<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Intermediate Representation on Nalar</title>
    <link>https://nalar.dev/tags/intermediate-representation/</link>
    <description>Recent content in Intermediate Representation on Nalar</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Sat, 19 Sep 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://nalar.dev/tags/intermediate-representation/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>LLVM Poison Values Defer Undefined Behavior Through the IR</title>
      <link>https://nalar.dev/llvm-poison-values-defer-undefined-behavior-through-the-ir/</link>
      <pubDate>Sat, 19 Sep 2026 00:00:00 +0000</pubDate>
      <guid>https://nalar.dev/llvm-poison-values-defer-undefined-behavior-through-the-ir/</guid>
      <description>&lt;p&gt;LLVM IR does not turn every invalid arithmetic or pointer condition into immediate undefined behavior. Many instructions produce a &lt;code&gt;poison&lt;/code&gt; value instead. That value can flow through later instructions, preserving the optimizer&amp;rsquo;s ability to rely on promises such as &amp;ldquo;this addition does not overflow&amp;rdquo; without forcing undefined behavior at the exact instruction where the promise is violated.&lt;/p&gt;&#xA;&lt;p&gt;This is part of LLVM&amp;rsquo;s semantics, not an optimizer implementation detail. A frontend that emits &lt;code&gt;nsw&lt;/code&gt;, &lt;code&gt;nuw&lt;/code&gt;, &lt;code&gt;inbounds&lt;/code&gt;, &lt;code&gt;noundef&lt;/code&gt;, or related constraints is stating facts that later passes may trust.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
