<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">

<html>

<head>
<title>cmp&lt;kind&gt;</title>
<link rel=stylesheet href="opcode.css">
</head>

<body>

<h1>cmp&lt;kind&gt;</h1>

<h2>Purpose</h2>

<p>
Perform the indicated floating point or long comparison, storing 0 if the two
arguments are equal, 1 if the second argument is larger, or -1 if the first
argument is larger. The "bias" listed for the floating point operations
indicates how NaN comparisons are treated: "Gt bias" instructions return 1 for
NaN comparisons, and "lt bias" instructions return -1.
</p>
<p>
For example, to check to see if floating point a < b, then it is advisable to
use cmpg-float; a result of -1 indicates that the test was true, and the other
values indicate it was false either due to a valid comparison or because one
or the other values was NaN.
</p>

<h2>Details</h2>

<table class="instruc">
<thead>
<tr>
  <th>Op &amp; Format</th>
  <th>Mnemonic / Syntax</th>
  <th>Arguments</th>
</tr>
</thead>
<tbody>
<tr>
  <td>2d..31 23x</td>
  <td>cmp<i>kind</i> vAA, vBB, vCC<br/>
    2d: cmpl-float <i>(lt bias)</i><br/>
    2e: cmpg-float <i>(gt bias)</i><br/>
    2f: cmpl-double <i>(lt bias)</i><br/>
    30: cmpg-double <i>(gt bias)</i><br/>
    31: cmp-long
  </td>
  <td><code>A:</code> destination register (8 bits)<br/>
    <code>B:</code> first source register or pair<br/>
    <code>C:</code> second source register or pair</td>
</tr>
</tbody>
</table>

<h2>Constraints</h2>

<ul>
  <li>
    A, B and C must be valid register indices in the current stack frame.
  </li>
  <li>
    For the two -float variants, both vB and vC must be of type float.
  </li>
  <li>
    For the two -double variants, both vB and vC must be the lower part of a
    register pair holding a double value.
  </li>
  <li>
    For the -long variant, both both vB and vC must be the lower part of a
    register pair holding a long value.
  </li>
</ul>

<h2>Behavior</h2>

<ul>
  <li>
    The values of registers vB and vC are compared. The result, which is stored
    in vA, is one of the following three:
    <ul>
      <li>
        If vB < vC, then vA'=-1.
      </li>
      <li>
        If vB == vC, then vA'=0.
      </li>
      <li>
        If vC > vC, then vA'=1.
      </li>
    </ul>
  </li>
  <li>
    For the -float and -double variants, an addition "bias" specifies what
    happens if one or both of the arguments are NaN:
    <ul>
      <li>
        A "lt bias" results in vA'=-1.
      </li>
      <li>
        A "gt bias" results in vA'=1.
      </li>
    </ul>
  </li>
  <li>
    If v(A-1) is the lower half of a register pair, v(A-1)' becomes undefined.
  </li>
  <li>
    If v(A+1) is the upper half of a register pair, v(A+1)' becomes undefined.
  </li>
</ul>

<h2>Exceptions</h2>

<p>
None.
</p>

</body>
</html>