<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 - if (auto x = y(z)) generates bad/slow code (found via analysis of dyn_cast)"
href="https://bugs.llvm.org/show_bug.cgi?id=35790">35790</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>if (auto x = y(z)) generates bad/slow code (found via analysis of dyn_cast)
</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>All
</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>Interprocedural Optimizations
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>dave@znu.io
</td>
</tr>
<tr>
<th>CC</th>
<td>llvm-bugs@lists.llvm.org
</td>
</tr></table>
<p>
<div>
<pre>While analyzing Swift compiler performance, I found that that LLVM's dyn_cast
generates suboptimal and surprising code gen. I've reduced it down to a test
case that still reproduces on top-of-tree on my Linux box when compiling with
-O3:
//__attribute__((used,noinline))
int *x(void *arg) {
return ((long long)arg & 1) ? (int *)arg : nullptr;
}
int test(void *arg) {
if (auto y = x(arg)) return *y;
return 42;
}
If 'x' is inlined, then the compiler effectively generates the following
surprising pseudo code:
int test(void *arg) {
if (arg != nullptr)
if (arg & 1)
return *arg;
return 42;
}
If 'x' is not inlined. then 'arg' passed to 'test' is not (as expected) NULL
tested by either 'test' or 'x'.
This seems like a a weird inlining versus "if (auto x = y(z))" bug.</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>