<html>
<head>
<base href="https://bugs.llvm.org/">
</head>
<body><table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>Bug ID</th>
<td><a class="bz_bug_link
bz_status_NEW "
title="NEW - LoopIdiomRecognize can turn infinite loops into countable loop using ctlz"
href="https://bugs.llvm.org/show_bug.cgi?id=37479">37479</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>LoopIdiomRecognize can turn infinite loops into countable loop using ctlz
</td>
</tr>
<tr>
<th>Product</th>
<td>libraries
</td>
</tr>
<tr>
<th>Version</th>
<td>trunk
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>Windows NT
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>enhancement
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>Scalar Optimizations
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>craig.topper@gmail.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvm-bugs@lists.llvm.org
</td>
</tr></table>
<p>
<div>
<pre>On targets that support ctlz, loop idiom recognize can transform code like this
int foo(int x) {
int cnt = 0;
while (x) {
x >>= 1;
++cnt;
}
return cnt;
}
Into just returning (32 - ctlz(x)). Or if the loop contains other code it will
create a loop that runs (32 - ctlz(x)) times.
But if x was negative the original code would have been an infinite loop since
the shift would just keep copying the sign bit preventing x from ever being 0.
This transform was added by D32605 and the motivating case cited there looks
pretty much like above with an additional "if (x < 0) x = -x;" before the loop.
Ideally I'd like to not lose the optimization. Are there any things we can
exploit to ensure this is safe for that motivating case.</pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are on the CC list for the bug.</li>
</ul>
</body>
</html>