<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 - llvm-mca markers interfere with loop vectorization"
href="https://bugs.llvm.org/show_bug.cgi?id=42173">42173</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>llvm-mca markers interfere with loop vectorization
</td>
</tr>
<tr>
<th>Product</th>
<td>libraries
</td>
</tr>
<tr>
<th>Version</th>
<td>8.0
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>Linux
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>normal
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>Loop Optimizer
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>max@marrone.nyc
</td>
</tr>
<tr>
<th>CC</th>
<td>llvm-bugs@lists.llvm.org
</td>
</tr></table>
<p>
<div>
<pre>The docs for llvm-mca
(<a href="https://llvm.org/docs/CommandGuide/llvm-mca.html#using-markers-to-analyze-specific-code-blocks">https://llvm.org/docs/CommandGuide/llvm-mca.html#using-markers-to-analyze-specific-code-blocks</a>)
suggest using inline assembly to mark the region that llvm-mca should examine,
e.g.
<span class="quote">> for(size_t index = 0; index < count; index++)
> {
> __asm volatile("# LLVM-MCA-BEGIN");
> result += source[index];
> __asm volatile("# LLVM-MCA-END");
> }</span >
However, these directives prevent auto-vectorization.
<span class="quote">> <source>:8:3: remark: loop not vectorized: call instruction cannot be
> vectorized [-Rpass-analysis=loop-vectorize]
> __asm volatile("# LLVM-MCA-BEGIN sum_marked");
> ^</span >
<span class="quote">> <source>:6:2: remark: loop not vectorized: read with atomic ordering or
> volatile read [-Rpass-analysis=loop-vectorize]
> for (size_t index = 0; index < count; index++)
> ^</span >
Compiler Explorer demo, on Clang 8.0.0: <a href="https://godbolt.org/z/NSQchu">https://godbolt.org/z/NSQchu</a>
We should be able to use llvm-mca markers without affecting optimization and
code generation.</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>