<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Lock-Free on LCS.Developer</title><link>https://www.lcs-dev.org/tags/lock-free/</link><description>Recent content in Lock-Free on LCS.Developer</description><generator>Hugo -- gohugo.io</generator><language>en-us</language><copyright>LCS.Developer</copyright><lastBuildDate>Tue, 10 Dec 2024 20:00:00 +0000</lastBuildDate><atom:link href="https://www.lcs-dev.org/tags/lock-free/index.xml" rel="self" type="application/rss+xml"/><item><title>Beyond Semaphores: Condition Variables and Lock-Free Ring Buffers</title><link>https://www.lcs-dev.org/posts/beyond-semaphores---condition-variables-and-lock-free-ring-buffers/</link><pubDate>Tue, 10 Dec 2024 20:00:00 +0000</pubDate><guid>https://www.lcs-dev.org/posts/beyond-semaphores---condition-variables-and-lock-free-ring-buffers/</guid><description>&lt;h2 id="where-counting-semaphores-hit-their-limits"&gt;Where Counting Semaphores Hit Their Limits
&lt;/h2&gt;&lt;p&gt;Counting semaphores are a natural fit when synchronization reduces to a single integer counter (such as available buffer slots). However, two common engineering requirements quickly push beyond what semaphores express cleanly:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Compound State Predicates and Graceful Shutdown&lt;/strong&gt;: In a Multi-Producer / Multi-Consumer (MPMC) thread pool, a worker thread typically waits for &lt;code&gt;(count &amp;gt; 0 || shutdown_requested)&lt;/code&gt;. Encoding compound boolean predicates across separate counting semaphores without race conditions or &amp;ldquo;poison-pill&amp;rdquo; dummy messages is brittle.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Sub-Microsecond Latency (Zero Kernel Traps)&lt;/strong&gt;: Contended semaphore operations trap into the OS kernel (&lt;code&gt;futex&lt;/code&gt; on Linux), costing thousands of clock cycles. In real-time audio, high-frequency telemetry, or network packet rings, a Single-Producer / Single-Consumer (SPSC) channel should transfer data across cores in a few nanoseconds &lt;strong&gt;without locks or system calls&lt;/strong&gt;.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Let&amp;rsquo;s build the idiomatic C solution for each regime: &lt;strong&gt;Monitors with POSIX Condition Variables&lt;/strong&gt; for MPMC thread pools, and a &lt;strong&gt;Wait-Free Atomic Ring Buffer&lt;/strong&gt; using C11 acquire-release memory ordering.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="1-mpmc-thread-queues-with-condition-variables"&gt;1. MPMC Thread Queues with Condition Variables
&lt;/h2&gt;&lt;p&gt;A &lt;strong&gt;Monitor&lt;/strong&gt; pairs a mutex (&lt;code&gt;pthread_mutex_t&lt;/code&gt;), shared state variables, and one or more &lt;strong&gt;Condition Variables&lt;/strong&gt; (&lt;code&gt;pthread_cond_t&lt;/code&gt;). Unlike a semaphore, a condition variable holds no counter; it is a kernel wait-queue that lets threads sleep until an arbitrary predicate protected by the mutex becomes true.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;div class="chroma"&gt;
&lt;table class="lntable"&gt;&lt;tr&gt;&lt;td class="lntd"&gt;
&lt;pre tabindex="0" class="chroma"&gt;&lt;code&gt;&lt;span class="lnt"&gt; 1
&lt;/span&gt;&lt;span class="lnt"&gt; 2
&lt;/span&gt;&lt;span class="lnt"&gt; 3
&lt;/span&gt;&lt;span class="lnt"&gt; 4
&lt;/span&gt;&lt;span class="lnt"&gt; 5
&lt;/span&gt;&lt;span class="lnt"&gt; 6
&lt;/span&gt;&lt;span class="lnt"&gt; 7
&lt;/span&gt;&lt;span class="lnt"&gt; 8
&lt;/span&gt;&lt;span class="lnt"&gt; 9
&lt;/span&gt;&lt;span class="lnt"&gt;10
&lt;/span&gt;&lt;span class="lnt"&gt;11
&lt;/span&gt;&lt;span class="lnt"&gt;12
&lt;/span&gt;&lt;span class="lnt"&gt;13
&lt;/span&gt;&lt;span class="lnt"&gt;14
&lt;/span&gt;&lt;span class="lnt"&gt;15
&lt;/span&gt;&lt;span class="lnt"&gt;16
&lt;/span&gt;&lt;span class="lnt"&gt;17
&lt;/span&gt;&lt;span class="lnt"&gt;18
&lt;/span&gt;&lt;span class="lnt"&gt;19
&lt;/span&gt;&lt;span class="lnt"&gt;20
&lt;/span&gt;&lt;span class="lnt"&gt;21
&lt;/span&gt;&lt;span class="lnt"&gt;22
&lt;/span&gt;&lt;span class="lnt"&gt;23
&lt;/span&gt;&lt;span class="lnt"&gt;24
&lt;/span&gt;&lt;span class="lnt"&gt;25
&lt;/span&gt;&lt;span class="lnt"&gt;26
&lt;/span&gt;&lt;span class="lnt"&gt;27
&lt;/span&gt;&lt;span class="lnt"&gt;28
&lt;/span&gt;&lt;span class="lnt"&gt;29
&lt;/span&gt;&lt;span class="lnt"&gt;30
&lt;/span&gt;&lt;span class="lnt"&gt;31
&lt;/span&gt;&lt;span class="lnt"&gt;32
&lt;/span&gt;&lt;span class="lnt"&gt;33
&lt;/span&gt;&lt;span class="lnt"&gt;34
&lt;/span&gt;&lt;span class="lnt"&gt;35
&lt;/span&gt;&lt;span class="lnt"&gt;36
&lt;/span&gt;&lt;span class="lnt"&gt;37
&lt;/span&gt;&lt;span class="lnt"&gt;38
&lt;/span&gt;&lt;span class="lnt"&gt;39
&lt;/span&gt;&lt;span class="lnt"&gt;40
&lt;/span&gt;&lt;span class="lnt"&gt;41
&lt;/span&gt;&lt;span class="lnt"&gt;42
&lt;/span&gt;&lt;span class="lnt"&gt;43
&lt;/span&gt;&lt;span class="lnt"&gt;44
&lt;/span&gt;&lt;span class="lnt"&gt;45
&lt;/span&gt;&lt;span class="lnt"&gt;46
&lt;/span&gt;&lt;span class="lnt"&gt;47
&lt;/span&gt;&lt;span class="lnt"&gt;48
&lt;/span&gt;&lt;span class="lnt"&gt;49
&lt;/span&gt;&lt;span class="lnt"&gt;50
&lt;/span&gt;&lt;span class="lnt"&gt;51
&lt;/span&gt;&lt;span class="lnt"&gt;52
&lt;/span&gt;&lt;span class="lnt"&gt;53
&lt;/span&gt;&lt;span class="lnt"&gt;54
&lt;/span&gt;&lt;span class="lnt"&gt;55
&lt;/span&gt;&lt;span class="lnt"&gt;56
&lt;/span&gt;&lt;span class="lnt"&gt;57
&lt;/span&gt;&lt;span class="lnt"&gt;58
&lt;/span&gt;&lt;span class="lnt"&gt;59
&lt;/span&gt;&lt;span class="lnt"&gt;60
&lt;/span&gt;&lt;span class="lnt"&gt;61
&lt;/span&gt;&lt;span class="lnt"&gt;62
&lt;/span&gt;&lt;span class="lnt"&gt;63
&lt;/span&gt;&lt;span class="lnt"&gt;64
&lt;/span&gt;&lt;span class="lnt"&gt;65
&lt;/span&gt;&lt;span class="lnt"&gt;66
&lt;/span&gt;&lt;span class="lnt"&gt;67
&lt;/span&gt;&lt;span class="lnt"&gt;68
&lt;/span&gt;&lt;span class="lnt"&gt;69
&lt;/span&gt;&lt;span class="lnt"&gt;70
&lt;/span&gt;&lt;span class="lnt"&gt;71
&lt;/span&gt;&lt;span class="lnt"&gt;72
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/td&gt;
&lt;td class="lntd"&gt;
&lt;pre tabindex="0" class="chroma"&gt;&lt;code class="language-c" data-lang="c"&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#include&lt;/span&gt; &lt;span class="cpf"&gt;&amp;lt;pthread.h&amp;gt;&lt;/span&gt;&lt;span class="cp"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#include&lt;/span&gt; &lt;span class="cpf"&gt;&amp;lt;stdbool.h&amp;gt;&lt;/span&gt;&lt;span class="cp"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#include&lt;/span&gt; &lt;span class="cpf"&gt;&amp;lt;stddef.h&amp;gt;&lt;/span&gt;&lt;span class="cp"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#include&lt;/span&gt; &lt;span class="cpf"&gt;&amp;lt;stdint.h&amp;gt;&lt;/span&gt;&lt;span class="cp"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#define QUEUE_CAPACITY 64
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="k"&gt;typedef&lt;/span&gt; &lt;span class="k"&gt;struct&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="kt"&gt;int32_t&lt;/span&gt; &lt;span class="n"&gt;slots&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;QUEUE_CAPACITY&lt;/span&gt;&lt;span class="p"&gt;];&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="kt"&gt;size_t&lt;/span&gt; &lt;span class="n"&gt;write_idx&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="kt"&gt;size_t&lt;/span&gt; &lt;span class="n"&gt;read_idx&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="kt"&gt;size_t&lt;/span&gt; &lt;span class="n"&gt;count&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="kt"&gt;bool&lt;/span&gt; &lt;span class="n"&gt;closed&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="kt"&gt;pthread_mutex_t&lt;/span&gt; &lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="kt"&gt;pthread_cond_t&lt;/span&gt; &lt;span class="n"&gt;can_push&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="kt"&gt;pthread_cond_t&lt;/span&gt; &lt;span class="n"&gt;can_pop&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="n"&gt;MpmcQueue&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="kt"&gt;void&lt;/span&gt; &lt;span class="nf"&gt;mpmc_init&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;MpmcQueue&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;MpmcQueue&lt;/span&gt;&lt;span class="p"&gt;){&lt;/span&gt;&lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;};&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_init&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;nullptr&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_cond_init&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;can_push&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;nullptr&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_cond_init&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;can_pop&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;nullptr&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="kt"&gt;bool&lt;/span&gt; &lt;span class="nf"&gt;mpmc_push&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;MpmcQueue&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="kt"&gt;int32_t&lt;/span&gt; &lt;span class="n"&gt;item&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_lock&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;while&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;count&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="n"&gt;QUEUE_CAPACITY&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;closed&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_cond_wait&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;can_push&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;closed&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_unlock&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;false&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;slots&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;write_idx&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;item&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;write_idx&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;write_idx&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;%&lt;/span&gt; &lt;span class="n"&gt;QUEUE_CAPACITY&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;count&lt;/span&gt;&lt;span class="o"&gt;++&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_cond_signal&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;can_pop&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_unlock&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="kt"&gt;bool&lt;/span&gt; &lt;span class="nf"&gt;mpmc_pop&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;MpmcQueue&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="kt"&gt;int32_t&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;out&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_lock&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;while&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;count&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="mi"&gt;0&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;closed&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_cond_wait&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;can_pop&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;count&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="c1"&gt;// Queue is both empty and closed
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_unlock&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;false&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;out&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;slots&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;read_idx&lt;/span&gt;&lt;span class="p"&gt;];&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;read_idx&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;read_idx&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;%&lt;/span&gt; &lt;span class="n"&gt;QUEUE_CAPACITY&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;count&lt;/span&gt;&lt;span class="o"&gt;--&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_cond_signal&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;can_push&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_unlock&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="kt"&gt;void&lt;/span&gt; &lt;span class="nf"&gt;mpmc_close&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;MpmcQueue&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_lock&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;closed&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_cond_broadcast&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;can_push&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_cond_broadcast&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;can_pop&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;pthread_mutex_unlock&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;q&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;lock&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/td&gt;&lt;/tr&gt;&lt;/table&gt;
&lt;/div&gt;
&lt;/div&gt;&lt;h3 id="why-the-while-loop-around-pthread_cond_wait-is-mandatory"&gt;Why the &lt;code&gt;while&lt;/code&gt; Loop Around &lt;code&gt;pthread_cond_wait&lt;/code&gt; Is Mandatory
&lt;/h3&gt;&lt;p&gt;Calling &lt;code&gt;pthread_cond_wait(&amp;amp;q-&amp;gt;can_pop, &amp;amp;q-&amp;gt;lock)&lt;/code&gt; atomically unlocks &lt;code&gt;q-&amp;gt;lock&lt;/code&gt; and puts the calling thread to sleep; when signaled, it re-acquires &lt;code&gt;q-&amp;gt;lock&lt;/code&gt; before returning. Re-testing the predicate in a &lt;code&gt;while&lt;/code&gt; loop—never an &lt;code&gt;if&lt;/code&gt;—is required for two reasons:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Mesa Semantics (Intercepted Wakeups)&lt;/strong&gt;: Between the instant a producer calls &lt;code&gt;pthread_cond_signal(&amp;amp;q-&amp;gt;can_pop)&lt;/code&gt; and the instant sleeping Consumer A wakes up and re-locks &lt;code&gt;q-&amp;gt;lock&lt;/code&gt;, a freshly scheduled Consumer B can acquire &lt;code&gt;q-&amp;gt;lock&lt;/code&gt; first and steal the item. When Consumer A returns from &lt;code&gt;pthread_cond_wait&lt;/code&gt;, &lt;code&gt;q-&amp;gt;count&lt;/code&gt; is &lt;code&gt;0&lt;/code&gt; again.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Spurious Wakeups&lt;/strong&gt;: On Linux (&lt;code&gt;futex&lt;/code&gt;) and other POSIX kernels, hardware interrupts or signal delivery can wake a waiting thread even when no thread called &lt;code&gt;signal&lt;/code&gt; or &lt;code&gt;broadcast&lt;/code&gt;.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Notice how cleanly &lt;code&gt;mpmc_close()&lt;/code&gt; shuts down an entire thread pool: setting &lt;code&gt;q-&amp;gt;closed = true&lt;/code&gt; and broadcasting wakes all blocked producers and consumers simultaneously, allowing consumers to drain any remaining items (&lt;code&gt;q-&amp;gt;count &amp;gt; 0&lt;/code&gt;) before exiting.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="2-wait-free-spsc-ring-buffer-with-c11-atomics"&gt;2. Wait-Free SPSC Ring Buffer with C11 Atomics
&lt;/h2&gt;&lt;p&gt;Now consider the &lt;strong&gt;Single-Producer / Single-Consumer (SPSC)&lt;/strong&gt; case. Look closely at data ownership:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Only the &lt;strong&gt;producer&lt;/strong&gt; ever modifies &lt;code&gt;write_idx&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Only the &lt;strong&gt;consumer&lt;/strong&gt; ever modifies &lt;code&gt;read_idx&lt;/code&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;If we drop the shared &lt;code&gt;count&lt;/code&gt; variable—which forced both threads to mutate the same integer—&lt;strong&gt;no variable is written by more than one thread&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;Instead of wrapping indices early, let &lt;code&gt;write_idx&lt;/code&gt; and &lt;code&gt;read_idx&lt;/code&gt; be monotonically increasing &lt;code&gt;uint64_t&lt;/code&gt; counters, and require &lt;code&gt;RING_CAPACITY&lt;/code&gt; to be a power of two ($2^k$). We index the array via bitwise masking (&lt;code&gt;idx &amp;amp; (RING_CAPACITY - 1)&lt;/code&gt;), and the exact number of items in flight is simply &lt;code&gt;write_idx - read_idx&lt;/code&gt;:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Empty&lt;/strong&gt;: &lt;code&gt;write_idx == read_idx&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Full&lt;/strong&gt;: &lt;code&gt;write_idx - read_idx == RING_CAPACITY&lt;/code&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;(Unlike the classic modulo approach, monotonic counters use all $2^k$ slots without wasting an element to distinguish full from empty).&lt;/em&gt;&lt;/p&gt;
&lt;h3 id="the-hardware-reordering-hazard-and-acquire-release-pairing"&gt;The Hardware Reordering Hazard and Acquire-Release Pairing
&lt;/h3&gt;&lt;p&gt;Even with single-writer ownership, plain C variables break on multi-core CPUs. Suppose the producer executes:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;div class="chroma"&gt;
&lt;table class="lntable"&gt;&lt;tr&gt;&lt;td class="lntd"&gt;
&lt;pre tabindex="0" class="chroma"&gt;&lt;code&gt;&lt;span class="lnt"&gt;1
&lt;/span&gt;&lt;span class="lnt"&gt;2
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/td&gt;
&lt;td class="lntd"&gt;
&lt;pre tabindex="0" class="chroma"&gt;&lt;code class="language-c" data-lang="c"&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="n"&gt;ring&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;slots&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&lt;/span&gt; &lt;span class="n"&gt;MASK&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;item&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// (1) Write payload to slot
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="n"&gt;ring&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;write_idx&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// (2) Publish new write_idx
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/td&gt;&lt;/tr&gt;&lt;/table&gt;
&lt;/div&gt;
&lt;/div&gt;&lt;p&gt;Without memory ordering constraints, both the compiler and weakly ordered CPUs (such as ARM64 / Apple Silicon) can reorder &lt;code&gt;(2)&lt;/code&gt; ahead of &lt;code&gt;(1)&lt;/code&gt; as seen by another core. The consumer observes the incremented &lt;code&gt;write_idx&lt;/code&gt;, reads &lt;code&gt;ring-&amp;gt;slots[r &amp;amp; MASK]&lt;/code&gt;, and gets stale, uninitialized memory!&lt;/p&gt;
&lt;p&gt;C11 &lt;code&gt;&amp;lt;stdatomic.h&amp;gt;&lt;/code&gt; solves this without expensive full memory barriers (&lt;code&gt;memory_order_seq_cst&lt;/code&gt;) by pairing &lt;strong&gt;&lt;code&gt;memory_order_release&lt;/code&gt;&lt;/strong&gt; stores with &lt;strong&gt;&lt;code&gt;memory_order_acquire&lt;/code&gt;&lt;/strong&gt; loads:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;A &lt;strong&gt;Store-Release&lt;/strong&gt; on &lt;code&gt;write_idx&lt;/code&gt; guarantees that the payload write to &lt;code&gt;slots[w &amp;amp; MASK]&lt;/code&gt; is globally visible &lt;em&gt;before&lt;/em&gt; the updated &lt;code&gt;write_idx&lt;/code&gt; becomes visible.&lt;/li&gt;
&lt;li&gt;A &lt;strong&gt;Load-Acquire&lt;/strong&gt; on &lt;code&gt;write_idx&lt;/code&gt; by the consumer guarantees that once it sees the new &lt;code&gt;write_idx&lt;/code&gt;, its subsequent read of &lt;code&gt;slots[r &amp;amp; MASK]&lt;/code&gt; will observe the written payload.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="cache-aligned-c11-implementation"&gt;Cache-Aligned C11 Implementation
&lt;/h3&gt;&lt;div class="highlight"&gt;&lt;div class="chroma"&gt;
&lt;table class="lntable"&gt;&lt;tr&gt;&lt;td class="lntd"&gt;
&lt;pre tabindex="0" class="chroma"&gt;&lt;code&gt;&lt;span class="lnt"&gt; 1
&lt;/span&gt;&lt;span class="lnt"&gt; 2
&lt;/span&gt;&lt;span class="lnt"&gt; 3
&lt;/span&gt;&lt;span class="lnt"&gt; 4
&lt;/span&gt;&lt;span class="lnt"&gt; 5
&lt;/span&gt;&lt;span class="lnt"&gt; 6
&lt;/span&gt;&lt;span class="lnt"&gt; 7
&lt;/span&gt;&lt;span class="lnt"&gt; 8
&lt;/span&gt;&lt;span class="lnt"&gt; 9
&lt;/span&gt;&lt;span class="lnt"&gt;10
&lt;/span&gt;&lt;span class="lnt"&gt;11
&lt;/span&gt;&lt;span class="lnt"&gt;12
&lt;/span&gt;&lt;span class="lnt"&gt;13
&lt;/span&gt;&lt;span class="lnt"&gt;14
&lt;/span&gt;&lt;span class="lnt"&gt;15
&lt;/span&gt;&lt;span class="lnt"&gt;16
&lt;/span&gt;&lt;span class="lnt"&gt;17
&lt;/span&gt;&lt;span class="lnt"&gt;18
&lt;/span&gt;&lt;span class="lnt"&gt;19
&lt;/span&gt;&lt;span class="lnt"&gt;20
&lt;/span&gt;&lt;span class="lnt"&gt;21
&lt;/span&gt;&lt;span class="lnt"&gt;22
&lt;/span&gt;&lt;span class="lnt"&gt;23
&lt;/span&gt;&lt;span class="lnt"&gt;24
&lt;/span&gt;&lt;span class="lnt"&gt;25
&lt;/span&gt;&lt;span class="lnt"&gt;26
&lt;/span&gt;&lt;span class="lnt"&gt;27
&lt;/span&gt;&lt;span class="lnt"&gt;28
&lt;/span&gt;&lt;span class="lnt"&gt;29
&lt;/span&gt;&lt;span class="lnt"&gt;30
&lt;/span&gt;&lt;span class="lnt"&gt;31
&lt;/span&gt;&lt;span class="lnt"&gt;32
&lt;/span&gt;&lt;span class="lnt"&gt;33
&lt;/span&gt;&lt;span class="lnt"&gt;34
&lt;/span&gt;&lt;span class="lnt"&gt;35
&lt;/span&gt;&lt;span class="lnt"&gt;36
&lt;/span&gt;&lt;span class="lnt"&gt;37
&lt;/span&gt;&lt;span class="lnt"&gt;38
&lt;/span&gt;&lt;span class="lnt"&gt;39
&lt;/span&gt;&lt;span class="lnt"&gt;40
&lt;/span&gt;&lt;span class="lnt"&gt;41
&lt;/span&gt;&lt;span class="lnt"&gt;42
&lt;/span&gt;&lt;span class="lnt"&gt;43
&lt;/span&gt;&lt;span class="lnt"&gt;44
&lt;/span&gt;&lt;span class="lnt"&gt;45
&lt;/span&gt;&lt;span class="lnt"&gt;46
&lt;/span&gt;&lt;span class="lnt"&gt;47
&lt;/span&gt;&lt;span class="lnt"&gt;48
&lt;/span&gt;&lt;span class="lnt"&gt;49
&lt;/span&gt;&lt;span class="lnt"&gt;50
&lt;/span&gt;&lt;span class="lnt"&gt;51
&lt;/span&gt;&lt;span class="lnt"&gt;52
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/td&gt;
&lt;td class="lntd"&gt;
&lt;pre tabindex="0" class="chroma"&gt;&lt;code class="language-c" data-lang="c"&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#include&lt;/span&gt; &lt;span class="cpf"&gt;&amp;lt;stdatomic.h&amp;gt;&lt;/span&gt;&lt;span class="cp"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#include&lt;/span&gt; &lt;span class="cpf"&gt;&amp;lt;stdbool.h&amp;gt;&lt;/span&gt;&lt;span class="cp"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#include&lt;/span&gt; &lt;span class="cpf"&gt;&amp;lt;stddef.h&amp;gt;&lt;/span&gt;&lt;span class="cp"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#include&lt;/span&gt; &lt;span class="cpf"&gt;&amp;lt;stdint.h&amp;gt;&lt;/span&gt;&lt;span class="cp"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#define SPSC_CAPACITY 1024U
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="cp"&gt;#define SPSC_MASK (SPSC_CAPACITY - 1U)
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="nf"&gt;_Static_assert&lt;/span&gt;&lt;span class="p"&gt;((&lt;/span&gt;&lt;span class="n"&gt;SPSC_CAPACITY&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&lt;/span&gt; &lt;span class="n"&gt;SPSC_MASK&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s"&gt;&amp;#34;SPSC_CAPACITY must be a power of 2&amp;#34;&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="k"&gt;typedef&lt;/span&gt; &lt;span class="k"&gt;struct&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="c1"&gt;// Isolate producer-owned, consumer-owned, and read-mostly data on 64-byte cache lines
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;_Alignas&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;64&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="n"&gt;_Atomic&lt;/span&gt; &lt;span class="kt"&gt;uint64_t&lt;/span&gt; &lt;span class="n"&gt;write_idx&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;_Alignas&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;64&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="kt"&gt;uint64_t&lt;/span&gt; &lt;span class="n"&gt;cached_read_idx&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// Producer-local shadow copy
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;_Alignas&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;64&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="n"&gt;_Atomic&lt;/span&gt; &lt;span class="kt"&gt;uint64_t&lt;/span&gt; &lt;span class="n"&gt;read_idx&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;_Alignas&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;64&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="kt"&gt;uint64_t&lt;/span&gt; &lt;span class="n"&gt;cached_write_idx&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// Consumer-local shadow copy
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;_Alignas&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;64&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="kt"&gt;int32_t&lt;/span&gt; &lt;span class="n"&gt;slots&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;SPSC_CAPACITY&lt;/span&gt;&lt;span class="p"&gt;];&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="n"&gt;SpscRing&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="k"&gt;static&lt;/span&gt; &lt;span class="kr"&gt;inline&lt;/span&gt; &lt;span class="kt"&gt;bool&lt;/span&gt; &lt;span class="nf"&gt;spsc_push&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;SpscRing&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="kt"&gt;int32_t&lt;/span&gt; &lt;span class="n"&gt;item&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;const&lt;/span&gt; &lt;span class="kt"&gt;uint64_t&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;atomic_load_explicit&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;write_idx&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;memory_order_relaxed&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;cached_read_idx&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="n"&gt;SPSC_CAPACITY&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="c1"&gt;// Only reload the consumer&amp;#39;s atomic read_idx when our cached view looks full
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;cached_read_idx&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;atomic_load_explicit&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;read_idx&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;memory_order_acquire&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;cached_read_idx&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="n"&gt;SPSC_CAPACITY&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;false&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// Ring is genuinely full
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;slots&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&lt;/span&gt; &lt;span class="n"&gt;SPSC_MASK&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;item&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;atomic_store_explicit&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;write_idx&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;w&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;memory_order_release&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="k"&gt;static&lt;/span&gt; &lt;span class="kr"&gt;inline&lt;/span&gt; &lt;span class="kt"&gt;bool&lt;/span&gt; &lt;span class="nf"&gt;spsc_pop&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;SpscRing&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="kt"&gt;int32_t&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;out&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;const&lt;/span&gt; &lt;span class="kt"&gt;uint64_t&lt;/span&gt; &lt;span class="n"&gt;rd&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;atomic_load_explicit&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;read_idx&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;memory_order_relaxed&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;rd&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;cached_write_idx&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="c1"&gt;// Only reload the producer&amp;#39;s atomic write_idx when our cached view looks empty
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;cached_write_idx&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;atomic_load_explicit&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;write_idx&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;memory_order_acquire&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;rd&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;cached_write_idx&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;false&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// Ring is genuinely empty
&lt;/span&gt;&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="o"&gt;*&lt;/span&gt;&lt;span class="n"&gt;out&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;slots&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;rd&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&lt;/span&gt; &lt;span class="n"&gt;SPSC_MASK&lt;/span&gt;&lt;span class="p"&gt;];&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="nf"&gt;atomic_store_explicit&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;&amp;amp;&lt;/span&gt;&lt;span class="n"&gt;r&lt;/span&gt;&lt;span class="o"&gt;-&amp;gt;&lt;/span&gt;&lt;span class="n"&gt;read_idx&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;rd&lt;/span&gt; &lt;span class="o"&gt;+&lt;/span&gt; &lt;span class="mi"&gt;1&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;memory_order_release&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nb"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span class="line"&gt;&lt;span class="cl"&gt;&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/td&gt;&lt;/tr&gt;&lt;/table&gt;
&lt;/div&gt;
&lt;/div&gt;&lt;h3 id="two-microarchitectural-optimizations"&gt;Two Microarchitectural Optimizations
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Eliminating False Sharing (&lt;code&gt;_Alignas(64)&lt;/code&gt;)&lt;/strong&gt;: CPU caches coherence protocols (MESI/MOESI) operate on 64-byte &lt;strong&gt;cache lines&lt;/strong&gt;. If &lt;code&gt;write_idx&lt;/code&gt; and &lt;code&gt;read_idx&lt;/code&gt; shared the same 64-byte line, every store by the producer would invalidate the consumer&amp;rsquo;s L1 cache line and vice versa. Aligning each atomic index to 64 bytes eliminates cross-core cache-line ping-ponging.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Shadow Index Caching (&lt;code&gt;cached_read_idx&lt;/code&gt; / &lt;code&gt;cached_write_idx&lt;/code&gt;)&lt;/strong&gt;: Even with separate cache lines, reading the remote core&amp;rsquo;s atomic index on &lt;em&gt;every&lt;/em&gt; push or pop triggers cross-core cache traffic whenever the remote core updates its counter. By caching the last observed &lt;code&gt;read_idx&lt;/code&gt; locally on the producer&amp;rsquo;s cache line, the producer only performs an &lt;code&gt;acquire&lt;/code&gt; load across cores when its local bound &lt;code&gt;w - cached_read_idx == SPSC_CAPACITY&lt;/code&gt; is reached—reducing cross-core coherence traffic by up to $1024\times$.&lt;/li&gt;
&lt;/ol&gt;</description></item></channel></rss>