<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Tcp on Luther Monson</title>
    <link>https://luth.io/tags/tcp/</link>
    <description>Recent content in Tcp on Luther Monson</description>
    <generator>Hugo</generator>
    <language>en</language>
    <lastBuildDate>Fri, 18 Sep 2026 20:05:00 -0700</lastBuildDate>
    <atom:link href="https://luth.io/tags/tcp/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>ePHPm Odds and Ends: OOM-Killed Links, Opcode Scanning, WebSockets and a 44 ms Stall</title>
      <link>https://luth.io/blog/2026/09/ephpm-smaller-angles/</link>
      <pubDate>Fri, 18 Sep 2026 20:05:00 -0700</pubDate>
      <guid>https://luth.io/blog/2026/09/ephpm-smaller-angles/</guid>
      <description>&lt;p&gt;Not every fix in &lt;a href=&#34;https://github.com/ephpm/ephpm&#34;&gt;ePHPm&lt;/a&gt; earns a full write-up. A few of the smaller ones have stuck&#xA;with me because the lesson in each was bigger than the diff. Here are four from this summer.&lt;/p&gt;&#xA;&lt;h3 id=&#34;a-release-build-the-oom-killer-kept-eating&#34;&gt;A release build the OOM killer kept eating&lt;/h3&gt;&#xA;&lt;p&gt;ePHPm statically links libphp, ICU and a pile of other C libraries into one Rust binary. It ships for four targets.&#xA;For the v0.4.2 performance release I set the release profile to &lt;code&gt;lto = &amp;quot;fat&amp;quot;&lt;/code&gt; with &lt;code&gt;codegen-units = 1&lt;/code&gt;&#xA;(&lt;a href=&#34;https://github.com/ephpm/ephpm/pull/171&#34;&gt;#171&lt;/a&gt;). Fat LTO pulls every crate&amp;rsquo;s LLVM IR into one module in one process&#xA;and optimizes it as a unit. One codegen unit removes the parallelism that would otherwise spread that memory out.&#xA;That&amp;rsquo;s one very large process at the final link.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
