  You just need to
make sure everything is properly sign-extended at the right times.  So
there is no need for 32- and 64-bit variants of the bytecodes.  Just
implement everything using the largest size you support.
</p>
<p>GDB should certainly check to see what sizes the target supports, so the
user can get an error earlier, rather than later.  But this information
is not necessary for correctness.
</p>

</dd>
<dt><b>Why don&rsquo;t you have <code>&gt;</code> or <code>&lt;=</code> operators?</b></dt>
<dd><p>I want to keep the interpreter small, and we don&rsquo;t need them.  We can
combine the <code>less_</code> opcodes with <code>log_not</code>, and swap the order
of the operands, yielding all four asymmetrical comparison operators.
For example, <code>(x &lt;= y)</code> is <code>! (x &gt; y)</code>, which is <code>! (y &lt;
x)</code>.
</p>
</dd>
<dt><b>Why do you have <code>log_not</code>?</b></dt>
<dt><b>Why do you have <code>ext</code>?</b></dt>
<dt><b>Why do you have <code>zero_ext</code>?</b></dt>
<dd><p>These are all easily synthesized from other instructions, but I expect
them to be used frequently, and they&rsquo;re simple, so I include them to
keep bytecode strings short.
</p>
<p><code>log_not</code> is equivalent to <code>const8 0 equal</code>; it&rsquo;s used in half
the relational operators.
</p>
<p><code>ext <var>n</var></code> is equivalent to <code>const8 <var>s-n</var> lsh const8
<var>s-n</var> rsh_signed</code>, where <var>s</var> is the size of the stack elements;
it follows <code>ref<var>m</var></code> and <var>reg</var> bytecodes when the value
should be signed.  See the next bulleted item.
</p>
<p><code>zero_ext <var>n</var></code> is equivalent to <code>const<var>m</var> <var>mask</var>
log_and</code>; it&rsquo;s used whenever we push the value of a register, because we
can&rsquo;t assume the upper bits of the register aren&rsquo;t garbage.
</p>
</dd>
<dt><b>Why not have sign-extending variants of the <code>ref</code> operators?</b></dt>
<dd><p>Because that would double the number of <code>ref</code> operators, and we
need the <code>ext</code> bytecode anyway for accessing bitfields.
</p>
</dd>
<dt><b>Why not have constant-address variants of the <code>ref</code> operators?</b></dt>
<dd><p>Because that would double the number of <code>ref</code> operators again, and
<code>const32 <var>address</var> ref32</code> is only one byte longer.
</p>
</dd>
<dt><b>Why do the <code>ref<var>n</var></code> operators have to support unaligned fetches?</b></dt>
<dd><p>GDB will generate bytecode that fetches multi-byte values at unaligned
addresses whenever the executable&rsquo;s debugging information tells it to.
Furthermore, GDB does not know the value the pointer will have when GDB
generates the bytecode, so it cannot determine whether a particular
fetch will be aligned or not.
</p>
<p>In particular, structure bitfields may be several bytes long, but follow
no alignment rules; members of packed structures are not necessarily
aligned either.
</p>
<p>In general, there are many cases where unaligned references occur in
correct C code, either at the programmer&rsquo;s explicit request, or at the
compiler&rsquo;s discretion.  Thus, it is simpler to make the GDB agent
bytecodes work correctly in all circumstances than to make GDB guess in
each case whether the compiler did the usual thing.
</p>
</dd>
<dt><b>Why are there no side-effecting operators?</b></dt>
<dd><p>Because our current client doesn&rsquo;t want them?  That&rsquo;s a cheap answer.  I
think the real answer is that I&rsquo;m afraid of implementing function
calls.  We should re-visit this issue after the present contract is
delivered.
</p>
</dd>
<dt><b>Why aren&rsquo;t the <code>goto</code> ops PC-relative?</b></dt>
<dd><p>The interpreter has the base address around anyway for PC bounds
checking, and it seemed simpler.
</p>
</dd>
<dt><b>Why is there only one offset size for the <code>goto</code> ops?</b></dt>
<dd><p>Offsets are currently sixteen bits.  I&rsquo;m not happy with this situation
either:
</p>
<p>Suppose we have multiple branch ops with different offset sizes.  As I
generate code left-to-right, all my jumps are forward jumps (there are
no loops in expressions), so I never know the target when I emit the
jump opcode.  Thus, I have to either always assume the largest offset
size, or do jump relaxation on the code after I generate it, which seems
like a big waste of time.
</p>
<p>I can imagine a reasonable expression being longer than 256 bytes.  I
can&rsquo;t imagine one being longer than 64k.  Thus, we need 16-bit offsets.
This kind of reasoning is so bogus, but relaxation is pathetic.
</p>
<p>The other approach would be to generate code right-to-left.  Then I&rsquo;d
always know my offset size.  That might be fun.
</p>
</dd>
<dt><b>Where is the function call bytecode?</b></dt>
<dd>
<p>When we add side-effects, we should add this.
</p>
</dd>
<dt><b>Why does the <code>reg</code> bytecode take a 16-bit register number?</b></dt>
<dd>
<p>Intel&rsquo;s IA-64 architecture has 128 general-purpose registers,
and 128 floating-point registers, and I&rsquo;m sure it has some random
control registers.
</p>
</dd>
<dt><b>Why do we need <code>trace</code> and <code>trace_quick</code>?</b></dt>
<dd><p>Because GDB needs to record all the memory contents and registers an
expression touches.  If the user wants to evaluate an expression
<code>x-&gt;y-&gt;z</code>, the agent must record the values of <code>x</code> and
<code>x-&gt;y</code> as well as the value of <code>x-&gt;y-&gt;z</code>.
</p>
</dd>
<dt><b>Don&rsquo;t the <code>trace</code> bytecodes make the interpreter less general?</b></dt>
<dd><p>They do mean that the interpreter contains special-purpose code, but
that doesn&rsquo;t mean the interpreter can only be used for that purpose.  If
an expression doesn&rsquo;t use the <code>trace</code> bytecodes, they don&rsquo;t get in
its way.
</p>
</dd>
<dt><b>Why doesn&rsquo;t <code>trace_quick</code> consume its arguments the way everything else does?</b></dt>
<dd><p>In general, you do want your operators to consume their arguments; it&rsquo;s
consistent, and generally reduces the amount of stack rearrangement
necessary.  However, <code>trace_quick</code> is a kludge to save space; it
only exists so we needn&rsquo;t write <code>dup const8 <var>SIZE</var> trace</code>
before every memory reference.  Therefore, it&rsquo;s okay for it not to
consume its arguments; it&rsquo;s meant for a specific context in which we
know exactly what it should do with the stack.  If we&rsquo;re going to have a
kludge, it should be an effective kludge.
</p>
</dd>
<dt><b>Why does <code>trace16</code> exist?</b></dt>
<dd><p>That opcode was added by the customer that contracted Cygnus for the
data tracing work.  I personally think it is unnecessary; objects that
large will be quite rare, so it is okay to use <code>dup const16
<var>size</var> trace</code> in those cases.
</p>
<p>Whatever we decide to do with <code>trace16</code>, we should at least leave
opcode 0x30 reserved, to remain compatible with the customer who added
it.
</p>
</dd>
</dl>

<hr>
<div class="header">
<p>
Previous: <a href="Varying-Target-Capabilities.html#Varying-Target-Capabilities" accesskey="p" rel="previous">Varying Target Capabilities</a>, Up: <a href="Agent-Expressions.html#Agent-Expressions" accesskey="u" rel="up">Agent Expressions</a> &nbsp; [<a href="index.html#SEC_Contents" title="Table of contents" rel="contents">Contents</a>][<a href="Concept-Index.html#Concept-Index" title="Index" rel="index">Index</a>]</p>
</div>


</body>
</html>