<html>
<head>
<base href="https://bugs.llvm.org/">
</head>
<body><span class="vcard"><a class="email" href="mailto:spatel+llvm@rotateright.com" title="Sanjay Patel <spatel+llvm@rotateright.com>"> <span class="fn">Sanjay Patel</span></a>
</span> changed
<a class="bz_bug_link
bz_status_RESOLVED bz_closed"
title="RESOLVED FIXED - Miscompile with SLP vectorizer + instcombine"
href="https://bugs.llvm.org/show_bug.cgi?id=50500">bug 50500</a>
<br>
<table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>What</th>
<th>Removed</th>
<th>Added</th>
</tr>
<tr>
<td style="text-align:right;">Status</td>
<td>NEW
</td>
<td>RESOLVED
</td>
</tr>
<tr>
<td style="text-align:right;">Fixed By Commit(s)</td>
<td>
</td>
<td>7bb8bfa0622b
</td>
</tr>
<tr>
<td style="text-align:right;">Resolution</td>
<td>---
</td>
<td>FIXED
</td>
</tr></table>
<p>
<div>
<b><a class="bz_bug_link
bz_status_RESOLVED bz_closed"
title="RESOLVED FIXED - Miscompile with SLP vectorizer + instcombine"
href="https://bugs.llvm.org/show_bug.cgi?id=50500#c2">Comment # 2</a>
on <a class="bz_bug_link
bz_status_RESOLVED bz_closed"
title="RESOLVED FIXED - Miscompile with SLP vectorizer + instcombine"
href="https://bugs.llvm.org/show_bug.cgi?id=50500">bug 50500</a>
from <span class="vcard"><a class="email" href="mailto:spatel+llvm@rotateright.com" title="Sanjay Patel <spatel+llvm@rotateright.com>"> <span class="fn">Sanjay Patel</span></a>
</span></b>
<pre>As with <a class="bz_bug_link
bz_status_RESOLVED bz_closed"
title="RESOLVED FIXED - InstCombine incorrectly folds select of vector"
href="show_bug.cgi?id=49832">bug 49832</a>, it's more general than partial undef - we don't account for
vector transforms in our equivalence substituation logic:
<a href="https://reviews.llvm.org/rG7bb8bfa0622b">https://reviews.llvm.org/rG7bb8bfa0622b</a>
I audited callers and don't see any other places we could hit this, so I put an
assert in the simplify code.
If we want this kind of transform for vectors, we might try to handle splats
only, but I don't see any evidence for it.
I think this bug was completely hidden before D101191, so it doesn't affect
release branches, but feel free to reopen if that's not right.</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>