<!DOCTYPE HTML>
<html lang="en" class="sidebar-visible no-js">
<head>
<!-- Book generated using mdBook -->
<meta charset="UTF-8">
<title>SIMD for faster computing - The Edition Guide</title>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type">
<meta name="description" content="">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="theme-color" content="#ffffff" />
<link rel="shortcut icon" href="../favicon.png">
<link rel="stylesheet" href="../css/variables.css">
<link rel="stylesheet" href="../css/general.css">
<link rel="stylesheet" href="../css/chrome.css">
<link rel="stylesheet" href="../css/print.css" media="print">
<!-- Fonts -->
<link rel="stylesheet" href="../FontAwesome/css/font-awesome.css">
<link href="https://fonts.googleapis.com/css?family=Open+Sans:300italic,400italic,600italic,700italic,800italic,400,300,600,700,800" rel="stylesheet" type="text/css">
<link href="https://fonts.googleapis.com/css?family=Source+Code+Pro:500" rel="stylesheet" type="text/css">
<!-- Highlight.js Stylesheets -->
<link rel="stylesheet" href="../highlight.css">
<link rel="stylesheet" href="../tomorrow-night.css">
<link rel="stylesheet" href="../ayu-highlight.css">
<!-- Custom theme stylesheets -->
</head>
<body class="light">
<!-- Provide site root to javascript -->
<script type="text/javascript">
var path_to_root = "../";
var default_theme = "light";
</script>
<!-- Work around some values being stored in localStorage wrapped in quotes -->
<script type="text/javascript">
try {
var theme = localStorage.getItem('mdbook-theme');
var sidebar = localStorage.getItem('mdbook-sidebar');
if (theme.startsWith('"') && theme.endsWith('"')) {
localStorage.setItem('mdbook-theme', theme.slice(1, theme.length - 1));
}
if (sidebar.startsWith('"') && sidebar.endsWith('"')) {
localStorage.setItem('mdbook-sidebar', sidebar.slice(1, sidebar.length - 1));
}
} catch (e) { }
</script>
<!-- Set the theme before any content is loaded, prevents flash -->
<script type="text/javascript">
var theme;
try { theme = localStorage.getItem('mdbook-theme'); } catch(e) { }
if (theme === null || theme === undefined) { theme = default_theme; }
document.body.className = theme;
document.querySelector('html').className = theme + ' js';
</script>
<!-- Hide / unhide sidebar before it is displayed -->
<script type="text/javascript">
var html = document.querySelector('html');
var sidebar = 'hidden';
if (document.body.clientWidth >= 1080) {
try { sidebar = localStorage.getItem('mdbook-sidebar'); } catch(e) { }
sidebar = sidebar || 'visible';
}
html.classList.remove('sidebar-visible');
html.classList.add("sidebar-" + sidebar);
</script>
<nav id="sidebar" class="sidebar" aria-label="Table of contents">
<ol class="chapter"><li class="affix"><a href="../introduction.html">Introduction</a></li><li><a href="../editions/index.html"><strong aria-hidden="true">1.</strong> What are editions?</a></li><li><ol class="section"><li><a href="../editions/creating-a-new-project.html"><strong aria-hidden="true">1.1.</strong> Creating a new project</a></li><li><a href="../editions/transitioning-an-existing-project-to-a-new-edition.html"><strong aria-hidden="true">1.2.</strong> Transitioning an existing project to a new edition</a></li></ol></li><li><a href="../rust-2015/index.html"><strong aria-hidden="true">2.</strong> Rust 2015</a></li><li><a href="../rust-2018/index.html"><strong aria-hidden="true">3.</strong> Rust 2018</a></li><li><ol class="section"><li><a href="../rust-2018/edition-changes.html"><strong aria-hidden="true">3.1.</strong> 2018-Specific Changes</a></li><li><a href="../rust-2018/module-system/index.html"><strong aria-hidden="true">3.2.</strong> Module system</a></li><li><ol class="section"><li><a href="../rust-2018/module-system/raw-identifiers.html"><strong aria-hidden="true">3.2.1.</strong> Raw identifiers</a></li><li><a href="../rust-2018/module-system/path-clarity.html"><strong aria-hidden="true">3.2.2.</strong> Path clarity</a></li><li><a href="../rust-2018/module-system/more-visibility-modifiers.html"><strong aria-hidden="true">3.2.3.</strong> More visibility modifiers</a></li><li><a href="../rust-2018/module-system/nested-imports-with-use.html"><strong aria-hidden="true">3.2.4.</strong> Nested imports with use</a></li></ol></li><li><a href="../rust-2018/error-handling-and-panics/index.html"><strong aria-hidden="true">3.3.</strong> Error handling and panics</a></li><li><ol class="section"><li><a href="../rust-2018/error-handling-and-panics/the-question-mark-operator-for-easier-error-handling.html"><strong aria-hidden="true">3.3.1.</strong> The ? operator for easier error handling</a></li><li><a href="../rust-2018/error-handling-and-panics/question-mark-in-main-and-tests.html"><strong aria-hidden="true">3.3.2.</strong> ? in main and tests</a></li><li><a href="../rust-2018/error-handling-and-panics/controlling-panics-with-std-panic.html"><strong aria-hidden="true">3.3.3.</strong> Controlling panics with std::panic</a></li><li><a href="../rust-2018/error-handling-and-panics/aborting-on-panic.html"><strong aria-hidden="true">3.3.4.</strong> Aborting on panic</a></li></ol></li><li><a href="../rust-2018/control-flow/index.html"><strong aria-hidden="true">3.4.</strong> Control flow</a></li><li><ol class="section"><li><a href="../rust-2018/control-flow/loops-can-break-with-a-value.html"><strong aria-hidden="true">3.4.1.</strong> Loops can break with a value</a></li><li><a href="../rust-2018/control-flow/async-await-for-easier-concurrency.html"><strong aria-hidden="true">3.4.2.</strong> async/await for easier concurrency</a></li></ol></li><li><a href="../rust-2018/trait-system/index.html"><strong aria-hidden="true">3.5.</strong> Trait system</a></li><li><ol class="section"><li><a href="../rust-2018/trait-system/impl-trait-for-returning-complex-types-with-ease.html"><strong aria-hidden="true">3.5.1.</strong> impl Trait for returning complex types with ease</a></li><li><a href="../rust-2018/trait-system/dyn-trait-for-trait-objects.html"><strong aria-hidden="true">3.5.2.</strong> dyn Trait for trait objects</a></li><li><a href="../rust-2018/trait-system/more-container-types-support-trait-objects.html"><strong aria-hidden="true">3.5.3.</strong> More container types support trait objects</a></li><li><a href="../rust-2018/trait-system/associated-constants.html"><strong aria-hidden="true">3.5.4.</strong> Associated constants</a></li><li><a href="../rust-2018/trait-system/no-anon-params.html"><strong aria-hidden="true">3.5.5.</strong> No more anonymous parameters</a></li></ol></li><li><a href="../rust-2018/slice-patterns.html"><strong aria-hidden="true">3.6.</strong> Slice patterns</a></li><li><a href="../rust-2018/ownership-and-lifetimes/index.html"><strong aria-hidden="true">3.7.</strong> Ownership and lifetimes</a></li><li><ol class="section"><li><a href="../rust-2018/ownership-and-lifetimes/non-lexical-lifetimes.html"><strong aria-hidden="true">3.7.1.</strong> Non-lexical lifetimes</a></li><li><a href="../rust-2018/ownership-and-lifetimes/default-match-bindings.html"><strong aria-hidden="true">3.7.2.</strong> Default match bindings</a></li><li><a href="../rust-2018/ownership-and-lifetimes/the-anonymous-lifetime.html"><strong aria-hidden="true">3.7.3.</strong> '_, the anonymous lifetime</a></li><li><a href="../rust-2018/ownership-and-lifetimes/lifetime-elision-in-impl.html"><strong aria-hidden="true">3.7.4.</strong> Lifetime elision in impl</a></li><li><a href="../rust-2018/ownership-and-lifetimes/inference-in-structs.html"><strong aria-hidden="true">3.7.5.</strong> T: 'a inference in structs</a></li><li><a href="../rust-2018/ownership-and-lifetimes/simpler-lifetimes-in-static-and-const.html"><strong aria-hidden="true">3.7.6.</strong> Simpler lifetimes in static and const</a></li></ol></li><li><a href="../rust-2018/data-types/index.html"><strong aria-hidden="true">3.8.</strong> Data types</a></li><li><ol class="section"><li><a href="../rust-2018/data-types/field-init-shorthand.html"><strong aria-hidden="true">3.8.1.</strong> Field init shorthand</a></li><li><a href="../rust-2018/data-types/inclusive-ranges.html"><strong aria-hidden="true">3.8.2.</strong> ..= for inclusive ranges</a></li><li><a href="../rust-2018/data-types/128-bit-integers.html"><strong aria-hidden="true">3.8.3.</strong> 128 bit integers</a></li><li><a href="../rust-2018/data-types/operator-equals-are-now-implementable.html"><strong aria-hidden="true">3.8.4.</strong> "Operator-equals" are now implementable</a></li><li><a href="../rust-2018/data-types/union-for-an-unsafe-form-of-enum.html"><strong aria-hidden="true">3.8.5.</strong> union for an unsafe form of enum</a></li><li><a href="../rust-2018/data-types/choosing-alignment-with-the-repr-attribute.html"><strong aria-hidden="true">3.8.6.</strong> Choosing alignment with the repr attribute</a></li></ol></li><li><a href="../rust-2018/simd-for-faster-computing.html" class="active"><strong aria-hidden="true">3.9.</strong> SIMD for faster computing</a></li><li><a href="../rust-2018/macros/index.html"><strong aria-hidden="true">3.10.</strong> Macros</a></li><li><ol class="section"><li><a href="../rust-2018/macros/custom-derive.html"><strong aria-hidden="true">3.10.1.</strong> Custom Derive</a></li><li><a href="../rust-2018/macros/macro-changes.html"><strong aria-hidden="true">3.10.2.</strong> Macro changes</a></li><li><a href="../rust-2018/macros/at-most-once.html"><strong aria-hidden="true">3.10.3.</strong> At most one repetition</a></li></ol></li><li><a href="../rust-2018/the-compiler/index.html"><strong aria-hidden="true">3.11.</strong> The compiler</a></li><li><ol class="section"><li><a href="../rust-2018/the-compiler/improved-error-messages.html"><strong aria-hidden="true">3.11.1.</strong> Improved error messages</a></li><li><a href="../rust-2018/the-compiler/incremental-compilation-for-faster-compiles.html"><strong aria-hidden="true">3.11.2.</strong> Incremental Compilation for faster compiles</a></li><li><a href="../rust-2018/the-compiler/an-attribute-for-deprecation.html"><strong aria-hidden="true">3.11.3.</strong> An attribute for deprecation</a></li></ol></li><li><a href="../rust-2018/rustup-for-managing-rust-versions.html"><strong aria-hidden="true">3.12.</strong> Rustup for managing Rust versions</a></li><li><a href="../rust-2018/cargo-and-crates-io/index.html"><strong aria-hidden="true">3.13.</strong> Cargo and crates.io</a></li><li><ol class="section"><li><a href="../rust-2018/cargo-and-crates-io/cargo-check-for-faster-checking.html"><strong aria-hidden="true">3.13.1.</strong> cargo check for faster checking</a></li><li><a href="../rust-2018/cargo-and-crates-io/cargo-install-for-easy-installation-of-tools.html"><strong aria-hidden="true">3.13.2.</strong> cargo install for easy installation of tools</a></li><li><a href="../rust-2018/cargo-and-crates-io/cargo-new-defaults-to-a-binary-project.html"><strong aria-hidden="true">3.13.3.</strong> cargo new defaults to a binary project</a></li><li><a href="../rust-2018/cargo-and-crates-io/cargo-rustc-for-passing-arbitrary-flags-to-rustc.html"><strong aria-hidden="true">3.13.4.</strong> cargo rustc for passing arbitrary flags to rustc</a></li><li><a href="../rust-2018/cargo-and-crates-io/cargo-workspaces-for-multi-package-projects.html"><strong aria-hidden="true">3.13.5.</strong> Cargo workspaces for multi-package projects</a></li><li><a href="../rust-2018/cargo-and-crates-io/multi-file-examples.html"><strong aria-hidden="true">3.13.6.</strong> Multi-file examples</a></li><li><a href="../rust-2018/cargo-and-crates-io/replacing-dependencies-with-patch.html"><strong aria-hidden="true">3.13.7.</strong> Replacing dependencies with patch</a></li><li><a href="../rust-2018/cargo-and-crates-io/cargo-can-use-a-local-registry-replacement.html"><strong aria-hidden="true">3.13.8.</strong> Cargo can use a local registry replacement</a></li><li><a href="../rust-2018/cargo-and-crates-io/crates-io-disallows-wildcard-dependencies.html"><strong aria-hidden="true">3.13.9.</strong> Crates.io disallows wildcard dependencies</a></li></ol></li><li><a href="../rust-2018/documentation/index.html"><strong aria-hidden="true">3.14.</strong> Documentation</a></li><li><ol class="section"><li><a href="../rust-2018/documentation/new-editions-of-the-book.html"><strong aria-hidden="true">3.14.1.</strong> New editions of the "the book"</a></li><li><a href="../rust-2018/documentation/the-rust-bookshelf.html"><strong aria-hidden="true">3.14.2.</strong> The Rust Bookshelf</a></li><li><a href="../rust-2018/documentation/the-rustonomicon.html"><strong aria-hidden="true">3.14.3.</strong> The Rustonomicon</a></li><li><a href="../rust-2018/documentation/std-os-has-documentation-for-all-platforms.html"><strong aria-hidden="true">3.14.4.</strong> Full documentation for std::os</a></li></ol></li><li><a href="../rust-2018/rustdoc/index.html"><strong aria-hidden="true">3.15.</strong> rustdoc</a></li><li><ol class="section"><li><a href="../rust-2018/rustdoc/documentation-tests-can-now-compile-fail.html"><strong aria-hidden="true">3.15.1.</strong> Documentation tests can now compile-fail</a></li><li><a href="../rust-2018/rustdoc/rustdoc-uses-commonmark.html"><strong aria-hidden="true">3.15.2.</strong> Rustdoc uses CommonMark</a></li></ol></li><li><a href="../rust-2018/platform-and-target-support/index.html"><strong aria-hidden="true">3.16.</strong> Platform and target support</a></li><li><ol class="section"><li><a href="../rust-2018/platform-and-target-support/libcore-for-low-level-rust.html"><strong aria-hidden="true">3.16.1.</strong> libcore for low-level Rust</a></li><li><a href="../rust-2018/platform-and-target-support/webassembly-support.html"><strong aria-hidden="true">3.16.2.</strong> WebAssembly support</a></li><li><a href="../rust-2018/platform-and-target-support/global-allocators.html"><strong aria-hidden="true">3.16.3.</strong> Global allocators</a></li><li><a href="../rust-2018/platform-and-target-support/msvc-toolchain-support.html"><strong aria-hidden="true">3.16.4.</strong> MSVC toolchain support</a></li><li><a href="../rust-2018/platform-and-target-support/musl-support-for-fully-static-binaries.html"><strong aria-hidden="true">3.16.5.</strong> MUSL support for fully static binaries</a></li><li><a href="../rust-2018/platform-and-target-support/cdylib-crates-for-c-interoperability.html"><strong aria-hidden="true">3.16.6.</strong> cdylib crates for C interoperability</a></li></ol></li></ol></li></ol>
</nav>
<div id="page-wrapper" class="page-wrapper">
<div class="page">
<div id="menu-bar" class="menu-bar">
<div id="menu-bar-sticky-container">
<div class="left-buttons">
<button id="sidebar-toggle" class="icon-button" type="button" title="Toggle Table of Contents" aria-label="Toggle Table of Contents" aria-controls="sidebar">
<i class="fa fa-bars"></i>
</button>
<button id="theme-toggle" class="icon-button" type="button" title="Change theme" aria-label="Change theme" aria-haspopup="true" aria-expanded="false" aria-controls="theme-list">
<i class="fa fa-paint-brush"></i>
</button>
<ul id="theme-list" class="theme-popup" aria-label="Themes" role="menu">
<li role="none"><button role="menuitem" class="theme" id="light">Light (default)</button></li>
<li role="none"><button role="menuitem" class="theme" id="rust">Rust</button></li>
<li role="none"><button role="menuitem" class="theme" id="coal">Coal</button></li>
<li role="none"><button role="menuitem" class="theme" id="navy">Navy</button></li>
<li role="none"><button role="menuitem" class="theme" id="ayu">Ayu</button></li>
</ul>
<button id="search-toggle" class="icon-button" type="button" title="Search. (Shortkey: s)" aria-label="Toggle Searchbar" aria-expanded="false" aria-keyshortcuts="S" aria-controls="searchbar">
<i class="fa fa-search"></i>
</button>
</div>
<h1 class="menu-title">The Edition Guide</h1>
<div class="right-buttons">
<a href="../print.html" title="Print this book" aria-label="Print this book">
<i id="print-button" class="fa fa-print"></i>
</a>
</div>
</div>
</div>
<div id="search-wrapper" class="hidden">
<form id="searchbar-outer" class="searchbar-outer">
<input type="search" name="search" id="searchbar" name="searchbar" placeholder="Search this book ..." aria-controls="searchresults-outer" aria-describedby="searchresults-header">
</form>
<div id="searchresults-outer" class="searchresults-outer hidden">
<div id="searchresults-header" class="searchresults-header"></div>
<ul id="searchresults">
</ul>
</div>
</div>
<!-- Apply ARIA attributes after the sidebar and the sidebar toggle button are added to the DOM -->
<script type="text/javascript">
document.getElementById('sidebar-toggle').setAttribute('aria-expanded', sidebar === 'visible');
document.getElementById('sidebar').setAttribute('aria-hidden', sidebar !== 'visible');
Array.from(document.querySelectorAll('#sidebar a')).forEach(function(link) {
link.setAttribute('tabIndex', sidebar === 'visible' ? 0 : -1);
});
</script>
<div id="content" class="content">
<main>
<a class="header" href="#simd-for-faster-computing" id="simd-for-faster-computing"><h1>SIMD for faster computing</h1></a>
<p><img src="https://img.shields.io/badge/Minimum%20Rust%20Version-1.27-brightgreen.svg" alt="Minimum Rust version: 1.27" /></p>
<p>The basics of <a href="https://en.wikipedia.org/wiki/SIMD">SIMD</a> are now available!
SIMD stands for “single instruction, multiple data.” Consider a function like
this:</p>
<pre><pre class="playpen"><code class="language-rust">
# #![allow(unused_variables)]
#fn main() {
pub fn foo(a: &[u8], b: &[u8], c: &mut [u8]) {
for ((a, b), c) in a.iter().zip(b).zip(c) {
*c = *a + *b;
}
}
#}</code></pre></pre>
<p>Here, we’re taking two slices, and adding the numbers together, placing the
result in a third slice. The simplest possible way to do this would be to do
exactly what the code does, and loop through each set of elements, add them
together, and store it in the result. However, compilers can often do better.
LLVM will usually “autovectorize” code like this, which is a fancy term for
“use SIMD.” Imagine that <code>a</code> and <code>b</code> were both 16 elements long. Each element
is a <code>u8</code>, and so that means that each slice would be 128 bits of data. Using
SIMD, we could put both <code>a</code> and <code>b</code> into 128 bit registers, add them together
in a <em>single</em> instruction, and then copy the resulting 128 bits into <code>c</code>.
That’d be much faster!</p>
<p>While stable Rust has always been able to take advantage of
autovectorization, sometimes, the compiler just isn’t smart enough to realize
that we can do something like this. Additionally, not every CPU has these
features, and so LLVM may not use them so your program can be used on a wide
variety of hardware. The <code>std::arch</code> module allows us to use these kinds of
instructions directly, which means we don’t need to rely on a smart compiler.
Additionally, it includes some features that allow us to choose a particular
implementation based on various criteria. For example:</p>
<pre><code class="language-rust ignore">#[cfg(all(any(target_arch = "x86", target_arch = "x86_64"),
target_feature = "avx2"))]
fn foo() {
#[cfg(target_arch = "x86")]
use std::arch::x86::_mm256_add_epi64;
#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::_mm256_add_epi64;
unsafe {
_mm256_add_epi64(...);
}
}
</code></pre>
<p>Here, we use cfg flags to choose the correct version based on the machine
we’re targeting; on x86 we use that version, and on x86_64 we use its
version. We can also choose at runtime:</p>
<pre><code class="language-rust ignore">fn foo() {
#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
{
if is_x86_feature_detected!("avx2") {
return unsafe { foo_avx2() };
}
}
foo_fallback();
}
</code></pre>
<p>Here, we have two versions of the function: one which uses AVX2, a specific
kind of SIMD feature that lets you do 256-bit operations. The
<code>is_x86_feature_detected!</code> macro will generate code that detects if your CPU
supports AVX2, and if so, calls the foo_avx2 function. If not, then we fall
back to a non-AVX implementation, foo_fallback. This means that our code will
run super fast on CPUs that support AVX2, but still work on ones that don’t,
albeit slower.</p>
<p>If all of this seems a bit low-level and fiddly, well, it is! <code>std::arch</code> is
specifically primitives for building these kinds of things. We hope to
eventually stabilize a <code>std::simd</code> module with higher-level stuff in the
future. But landing the basics now lets the ecosystem experiment with higher
level libraries starting today. For example, check out the
<a href="https://github.com/AdamNiederer/faster">faster</a> crate. Here’s a code snippet
with no SIMD:</p>
<pre><code class="language-rust ignore">let lots_of_3s = (&[-123.456f32; 128][..]).iter()
.map(|v| {
9.0 * v.abs().sqrt().sqrt().recip().ceil().sqrt() - 4.0 - 2.0
})
.collect::<Vec<f32>>();
</code></pre>
<p>To use SIMD with this code via faster, you’d change it to this:</p>
<pre><code class="language-rust ignore">let lots_of_3s = (&[-123.456f32; 128][..]).simd_iter()
.simd_map(f32s(0.0), |v| {
f32s(9.0) * v.abs().sqrt().rsqrt().ceil().sqrt() - f32s(4.0) - f32s(2.0)
})
.scalar_collect();
</code></pre>
<p>It looks almost the same: <code>simd_iter</code> instead of <code>iter</code>, <code>simd_map</code> instead of <code>map</code>,
<code>f32s(2.0)</code> instead of <code>2.0</code>. But you get a SIMD-ified version generated for you.</p>
<p>Beyond that, you may never write any of this yourself, but as always, the
libraries you depend on may. For example, the regex crate contains these SIMD
speedups without you needing to do anything at all!</p>
</main>
<nav class="nav-wrapper" aria-label="Page navigation">
<!-- Mobile navigation buttons -->
<a rel="prev" href="../rust-2018/data-types/choosing-alignment-with-the-repr-attribute.html" class="mobile-nav-chapters previous" title="Previous chapter" aria-label="Previous chapter" aria-keyshortcuts="Left">
<i class="fa fa-angle-left"></i>
</a>
<a rel="next" href="../rust-2018/macros/index.html" class="mobile-nav-chapters next" title="Next chapter" aria-label="Next chapter" aria-keyshortcuts="Right">
<i class="fa fa-angle-right"></i>
</a>
<div style="clear: both"></div>
</nav>
</div>
</div>
<nav class="nav-wide-wrapper" aria-label="Page navigation">
<a href="../rust-2018/data-types/choosing-alignment-with-the-repr-attribute.html" class="nav-chapters previous" title="Previous chapter" aria-label="Previous chapter" aria-keyshortcuts="Left">
<i class="fa fa-angle-left"></i>
</a>
<a href="../rust-2018/macros/index.html" class="nav-chapters next" title="Next chapter" aria-label="Next chapter" aria-keyshortcuts="Right">
<i class="fa fa-angle-right"></i>
</a>
</nav>
</div>
<script src="../elasticlunr.min.js" type="text/javascript" charset="utf-8"></script>
<script src="../mark.min.js" type="text/javascript" charset="utf-8"></script>
<script src="../searcher.js" type="text/javascript" charset="utf-8"></script>
<script src="../clipboard.min.js" type="text/javascript" charset="utf-8"></script>
<script src="../highlight.js" type="text/javascript" charset="utf-8"></script>
<script src="../book.js" type="text/javascript" charset="utf-8"></script>
<!-- Custom JS scripts -->
</body>
</html>
distrib
>
Mageia
>
7
>
armv7hl
>
media
>
core-release
>
by-pkgid
>
0c2243f8a1696816431e7210e991fa52
>
files
>
12688
