[www-releases] r257950 - Add 3.7.1 docs
Tom Stellard via llvm-commits
llvm-commits at lists.llvm.org
Fri Jan 15 15:13:20 PST 2016
Added: www-releases/trunk/3.7.1/docs/LangRef.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/3.7.1/docs/LangRef.html?rev=257950&view=auto
==============================================================================
--- www-releases/trunk/3.7.1/docs/LangRef.html (added)
+++ www-releases/trunk/3.7.1/docs/LangRef.html Fri Jan 15 17:13:16 2016
@@ -0,0 +1,11398 @@
+
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
+ "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+
+
+<html xmlns="http://www.w3.org/1999/xhtml">
+ <head>
+ <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+
+ <title>LLVM Language Reference Manual — LLVM 3.7 documentation</title>
+
+ <link rel="stylesheet" href="_static/llvm-theme.css" type="text/css" />
+ <link rel="stylesheet" href="_static/pygments.css" type="text/css" />
+
+ <script type="text/javascript">
+ var DOCUMENTATION_OPTIONS = {
+ URL_ROOT: './',
+ VERSION: '3.7',
+ COLLAPSE_INDEX: false,
+ FILE_SUFFIX: '.html',
+ HAS_SOURCE: true
+ };
+ </script>
+ <script type="text/javascript" src="_static/jquery.js"></script>
+ <script type="text/javascript" src="_static/underscore.js"></script>
+ <script type="text/javascript" src="_static/doctools.js"></script>
+ <link rel="top" title="LLVM 3.7 documentation" href="index.html" />
+ <link rel="next" title="Building LLVM with CMake" href="CMake.html" />
+ <link rel="prev" title="Overview" href="index.html" />
+<style type="text/css">
+ table.right { float: right; margin-left: 20px; }
+ table.right td { border: 1px solid #ccc; }
+</style>
+
+ </head>
+ <body>
+<div class="logo">
+ <a href="index.html">
+ <img src="_static/logo.png"
+ alt="LLVM Logo" width="250" height="88"/></a>
+</div>
+
+ <div class="related">
+ <h3>Navigation</h3>
+ <ul>
+ <li class="right" style="margin-right: 10px">
+ <a href="genindex.html" title="General Index"
+ accesskey="I">index</a></li>
+ <li class="right" >
+ <a href="CMake.html" title="Building LLVM with CMake"
+ accesskey="N">next</a> |</li>
+ <li class="right" >
+ <a href="index.html" title="Overview"
+ accesskey="P">previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="index.html">Documentation</a>»</li>
+
+ </ul>
+ </div>
+
+
+ <div class="document">
+ <div class="documentwrapper">
+ <div class="body">
+
+ <div class="section" id="llvm-language-reference-manual">
+<h1>LLVM Language Reference Manual<a class="headerlink" href="#llvm-language-reference-manual" title="Permalink to this headline">¶</a></h1>
+<div class="contents local topic" id="contents">
+<ul class="simple">
+<li><a class="reference internal" href="#abstract" id="id607">Abstract</a></li>
+<li><a class="reference internal" href="#introduction" id="id608">Introduction</a><ul>
+<li><a class="reference internal" href="#well-formedness" id="id609">Well-Formedness</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#identifiers" id="id610">Identifiers</a></li>
+<li><a class="reference internal" href="#high-level-structure" id="id611">High Level Structure</a><ul>
+<li><a class="reference internal" href="#module-structure" id="id612">Module Structure</a></li>
+<li><a class="reference internal" href="#linkage-types" id="id613">Linkage Types</a></li>
+<li><a class="reference internal" href="#calling-conventions" id="id614">Calling Conventions</a></li>
+<li><a class="reference internal" href="#visibility-styles" id="id615">Visibility Styles</a></li>
+<li><a class="reference internal" href="#dll-storage-classes" id="id616">DLL Storage Classes</a></li>
+<li><a class="reference internal" href="#thread-local-storage-models" id="id617">Thread Local Storage Models</a></li>
+<li><a class="reference internal" href="#structure-types" id="id618">Structure Types</a></li>
+<li><a class="reference internal" href="#global-variables" id="id619">Global Variables</a></li>
+<li><a class="reference internal" href="#functions" id="id620">Functions</a></li>
+<li><a class="reference internal" href="#aliases" id="id621">Aliases</a></li>
+<li><a class="reference internal" href="#comdats" id="id622">Comdats</a></li>
+<li><a class="reference internal" href="#named-metadata" id="id623">Named Metadata</a></li>
+<li><a class="reference internal" href="#parameter-attributes" id="id624">Parameter Attributes</a></li>
+<li><a class="reference internal" href="#garbage-collector-strategy-names" id="id625">Garbage Collector Strategy Names</a></li>
+<li><a class="reference internal" href="#prefix-data" id="id626">Prefix Data</a></li>
+<li><a class="reference internal" href="#prologue-data" id="id627">Prologue Data</a></li>
+<li><a class="reference internal" href="#personality-function" id="id628">Personality Function</a></li>
+<li><a class="reference internal" href="#attribute-groups" id="id629">Attribute Groups</a></li>
+<li><a class="reference internal" href="#function-attributes" id="id630">Function Attributes</a></li>
+<li><a class="reference internal" href="#module-level-inline-assembly" id="id631">Module-Level Inline Assembly</a></li>
+<li><a class="reference internal" href="#data-layout" id="id632">Data Layout</a></li>
+<li><a class="reference internal" href="#target-triple" id="id633">Target Triple</a></li>
+<li><a class="reference internal" href="#pointer-aliasing-rules" id="id634">Pointer Aliasing Rules</a></li>
+<li><a class="reference internal" href="#volatile-memory-accesses" id="id635">Volatile Memory Accesses</a></li>
+<li><a class="reference internal" href="#memory-model-for-concurrent-operations" id="id636">Memory Model for Concurrent Operations</a></li>
+<li><a class="reference internal" href="#atomic-memory-ordering-constraints" id="id637">Atomic Memory Ordering Constraints</a></li>
+<li><a class="reference internal" href="#fast-math-flags" id="id638">Fast-Math Flags</a></li>
+<li><a class="reference internal" href="#use-list-order-directives" id="id639">Use-list Order Directives</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#type-system" id="id640">Type System</a><ul>
+<li><a class="reference internal" href="#void-type" id="id641">Void Type</a></li>
+<li><a class="reference internal" href="#function-type" id="id642">Function Type</a></li>
+<li><a class="reference internal" href="#first-class-types" id="id643">First Class Types</a><ul>
+<li><a class="reference internal" href="#single-value-types" id="id644">Single Value Types</a><ul>
+<li><a class="reference internal" href="#integer-type" id="id645">Integer Type</a></li>
+<li><a class="reference internal" href="#floating-point-types" id="id646">Floating Point Types</a></li>
+<li><a class="reference internal" href="#x86-mmx-type" id="id647">X86_mmx Type</a></li>
+<li><a class="reference internal" href="#pointer-type" id="id648">Pointer Type</a></li>
+<li><a class="reference internal" href="#vector-type" id="id649">Vector Type</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#label-type" id="id650">Label Type</a></li>
+<li><a class="reference internal" href="#metadata-type" id="id651">Metadata Type</a></li>
+<li><a class="reference internal" href="#aggregate-types" id="id652">Aggregate Types</a><ul>
+<li><a class="reference internal" href="#array-type" id="id653">Array Type</a></li>
+<li><a class="reference internal" href="#structure-type" id="id654">Structure Type</a></li>
+<li><a class="reference internal" href="#opaque-structure-types" id="id655">Opaque Structure Types</a></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#constants" id="id656">Constants</a><ul>
+<li><a class="reference internal" href="#simple-constants" id="id657">Simple Constants</a></li>
+<li><a class="reference internal" href="#complex-constants" id="id658">Complex Constants</a></li>
+<li><a class="reference internal" href="#global-variable-and-function-addresses" id="id659">Global Variable and Function Addresses</a></li>
+<li><a class="reference internal" href="#undefined-values" id="id660">Undefined Values</a></li>
+<li><a class="reference internal" href="#poison-values" id="id661">Poison Values</a></li>
+<li><a class="reference internal" href="#addresses-of-basic-blocks" id="id662">Addresses of Basic Blocks</a></li>
+<li><a class="reference internal" href="#constant-expressions" id="id663">Constant Expressions</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#other-values" id="id664">Other Values</a><ul>
+<li><a class="reference internal" href="#inline-assembler-expressions" id="id665">Inline Assembler Expressions</a><ul>
+<li><a class="reference internal" href="#inline-asm-constraint-string" id="id666">Inline Asm Constraint String</a><ul>
+<li><a class="reference internal" href="#output-constraints" id="id667">Output constraints</a></li>
+<li><a class="reference internal" href="#input-constraints" id="id668">Input constraints</a></li>
+<li><a class="reference internal" href="#indirect-inputs-and-outputs" id="id669">Indirect inputs and outputs</a></li>
+<li><a class="reference internal" href="#clobber-constraints" id="id670">Clobber constraints</a></li>
+<li><a class="reference internal" href="#constraint-codes" id="id671">Constraint Codes</a></li>
+<li><a class="reference internal" href="#supported-constraint-code-list" id="id672">Supported Constraint Code List</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#asm-template-argument-modifiers" id="id673">Asm template argument modifiers</a></li>
+<li><a class="reference internal" href="#inline-asm-metadata" id="id674">Inline Asm Metadata</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#metadata" id="id675">Metadata</a><ul>
+<li><a class="reference internal" href="#metadata-nodes-and-metadata-strings" id="id676">Metadata Nodes and Metadata Strings</a><ul>
+<li><a class="reference internal" href="#specialized-metadata-nodes" id="id677">Specialized Metadata Nodes</a><ul>
+<li><a class="reference internal" href="#dicompileunit" id="id678">DICompileUnit</a></li>
+<li><a class="reference internal" href="#difile" id="id679">DIFile</a></li>
+<li><a class="reference internal" href="#dibasictype" id="id680">DIBasicType</a></li>
+<li><a class="reference internal" href="#disubroutinetype" id="id681">DISubroutineType</a></li>
+<li><a class="reference internal" href="#diderivedtype" id="id682">DIDerivedType</a></li>
+<li><a class="reference internal" href="#dicompositetype" id="id683">DICompositeType</a></li>
+<li><a class="reference internal" href="#disubrange" id="id684">DISubrange</a></li>
+<li><a class="reference internal" href="#dienumerator" id="id685">DIEnumerator</a></li>
+<li><a class="reference internal" href="#ditemplatetypeparameter" id="id686">DITemplateTypeParameter</a></li>
+<li><a class="reference internal" href="#ditemplatevalueparameter" id="id687">DITemplateValueParameter</a></li>
+<li><a class="reference internal" href="#dinamespace" id="id688">DINamespace</a></li>
+<li><a class="reference internal" href="#diglobalvariable" id="id689">DIGlobalVariable</a></li>
+<li><a class="reference internal" href="#disubprogram" id="id690">DISubprogram</a></li>
+<li><a class="reference internal" href="#dilexicalblock" id="id691">DILexicalBlock</a></li>
+<li><a class="reference internal" href="#dilexicalblockfile" id="id692">DILexicalBlockFile</a></li>
+<li><a class="reference internal" href="#dilocation" id="id693">DILocation</a></li>
+<li><a class="reference internal" href="#dilocalvariable" id="id694">DILocalVariable</a></li>
+<li><a class="reference internal" href="#diexpression" id="id695">DIExpression</a></li>
+<li><a class="reference internal" href="#diobjcproperty" id="id696">DIObjCProperty</a></li>
+<li><a class="reference internal" href="#diimportedentity" id="id697">DIImportedEntity</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#tbaa-metadata" id="id698">‘<tt class="docutils literal"><span class="pre">tbaa</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#tbaa-struct-metadata" id="id699">‘<tt class="docutils literal"><span class="pre">tbaa.struct</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#noalias-and-alias-scope-metadata" id="id700">‘<tt class="docutils literal"><span class="pre">noalias</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">alias.scope</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#fpmath-metadata" id="id701">‘<tt class="docutils literal"><span class="pre">fpmath</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#range-metadata" id="id702">‘<tt class="docutils literal"><span class="pre">range</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-loop" id="id703">‘<tt class="docutils literal"><span class="pre">llvm.loop</span></tt>‘</a></li>
+<li><a class="reference internal" href="#llvm-loop-vectorize-and-llvm-loop-interleave" id="id704">‘<tt class="docutils literal"><span class="pre">llvm.loop.vectorize</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">llvm.loop.interleave</span></tt>‘</a></li>
+<li><a class="reference internal" href="#llvm-loop-interleave-count-metadata" id="id705">‘<tt class="docutils literal"><span class="pre">llvm.loop.interleave.count</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-loop-vectorize-enable-metadata" id="id706">‘<tt class="docutils literal"><span class="pre">llvm.loop.vectorize.enable</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-loop-vectorize-width-metadata" id="id707">‘<tt class="docutils literal"><span class="pre">llvm.loop.vectorize.width</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-loop-unroll" id="id708">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll</span></tt>‘</a></li>
+<li><a class="reference internal" href="#llvm-loop-unroll-count-metadata" id="id709">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll.count</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-loop-unroll-disable-metadata" id="id710">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll.disable</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-loop-unroll-runtime-disable-metadata" id="id711">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll.runtime.disable</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-loop-unroll-full-metadata" id="id712">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll.full</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-mem" id="id713">‘<tt class="docutils literal"><span class="pre">llvm.mem</span></tt>‘</a></li>
+<li><a class="reference internal" href="#llvm-mem-parallel-loop-access-metadata" id="id714">‘<tt class="docutils literal"><span class="pre">llvm.mem.parallel_loop_access</span></tt>‘ Metadata</a></li>
+<li><a class="reference internal" href="#llvm-bitsets" id="id715">‘<tt class="docutils literal"><span class="pre">llvm.bitsets</span></tt>‘</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#module-flags-metadata" id="id716">Module Flags Metadata</a><ul>
+<li><a class="reference internal" href="#objective-c-garbage-collection-module-flags-metadata" id="id717">Objective-C Garbage Collection Module Flags Metadata</a></li>
+<li><a class="reference internal" href="#automatic-linker-flags-module-flags-metadata" id="id718">Automatic Linker Flags Module Flags Metadata</a></li>
+<li><a class="reference internal" href="#c-type-width-module-flags-metadata" id="id719">C type width Module Flags Metadata</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#intrinsic-global-variables" id="id720">Intrinsic Global Variables</a><ul>
+<li><a class="reference internal" href="#the-llvm-used-global-variable" id="id721">The ‘<tt class="docutils literal"><span class="pre">llvm.used</span></tt>‘ Global Variable</a></li>
+<li><a class="reference internal" href="#the-llvm-compiler-used-global-variable" id="id722">The ‘<tt class="docutils literal"><span class="pre">llvm.compiler.used</span></tt>‘ Global Variable</a></li>
+<li><a class="reference internal" href="#the-llvm-global-ctors-global-variable" id="id723">The ‘<tt class="docutils literal"><span class="pre">llvm.global_ctors</span></tt>‘ Global Variable</a></li>
+<li><a class="reference internal" href="#the-llvm-global-dtors-global-variable" id="id724">The ‘<tt class="docutils literal"><span class="pre">llvm.global_dtors</span></tt>‘ Global Variable</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#instruction-reference" id="id725">Instruction Reference</a><ul>
+<li><a class="reference internal" href="#terminator-instructions" id="id726">Terminator Instructions</a><ul>
+<li><a class="reference internal" href="#ret-instruction" id="id727">‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#syntax" id="id728">Syntax:</a></li>
+<li><a class="reference internal" href="#overview" id="id729">Overview:</a></li>
+<li><a class="reference internal" href="#arguments" id="id730">Arguments:</a></li>
+<li><a class="reference internal" href="#semantics" id="id731">Semantics:</a></li>
+<li><a class="reference internal" href="#example" id="id732">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#br-instruction" id="id733">‘<tt class="docutils literal"><span class="pre">br</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id19" id="id734">Syntax:</a></li>
+<li><a class="reference internal" href="#id20" id="id735">Overview:</a></li>
+<li><a class="reference internal" href="#id21" id="id736">Arguments:</a></li>
+<li><a class="reference internal" href="#id22" id="id737">Semantics:</a></li>
+<li><a class="reference internal" href="#id23" id="id738">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#switch-instruction" id="id739">‘<tt class="docutils literal"><span class="pre">switch</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id24" id="id740">Syntax:</a></li>
+<li><a class="reference internal" href="#id25" id="id741">Overview:</a></li>
+<li><a class="reference internal" href="#id26" id="id742">Arguments:</a></li>
+<li><a class="reference internal" href="#id27" id="id743">Semantics:</a></li>
+<li><a class="reference internal" href="#implementation" id="id744">Implementation:</a></li>
+<li><a class="reference internal" href="#id28" id="id745">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#indirectbr-instruction" id="id746">‘<tt class="docutils literal"><span class="pre">indirectbr</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id29" id="id747">Syntax:</a></li>
+<li><a class="reference internal" href="#id30" id="id748">Overview:</a></li>
+<li><a class="reference internal" href="#id31" id="id749">Arguments:</a></li>
+<li><a class="reference internal" href="#id32" id="id750">Semantics:</a></li>
+<li><a class="reference internal" href="#id33" id="id751">Implementation:</a></li>
+<li><a class="reference internal" href="#id34" id="id752">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#invoke-instruction" id="id753">‘<tt class="docutils literal"><span class="pre">invoke</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id35" id="id754">Syntax:</a></li>
+<li><a class="reference internal" href="#id36" id="id755">Overview:</a></li>
+<li><a class="reference internal" href="#id37" id="id756">Arguments:</a></li>
+<li><a class="reference internal" href="#id38" id="id757">Semantics:</a></li>
+<li><a class="reference internal" href="#id39" id="id758">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#resume-instruction" id="id759">‘<tt class="docutils literal"><span class="pre">resume</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id40" id="id760">Syntax:</a></li>
+<li><a class="reference internal" href="#id41" id="id761">Overview:</a></li>
+<li><a class="reference internal" href="#id42" id="id762">Arguments:</a></li>
+<li><a class="reference internal" href="#id43" id="id763">Semantics:</a></li>
+<li><a class="reference internal" href="#id44" id="id764">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#unreachable-instruction" id="id765">‘<tt class="docutils literal"><span class="pre">unreachable</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id45" id="id766">Syntax:</a></li>
+<li><a class="reference internal" href="#id46" id="id767">Overview:</a></li>
+<li><a class="reference internal" href="#id47" id="id768">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#binary-operations" id="id769">Binary Operations</a><ul>
+<li><a class="reference internal" href="#add-instruction" id="id770">‘<tt class="docutils literal"><span class="pre">add</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id48" id="id771">Syntax:</a></li>
+<li><a class="reference internal" href="#id49" id="id772">Overview:</a></li>
+<li><a class="reference internal" href="#id50" id="id773">Arguments:</a></li>
+<li><a class="reference internal" href="#id51" id="id774">Semantics:</a></li>
+<li><a class="reference internal" href="#id52" id="id775">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fadd-instruction" id="id776">‘<tt class="docutils literal"><span class="pre">fadd</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id53" id="id777">Syntax:</a></li>
+<li><a class="reference internal" href="#id54" id="id778">Overview:</a></li>
+<li><a class="reference internal" href="#id55" id="id779">Arguments:</a></li>
+<li><a class="reference internal" href="#id56" id="id780">Semantics:</a></li>
+<li><a class="reference internal" href="#id57" id="id781">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#sub-instruction" id="id782">‘<tt class="docutils literal"><span class="pre">sub</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id58" id="id783">Syntax:</a></li>
+<li><a class="reference internal" href="#id59" id="id784">Overview:</a></li>
+<li><a class="reference internal" href="#id60" id="id785">Arguments:</a></li>
+<li><a class="reference internal" href="#id61" id="id786">Semantics:</a></li>
+<li><a class="reference internal" href="#id62" id="id787">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fsub-instruction" id="id788">‘<tt class="docutils literal"><span class="pre">fsub</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id63" id="id789">Syntax:</a></li>
+<li><a class="reference internal" href="#id64" id="id790">Overview:</a></li>
+<li><a class="reference internal" href="#id65" id="id791">Arguments:</a></li>
+<li><a class="reference internal" href="#id66" id="id792">Semantics:</a></li>
+<li><a class="reference internal" href="#id67" id="id793">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#mul-instruction" id="id794">‘<tt class="docutils literal"><span class="pre">mul</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id68" id="id795">Syntax:</a></li>
+<li><a class="reference internal" href="#id69" id="id796">Overview:</a></li>
+<li><a class="reference internal" href="#id70" id="id797">Arguments:</a></li>
+<li><a class="reference internal" href="#id71" id="id798">Semantics:</a></li>
+<li><a class="reference internal" href="#id72" id="id799">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fmul-instruction" id="id800">‘<tt class="docutils literal"><span class="pre">fmul</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id73" id="id801">Syntax:</a></li>
+<li><a class="reference internal" href="#id74" id="id802">Overview:</a></li>
+<li><a class="reference internal" href="#id75" id="id803">Arguments:</a></li>
+<li><a class="reference internal" href="#id76" id="id804">Semantics:</a></li>
+<li><a class="reference internal" href="#id77" id="id805">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#udiv-instruction" id="id806">‘<tt class="docutils literal"><span class="pre">udiv</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id78" id="id807">Syntax:</a></li>
+<li><a class="reference internal" href="#id79" id="id808">Overview:</a></li>
+<li><a class="reference internal" href="#id80" id="id809">Arguments:</a></li>
+<li><a class="reference internal" href="#id81" id="id810">Semantics:</a></li>
+<li><a class="reference internal" href="#id82" id="id811">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#sdiv-instruction" id="id812">‘<tt class="docutils literal"><span class="pre">sdiv</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id83" id="id813">Syntax:</a></li>
+<li><a class="reference internal" href="#id84" id="id814">Overview:</a></li>
+<li><a class="reference internal" href="#id85" id="id815">Arguments:</a></li>
+<li><a class="reference internal" href="#id86" id="id816">Semantics:</a></li>
+<li><a class="reference internal" href="#id87" id="id817">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fdiv-instruction" id="id818">‘<tt class="docutils literal"><span class="pre">fdiv</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id88" id="id819">Syntax:</a></li>
+<li><a class="reference internal" href="#id89" id="id820">Overview:</a></li>
+<li><a class="reference internal" href="#id90" id="id821">Arguments:</a></li>
+<li><a class="reference internal" href="#id91" id="id822">Semantics:</a></li>
+<li><a class="reference internal" href="#id92" id="id823">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#urem-instruction" id="id824">‘<tt class="docutils literal"><span class="pre">urem</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id93" id="id825">Syntax:</a></li>
+<li><a class="reference internal" href="#id94" id="id826">Overview:</a></li>
+<li><a class="reference internal" href="#id95" id="id827">Arguments:</a></li>
+<li><a class="reference internal" href="#id96" id="id828">Semantics:</a></li>
+<li><a class="reference internal" href="#id97" id="id829">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#srem-instruction" id="id830">‘<tt class="docutils literal"><span class="pre">srem</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id98" id="id831">Syntax:</a></li>
+<li><a class="reference internal" href="#id99" id="id832">Overview:</a></li>
+<li><a class="reference internal" href="#id100" id="id833">Arguments:</a></li>
+<li><a class="reference internal" href="#id101" id="id834">Semantics:</a></li>
+<li><a class="reference internal" href="#id102" id="id835">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#frem-instruction" id="id836">‘<tt class="docutils literal"><span class="pre">frem</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id103" id="id837">Syntax:</a></li>
+<li><a class="reference internal" href="#id104" id="id838">Overview:</a></li>
+<li><a class="reference internal" href="#id105" id="id839">Arguments:</a></li>
+<li><a class="reference internal" href="#id106" id="id840">Semantics:</a></li>
+<li><a class="reference internal" href="#id107" id="id841">Example:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#bitwise-binary-operations" id="id842">Bitwise Binary Operations</a><ul>
+<li><a class="reference internal" href="#shl-instruction" id="id843">‘<tt class="docutils literal"><span class="pre">shl</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id108" id="id844">Syntax:</a></li>
+<li><a class="reference internal" href="#id109" id="id845">Overview:</a></li>
+<li><a class="reference internal" href="#id110" id="id846">Arguments:</a></li>
+<li><a class="reference internal" href="#id111" id="id847">Semantics:</a></li>
+<li><a class="reference internal" href="#id112" id="id848">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#lshr-instruction" id="id849">‘<tt class="docutils literal"><span class="pre">lshr</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id113" id="id850">Syntax:</a></li>
+<li><a class="reference internal" href="#id114" id="id851">Overview:</a></li>
+<li><a class="reference internal" href="#id115" id="id852">Arguments:</a></li>
+<li><a class="reference internal" href="#id116" id="id853">Semantics:</a></li>
+<li><a class="reference internal" href="#id117" id="id854">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#ashr-instruction" id="id855">‘<tt class="docutils literal"><span class="pre">ashr</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id118" id="id856">Syntax:</a></li>
+<li><a class="reference internal" href="#id119" id="id857">Overview:</a></li>
+<li><a class="reference internal" href="#id120" id="id858">Arguments:</a></li>
+<li><a class="reference internal" href="#id121" id="id859">Semantics:</a></li>
+<li><a class="reference internal" href="#id122" id="id860">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#and-instruction" id="id861">‘<tt class="docutils literal"><span class="pre">and</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id123" id="id862">Syntax:</a></li>
+<li><a class="reference internal" href="#id124" id="id863">Overview:</a></li>
+<li><a class="reference internal" href="#id125" id="id864">Arguments:</a></li>
+<li><a class="reference internal" href="#id126" id="id865">Semantics:</a></li>
+<li><a class="reference internal" href="#id127" id="id866">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#or-instruction" id="id867">‘<tt class="docutils literal"><span class="pre">or</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id128" id="id868">Syntax:</a></li>
+<li><a class="reference internal" href="#id129" id="id869">Overview:</a></li>
+<li><a class="reference internal" href="#id130" id="id870">Arguments:</a></li>
+<li><a class="reference internal" href="#id131" id="id871">Semantics:</a></li>
+<li><a class="reference internal" href="#id132" id="id872">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#xor-instruction" id="id873">‘<tt class="docutils literal"><span class="pre">xor</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id133" id="id874">Syntax:</a></li>
+<li><a class="reference internal" href="#id134" id="id875">Overview:</a></li>
+<li><a class="reference internal" href="#id135" id="id876">Arguments:</a></li>
+<li><a class="reference internal" href="#id136" id="id877">Semantics:</a></li>
+<li><a class="reference internal" href="#id137" id="id878">Example:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#vector-operations" id="id879">Vector Operations</a><ul>
+<li><a class="reference internal" href="#extractelement-instruction" id="id880">‘<tt class="docutils literal"><span class="pre">extractelement</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id138" id="id881">Syntax:</a></li>
+<li><a class="reference internal" href="#id139" id="id882">Overview:</a></li>
+<li><a class="reference internal" href="#id140" id="id883">Arguments:</a></li>
+<li><a class="reference internal" href="#id141" id="id884">Semantics:</a></li>
+<li><a class="reference internal" href="#id142" id="id885">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#insertelement-instruction" id="id886">‘<tt class="docutils literal"><span class="pre">insertelement</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id143" id="id887">Syntax:</a></li>
+<li><a class="reference internal" href="#id144" id="id888">Overview:</a></li>
+<li><a class="reference internal" href="#id145" id="id889">Arguments:</a></li>
+<li><a class="reference internal" href="#id146" id="id890">Semantics:</a></li>
+<li><a class="reference internal" href="#id147" id="id891">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#shufflevector-instruction" id="id892">‘<tt class="docutils literal"><span class="pre">shufflevector</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id148" id="id893">Syntax:</a></li>
+<li><a class="reference internal" href="#id149" id="id894">Overview:</a></li>
+<li><a class="reference internal" href="#id150" id="id895">Arguments:</a></li>
+<li><a class="reference internal" href="#id151" id="id896">Semantics:</a></li>
+<li><a class="reference internal" href="#id152" id="id897">Example:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#aggregate-operations" id="id898">Aggregate Operations</a><ul>
+<li><a class="reference internal" href="#extractvalue-instruction" id="id899">‘<tt class="docutils literal"><span class="pre">extractvalue</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id153" id="id900">Syntax:</a></li>
+<li><a class="reference internal" href="#id154" id="id901">Overview:</a></li>
+<li><a class="reference internal" href="#id155" id="id902">Arguments:</a></li>
+<li><a class="reference internal" href="#id156" id="id903">Semantics:</a></li>
+<li><a class="reference internal" href="#id157" id="id904">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#insertvalue-instruction" id="id905">‘<tt class="docutils literal"><span class="pre">insertvalue</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id158" id="id906">Syntax:</a></li>
+<li><a class="reference internal" href="#id159" id="id907">Overview:</a></li>
+<li><a class="reference internal" href="#id160" id="id908">Arguments:</a></li>
+<li><a class="reference internal" href="#id161" id="id909">Semantics:</a></li>
+<li><a class="reference internal" href="#id162" id="id910">Example:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#memory-access-and-addressing-operations" id="id911">Memory Access and Addressing Operations</a><ul>
+<li><a class="reference internal" href="#alloca-instruction" id="id912">‘<tt class="docutils literal"><span class="pre">alloca</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id163" id="id913">Syntax:</a></li>
+<li><a class="reference internal" href="#id164" id="id914">Overview:</a></li>
+<li><a class="reference internal" href="#id165" id="id915">Arguments:</a></li>
+<li><a class="reference internal" href="#id166" id="id916">Semantics:</a></li>
+<li><a class="reference internal" href="#id167" id="id917">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#load-instruction" id="id918">‘<tt class="docutils literal"><span class="pre">load</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id168" id="id919">Syntax:</a></li>
+<li><a class="reference internal" href="#id169" id="id920">Overview:</a></li>
+<li><a class="reference internal" href="#id170" id="id921">Arguments:</a></li>
+<li><a class="reference internal" href="#id171" id="id922">Semantics:</a></li>
+<li><a class="reference internal" href="#id172" id="id923">Examples:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#store-instruction" id="id924">‘<tt class="docutils literal"><span class="pre">store</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id173" id="id925">Syntax:</a></li>
+<li><a class="reference internal" href="#id174" id="id926">Overview:</a></li>
+<li><a class="reference internal" href="#id175" id="id927">Arguments:</a></li>
+<li><a class="reference internal" href="#id176" id="id928">Semantics:</a></li>
+<li><a class="reference internal" href="#id177" id="id929">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fence-instruction" id="id930">‘<tt class="docutils literal"><span class="pre">fence</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id178" id="id931">Syntax:</a></li>
+<li><a class="reference internal" href="#id179" id="id932">Overview:</a></li>
+<li><a class="reference internal" href="#id180" id="id933">Arguments:</a></li>
+<li><a class="reference internal" href="#id181" id="id934">Semantics:</a></li>
+<li><a class="reference internal" href="#id182" id="id935">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#cmpxchg-instruction" id="id936">‘<tt class="docutils literal"><span class="pre">cmpxchg</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id183" id="id937">Syntax:</a></li>
+<li><a class="reference internal" href="#id184" id="id938">Overview:</a></li>
+<li><a class="reference internal" href="#id185" id="id939">Arguments:</a></li>
+<li><a class="reference internal" href="#id186" id="id940">Semantics:</a></li>
+<li><a class="reference internal" href="#id187" id="id941">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#atomicrmw-instruction" id="id942">‘<tt class="docutils literal"><span class="pre">atomicrmw</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id188" id="id943">Syntax:</a></li>
+<li><a class="reference internal" href="#id189" id="id944">Overview:</a></li>
+<li><a class="reference internal" href="#id190" id="id945">Arguments:</a></li>
+<li><a class="reference internal" href="#id191" id="id946">Semantics:</a></li>
+<li><a class="reference internal" href="#id192" id="id947">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#getelementptr-instruction" id="id948">‘<tt class="docutils literal"><span class="pre">getelementptr</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id193" id="id949">Syntax:</a></li>
+<li><a class="reference internal" href="#id194" id="id950">Overview:</a></li>
+<li><a class="reference internal" href="#id195" id="id951">Arguments:</a></li>
+<li><a class="reference internal" href="#id196" id="id952">Semantics:</a></li>
+<li><a class="reference internal" href="#id197" id="id953">Example:</a></li>
+<li><a class="reference internal" href="#vector-of-pointers" id="id954">Vector of pointers:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#conversion-operations" id="id955">Conversion Operations</a><ul>
+<li><a class="reference internal" href="#trunc-to-instruction" id="id956">‘<tt class="docutils literal"><span class="pre">trunc</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id198" id="id957">Syntax:</a></li>
+<li><a class="reference internal" href="#id199" id="id958">Overview:</a></li>
+<li><a class="reference internal" href="#id200" id="id959">Arguments:</a></li>
+<li><a class="reference internal" href="#id201" id="id960">Semantics:</a></li>
+<li><a class="reference internal" href="#id202" id="id961">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#zext-to-instruction" id="id962">‘<tt class="docutils literal"><span class="pre">zext</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id203" id="id963">Syntax:</a></li>
+<li><a class="reference internal" href="#id204" id="id964">Overview:</a></li>
+<li><a class="reference internal" href="#id205" id="id965">Arguments:</a></li>
+<li><a class="reference internal" href="#id206" id="id966">Semantics:</a></li>
+<li><a class="reference internal" href="#id207" id="id967">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#sext-to-instruction" id="id968">‘<tt class="docutils literal"><span class="pre">sext</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id208" id="id969">Syntax:</a></li>
+<li><a class="reference internal" href="#id209" id="id970">Overview:</a></li>
+<li><a class="reference internal" href="#id210" id="id971">Arguments:</a></li>
+<li><a class="reference internal" href="#id211" id="id972">Semantics:</a></li>
+<li><a class="reference internal" href="#id212" id="id973">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fptrunc-to-instruction" id="id974">‘<tt class="docutils literal"><span class="pre">fptrunc</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id213" id="id975">Syntax:</a></li>
+<li><a class="reference internal" href="#id214" id="id976">Overview:</a></li>
+<li><a class="reference internal" href="#id215" id="id977">Arguments:</a></li>
+<li><a class="reference internal" href="#id216" id="id978">Semantics:</a></li>
+<li><a class="reference internal" href="#id217" id="id979">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fpext-to-instruction" id="id980">‘<tt class="docutils literal"><span class="pre">fpext</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id218" id="id981">Syntax:</a></li>
+<li><a class="reference internal" href="#id219" id="id982">Overview:</a></li>
+<li><a class="reference internal" href="#id220" id="id983">Arguments:</a></li>
+<li><a class="reference internal" href="#id221" id="id984">Semantics:</a></li>
+<li><a class="reference internal" href="#id222" id="id985">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fptoui-to-instruction" id="id986">‘<tt class="docutils literal"><span class="pre">fptoui</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id223" id="id987">Syntax:</a></li>
+<li><a class="reference internal" href="#id224" id="id988">Overview:</a></li>
+<li><a class="reference internal" href="#id225" id="id989">Arguments:</a></li>
+<li><a class="reference internal" href="#id226" id="id990">Semantics:</a></li>
+<li><a class="reference internal" href="#id227" id="id991">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fptosi-to-instruction" id="id992">‘<tt class="docutils literal"><span class="pre">fptosi</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id228" id="id993">Syntax:</a></li>
+<li><a class="reference internal" href="#id229" id="id994">Overview:</a></li>
+<li><a class="reference internal" href="#id230" id="id995">Arguments:</a></li>
+<li><a class="reference internal" href="#id231" id="id996">Semantics:</a></li>
+<li><a class="reference internal" href="#id232" id="id997">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#uitofp-to-instruction" id="id998">‘<tt class="docutils literal"><span class="pre">uitofp</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id233" id="id999">Syntax:</a></li>
+<li><a class="reference internal" href="#id234" id="id1000">Overview:</a></li>
+<li><a class="reference internal" href="#id235" id="id1001">Arguments:</a></li>
+<li><a class="reference internal" href="#id236" id="id1002">Semantics:</a></li>
+<li><a class="reference internal" href="#id237" id="id1003">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#sitofp-to-instruction" id="id1004">‘<tt class="docutils literal"><span class="pre">sitofp</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id238" id="id1005">Syntax:</a></li>
+<li><a class="reference internal" href="#id239" id="id1006">Overview:</a></li>
+<li><a class="reference internal" href="#id240" id="id1007">Arguments:</a></li>
+<li><a class="reference internal" href="#id241" id="id1008">Semantics:</a></li>
+<li><a class="reference internal" href="#id242" id="id1009">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#ptrtoint-to-instruction" id="id1010">‘<tt class="docutils literal"><span class="pre">ptrtoint</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id243" id="id1011">Syntax:</a></li>
+<li><a class="reference internal" href="#id244" id="id1012">Overview:</a></li>
+<li><a class="reference internal" href="#id245" id="id1013">Arguments:</a></li>
+<li><a class="reference internal" href="#id246" id="id1014">Semantics:</a></li>
+<li><a class="reference internal" href="#id247" id="id1015">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#inttoptr-to-instruction" id="id1016">‘<tt class="docutils literal"><span class="pre">inttoptr</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id248" id="id1017">Syntax:</a></li>
+<li><a class="reference internal" href="#id249" id="id1018">Overview:</a></li>
+<li><a class="reference internal" href="#id250" id="id1019">Arguments:</a></li>
+<li><a class="reference internal" href="#id251" id="id1020">Semantics:</a></li>
+<li><a class="reference internal" href="#id252" id="id1021">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#bitcast-to-instruction" id="id1022">‘<tt class="docutils literal"><span class="pre">bitcast</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id253" id="id1023">Syntax:</a></li>
+<li><a class="reference internal" href="#id254" id="id1024">Overview:</a></li>
+<li><a class="reference internal" href="#id255" id="id1025">Arguments:</a></li>
+<li><a class="reference internal" href="#id256" id="id1026">Semantics:</a></li>
+<li><a class="reference internal" href="#id257" id="id1027">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#addrspacecast-to-instruction" id="id1028">‘<tt class="docutils literal"><span class="pre">addrspacecast</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id258" id="id1029">Syntax:</a></li>
+<li><a class="reference internal" href="#id259" id="id1030">Overview:</a></li>
+<li><a class="reference internal" href="#id260" id="id1031">Arguments:</a></li>
+<li><a class="reference internal" href="#id261" id="id1032">Semantics:</a></li>
+<li><a class="reference internal" href="#id262" id="id1033">Example:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#other-operations" id="id1034">Other Operations</a><ul>
+<li><a class="reference internal" href="#icmp-instruction" id="id1035">‘<tt class="docutils literal"><span class="pre">icmp</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id263" id="id1036">Syntax:</a></li>
+<li><a class="reference internal" href="#id264" id="id1037">Overview:</a></li>
+<li><a class="reference internal" href="#id265" id="id1038">Arguments:</a></li>
+<li><a class="reference internal" href="#id266" id="id1039">Semantics:</a></li>
+<li><a class="reference internal" href="#id267" id="id1040">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fcmp-instruction" id="id1041">‘<tt class="docutils literal"><span class="pre">fcmp</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id268" id="id1042">Syntax:</a></li>
+<li><a class="reference internal" href="#id269" id="id1043">Overview:</a></li>
+<li><a class="reference internal" href="#id270" id="id1044">Arguments:</a></li>
+<li><a class="reference internal" href="#id271" id="id1045">Semantics:</a></li>
+<li><a class="reference internal" href="#id272" id="id1046">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#phi-instruction" id="id1047">‘<tt class="docutils literal"><span class="pre">phi</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id273" id="id1048">Syntax:</a></li>
+<li><a class="reference internal" href="#id274" id="id1049">Overview:</a></li>
+<li><a class="reference internal" href="#id275" id="id1050">Arguments:</a></li>
+<li><a class="reference internal" href="#id276" id="id1051">Semantics:</a></li>
+<li><a class="reference internal" href="#id277" id="id1052">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#select-instruction" id="id1053">‘<tt class="docutils literal"><span class="pre">select</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id278" id="id1054">Syntax:</a></li>
+<li><a class="reference internal" href="#id279" id="id1055">Overview:</a></li>
+<li><a class="reference internal" href="#id280" id="id1056">Arguments:</a></li>
+<li><a class="reference internal" href="#id281" id="id1057">Semantics:</a></li>
+<li><a class="reference internal" href="#id282" id="id1058">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#call-instruction" id="id1059">‘<tt class="docutils literal"><span class="pre">call</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id283" id="id1060">Syntax:</a></li>
+<li><a class="reference internal" href="#id284" id="id1061">Overview:</a></li>
+<li><a class="reference internal" href="#id285" id="id1062">Arguments:</a></li>
+<li><a class="reference internal" href="#id286" id="id1063">Semantics:</a></li>
+<li><a class="reference internal" href="#id287" id="id1064">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#va-arg-instruction" id="id1065">‘<tt class="docutils literal"><span class="pre">va_arg</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id288" id="id1066">Syntax:</a></li>
+<li><a class="reference internal" href="#id289" id="id1067">Overview:</a></li>
+<li><a class="reference internal" href="#id290" id="id1068">Arguments:</a></li>
+<li><a class="reference internal" href="#id291" id="id1069">Semantics:</a></li>
+<li><a class="reference internal" href="#id292" id="id1070">Example:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#landingpad-instruction" id="id1071">‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ Instruction</a><ul>
+<li><a class="reference internal" href="#id293" id="id1072">Syntax:</a></li>
+<li><a class="reference internal" href="#id294" id="id1073">Overview:</a></li>
+<li><a class="reference internal" href="#id295" id="id1074">Arguments:</a></li>
+<li><a class="reference internal" href="#id296" id="id1075">Semantics:</a></li>
+<li><a class="reference internal" href="#id297" id="id1076">Example:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#intrinsic-functions" id="id1077">Intrinsic Functions</a><ul>
+<li><a class="reference internal" href="#variable-argument-handling-intrinsics" id="id1078">Variable Argument Handling Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-va-start-intrinsic" id="id1079">‘<tt class="docutils literal"><span class="pre">llvm.va_start</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id298" id="id1080">Syntax:</a></li>
+<li><a class="reference internal" href="#id299" id="id1081">Overview:</a></li>
+<li><a class="reference internal" href="#id300" id="id1082">Arguments:</a></li>
+<li><a class="reference internal" href="#id301" id="id1083">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-va-end-intrinsic" id="id1084">‘<tt class="docutils literal"><span class="pre">llvm.va_end</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id302" id="id1085">Syntax:</a></li>
+<li><a class="reference internal" href="#id303" id="id1086">Overview:</a></li>
+<li><a class="reference internal" href="#id304" id="id1087">Arguments:</a></li>
+<li><a class="reference internal" href="#id305" id="id1088">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-va-copy-intrinsic" id="id1089">‘<tt class="docutils literal"><span class="pre">llvm.va_copy</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id306" id="id1090">Syntax:</a></li>
+<li><a class="reference internal" href="#id307" id="id1091">Overview:</a></li>
+<li><a class="reference internal" href="#id308" id="id1092">Arguments:</a></li>
+<li><a class="reference internal" href="#id309" id="id1093">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#accurate-garbage-collection-intrinsics" id="id1094">Accurate Garbage Collection Intrinsics</a><ul>
+<li><a class="reference internal" href="#experimental-statepoint-intrinsics" id="id1095">Experimental Statepoint Intrinsics</a></li>
+<li><a class="reference internal" href="#llvm-gcroot-intrinsic" id="id1096">‘<tt class="docutils literal"><span class="pre">llvm.gcroot</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id311" id="id1097">Syntax:</a></li>
+<li><a class="reference internal" href="#id312" id="id1098">Overview:</a></li>
+<li><a class="reference internal" href="#id313" id="id1099">Arguments:</a></li>
+<li><a class="reference internal" href="#id314" id="id1100">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-gcread-intrinsic" id="id1101">‘<tt class="docutils literal"><span class="pre">llvm.gcread</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id315" id="id1102">Syntax:</a></li>
+<li><a class="reference internal" href="#id316" id="id1103">Overview:</a></li>
+<li><a class="reference internal" href="#id317" id="id1104">Arguments:</a></li>
+<li><a class="reference internal" href="#id318" id="id1105">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-gcwrite-intrinsic" id="id1106">‘<tt class="docutils literal"><span class="pre">llvm.gcwrite</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id319" id="id1107">Syntax:</a></li>
+<li><a class="reference internal" href="#id320" id="id1108">Overview:</a></li>
+<li><a class="reference internal" href="#id321" id="id1109">Arguments:</a></li>
+<li><a class="reference internal" href="#id322" id="id1110">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#code-generator-intrinsics" id="id1111">Code Generator Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-returnaddress-intrinsic" id="id1112">‘<tt class="docutils literal"><span class="pre">llvm.returnaddress</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id323" id="id1113">Syntax:</a></li>
+<li><a class="reference internal" href="#id324" id="id1114">Overview:</a></li>
+<li><a class="reference internal" href="#id325" id="id1115">Arguments:</a></li>
+<li><a class="reference internal" href="#id326" id="id1116">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-frameaddress-intrinsic" id="id1117">‘<tt class="docutils literal"><span class="pre">llvm.frameaddress</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id327" id="id1118">Syntax:</a></li>
+<li><a class="reference internal" href="#id328" id="id1119">Overview:</a></li>
+<li><a class="reference internal" href="#id329" id="id1120">Arguments:</a></li>
+<li><a class="reference internal" href="#id330" id="id1121">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-localescape-and-llvm-localrecover-intrinsics" id="id1122">‘<tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">llvm.localrecover</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id331" id="id1123">Syntax:</a></li>
+<li><a class="reference internal" href="#id332" id="id1124">Overview:</a></li>
+<li><a class="reference internal" href="#id333" id="id1125">Arguments:</a></li>
+<li><a class="reference internal" href="#id334" id="id1126">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-read-register-and-llvm-write-register-intrinsics" id="id1127">‘<tt class="docutils literal"><span class="pre">llvm.read_register</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">llvm.write_register</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id335" id="id1128">Syntax:</a></li>
+<li><a class="reference internal" href="#id336" id="id1129">Overview:</a></li>
+<li><a class="reference internal" href="#id337" id="id1130">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-stacksave-intrinsic" id="id1131">‘<tt class="docutils literal"><span class="pre">llvm.stacksave</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id338" id="id1132">Syntax:</a></li>
+<li><a class="reference internal" href="#id339" id="id1133">Overview:</a></li>
+<li><a class="reference internal" href="#id340" id="id1134">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-stackrestore-intrinsic" id="id1135">‘<tt class="docutils literal"><span class="pre">llvm.stackrestore</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id341" id="id1136">Syntax:</a></li>
+<li><a class="reference internal" href="#id342" id="id1137">Overview:</a></li>
+<li><a class="reference internal" href="#id343" id="id1138">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-prefetch-intrinsic" id="id1139">‘<tt class="docutils literal"><span class="pre">llvm.prefetch</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id344" id="id1140">Syntax:</a></li>
+<li><a class="reference internal" href="#id345" id="id1141">Overview:</a></li>
+<li><a class="reference internal" href="#id346" id="id1142">Arguments:</a></li>
+<li><a class="reference internal" href="#id347" id="id1143">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-pcmarker-intrinsic" id="id1144">‘<tt class="docutils literal"><span class="pre">llvm.pcmarker</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id348" id="id1145">Syntax:</a></li>
+<li><a class="reference internal" href="#id349" id="id1146">Overview:</a></li>
+<li><a class="reference internal" href="#id350" id="id1147">Arguments:</a></li>
+<li><a class="reference internal" href="#id351" id="id1148">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-readcyclecounter-intrinsic" id="id1149">‘<tt class="docutils literal"><span class="pre">llvm.readcyclecounter</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id352" id="id1150">Syntax:</a></li>
+<li><a class="reference internal" href="#id353" id="id1151">Overview:</a></li>
+<li><a class="reference internal" href="#id354" id="id1152">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-clear-cache-intrinsic" id="id1153">‘<tt class="docutils literal"><span class="pre">llvm.clear_cache</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id355" id="id1154">Syntax:</a></li>
+<li><a class="reference internal" href="#id356" id="id1155">Overview:</a></li>
+<li><a class="reference internal" href="#id357" id="id1156">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-instrprof-increment-intrinsic" id="id1157">‘<tt class="docutils literal"><span class="pre">llvm.instrprof_increment</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id358" id="id1158">Syntax:</a></li>
+<li><a class="reference internal" href="#id359" id="id1159">Overview:</a></li>
+<li><a class="reference internal" href="#id360" id="id1160">Arguments:</a></li>
+<li><a class="reference internal" href="#id361" id="id1161">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#standard-c-library-intrinsics" id="id1162">Standard C Library Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-memcpy-intrinsic" id="id1163">‘<tt class="docutils literal"><span class="pre">llvm.memcpy</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id362" id="id1164">Syntax:</a></li>
+<li><a class="reference internal" href="#id363" id="id1165">Overview:</a></li>
+<li><a class="reference internal" href="#id364" id="id1166">Arguments:</a></li>
+<li><a class="reference internal" href="#id365" id="id1167">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-memmove-intrinsic" id="id1168">‘<tt class="docutils literal"><span class="pre">llvm.memmove</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id366" id="id1169">Syntax:</a></li>
+<li><a class="reference internal" href="#id367" id="id1170">Overview:</a></li>
+<li><a class="reference internal" href="#id368" id="id1171">Arguments:</a></li>
+<li><a class="reference internal" href="#id369" id="id1172">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-memset-intrinsics" id="id1173">‘<tt class="docutils literal"><span class="pre">llvm.memset.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id370" id="id1174">Syntax:</a></li>
+<li><a class="reference internal" href="#id371" id="id1175">Overview:</a></li>
+<li><a class="reference internal" href="#id372" id="id1176">Arguments:</a></li>
+<li><a class="reference internal" href="#id373" id="id1177">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-sqrt-intrinsic" id="id1178">‘<tt class="docutils literal"><span class="pre">llvm.sqrt.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id374" id="id1179">Syntax:</a></li>
+<li><a class="reference internal" href="#id375" id="id1180">Overview:</a></li>
+<li><a class="reference internal" href="#id376" id="id1181">Arguments:</a></li>
+<li><a class="reference internal" href="#id377" id="id1182">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-powi-intrinsic" id="id1183">‘<tt class="docutils literal"><span class="pre">llvm.powi.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id378" id="id1184">Syntax:</a></li>
+<li><a class="reference internal" href="#id379" id="id1185">Overview:</a></li>
+<li><a class="reference internal" href="#id380" id="id1186">Arguments:</a></li>
+<li><a class="reference internal" href="#id381" id="id1187">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-sin-intrinsic" id="id1188">‘<tt class="docutils literal"><span class="pre">llvm.sin.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id382" id="id1189">Syntax:</a></li>
+<li><a class="reference internal" href="#id383" id="id1190">Overview:</a></li>
+<li><a class="reference internal" href="#id384" id="id1191">Arguments:</a></li>
+<li><a class="reference internal" href="#id385" id="id1192">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-cos-intrinsic" id="id1193">‘<tt class="docutils literal"><span class="pre">llvm.cos.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id386" id="id1194">Syntax:</a></li>
+<li><a class="reference internal" href="#id387" id="id1195">Overview:</a></li>
+<li><a class="reference internal" href="#id388" id="id1196">Arguments:</a></li>
+<li><a class="reference internal" href="#id389" id="id1197">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-pow-intrinsic" id="id1198">‘<tt class="docutils literal"><span class="pre">llvm.pow.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id390" id="id1199">Syntax:</a></li>
+<li><a class="reference internal" href="#id391" id="id1200">Overview:</a></li>
+<li><a class="reference internal" href="#id392" id="id1201">Arguments:</a></li>
+<li><a class="reference internal" href="#id393" id="id1202">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-exp-intrinsic" id="id1203">‘<tt class="docutils literal"><span class="pre">llvm.exp.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id394" id="id1204">Syntax:</a></li>
+<li><a class="reference internal" href="#id395" id="id1205">Overview:</a></li>
+<li><a class="reference internal" href="#id396" id="id1206">Arguments:</a></li>
+<li><a class="reference internal" href="#id397" id="id1207">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-exp2-intrinsic" id="id1208">‘<tt class="docutils literal"><span class="pre">llvm.exp2.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id398" id="id1209">Syntax:</a></li>
+<li><a class="reference internal" href="#id399" id="id1210">Overview:</a></li>
+<li><a class="reference internal" href="#id400" id="id1211">Arguments:</a></li>
+<li><a class="reference internal" href="#id401" id="id1212">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-log-intrinsic" id="id1213">‘<tt class="docutils literal"><span class="pre">llvm.log.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id402" id="id1214">Syntax:</a></li>
+<li><a class="reference internal" href="#id403" id="id1215">Overview:</a></li>
+<li><a class="reference internal" href="#id404" id="id1216">Arguments:</a></li>
+<li><a class="reference internal" href="#id405" id="id1217">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-log10-intrinsic" id="id1218">‘<tt class="docutils literal"><span class="pre">llvm.log10.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id406" id="id1219">Syntax:</a></li>
+<li><a class="reference internal" href="#id407" id="id1220">Overview:</a></li>
+<li><a class="reference internal" href="#id408" id="id1221">Arguments:</a></li>
+<li><a class="reference internal" href="#id409" id="id1222">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-log2-intrinsic" id="id1223">‘<tt class="docutils literal"><span class="pre">llvm.log2.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id410" id="id1224">Syntax:</a></li>
+<li><a class="reference internal" href="#id411" id="id1225">Overview:</a></li>
+<li><a class="reference internal" href="#id412" id="id1226">Arguments:</a></li>
+<li><a class="reference internal" href="#id413" id="id1227">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-fma-intrinsic" id="id1228">‘<tt class="docutils literal"><span class="pre">llvm.fma.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id414" id="id1229">Syntax:</a></li>
+<li><a class="reference internal" href="#id415" id="id1230">Overview:</a></li>
+<li><a class="reference internal" href="#id416" id="id1231">Arguments:</a></li>
+<li><a class="reference internal" href="#id417" id="id1232">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-fabs-intrinsic" id="id1233">‘<tt class="docutils literal"><span class="pre">llvm.fabs.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id418" id="id1234">Syntax:</a></li>
+<li><a class="reference internal" href="#id419" id="id1235">Overview:</a></li>
+<li><a class="reference internal" href="#id420" id="id1236">Arguments:</a></li>
+<li><a class="reference internal" href="#id421" id="id1237">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-minnum-intrinsic" id="id1238">‘<tt class="docutils literal"><span class="pre">llvm.minnum.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id422" id="id1239">Syntax:</a></li>
+<li><a class="reference internal" href="#id423" id="id1240">Overview:</a></li>
+<li><a class="reference internal" href="#id424" id="id1241">Arguments:</a></li>
+<li><a class="reference internal" href="#id425" id="id1242">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-maxnum-intrinsic" id="id1243">‘<tt class="docutils literal"><span class="pre">llvm.maxnum.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id426" id="id1244">Syntax:</a></li>
+<li><a class="reference internal" href="#id427" id="id1245">Overview:</a></li>
+<li><a class="reference internal" href="#id428" id="id1246">Arguments:</a></li>
+<li><a class="reference internal" href="#id429" id="id1247">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-copysign-intrinsic" id="id1248">‘<tt class="docutils literal"><span class="pre">llvm.copysign.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id430" id="id1249">Syntax:</a></li>
+<li><a class="reference internal" href="#id431" id="id1250">Overview:</a></li>
+<li><a class="reference internal" href="#id432" id="id1251">Arguments:</a></li>
+<li><a class="reference internal" href="#id433" id="id1252">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-floor-intrinsic" id="id1253">‘<tt class="docutils literal"><span class="pre">llvm.floor.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id434" id="id1254">Syntax:</a></li>
+<li><a class="reference internal" href="#id435" id="id1255">Overview:</a></li>
+<li><a class="reference internal" href="#id436" id="id1256">Arguments:</a></li>
+<li><a class="reference internal" href="#id437" id="id1257">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-ceil-intrinsic" id="id1258">‘<tt class="docutils literal"><span class="pre">llvm.ceil.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id438" id="id1259">Syntax:</a></li>
+<li><a class="reference internal" href="#id439" id="id1260">Overview:</a></li>
+<li><a class="reference internal" href="#id440" id="id1261">Arguments:</a></li>
+<li><a class="reference internal" href="#id441" id="id1262">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-trunc-intrinsic" id="id1263">‘<tt class="docutils literal"><span class="pre">llvm.trunc.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id442" id="id1264">Syntax:</a></li>
+<li><a class="reference internal" href="#id443" id="id1265">Overview:</a></li>
+<li><a class="reference internal" href="#id444" id="id1266">Arguments:</a></li>
+<li><a class="reference internal" href="#id445" id="id1267">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-rint-intrinsic" id="id1268">‘<tt class="docutils literal"><span class="pre">llvm.rint.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id446" id="id1269">Syntax:</a></li>
+<li><a class="reference internal" href="#id447" id="id1270">Overview:</a></li>
+<li><a class="reference internal" href="#id448" id="id1271">Arguments:</a></li>
+<li><a class="reference internal" href="#id449" id="id1272">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-nearbyint-intrinsic" id="id1273">‘<tt class="docutils literal"><span class="pre">llvm.nearbyint.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id450" id="id1274">Syntax:</a></li>
+<li><a class="reference internal" href="#id451" id="id1275">Overview:</a></li>
+<li><a class="reference internal" href="#id452" id="id1276">Arguments:</a></li>
+<li><a class="reference internal" href="#id453" id="id1277">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-round-intrinsic" id="id1278">‘<tt class="docutils literal"><span class="pre">llvm.round.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id454" id="id1279">Syntax:</a></li>
+<li><a class="reference internal" href="#id455" id="id1280">Overview:</a></li>
+<li><a class="reference internal" href="#id456" id="id1281">Arguments:</a></li>
+<li><a class="reference internal" href="#id457" id="id1282">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#bit-manipulation-intrinsics" id="id1283">Bit Manipulation Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-bswap-intrinsics" id="id1284">‘<tt class="docutils literal"><span class="pre">llvm.bswap.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id458" id="id1285">Syntax:</a></li>
+<li><a class="reference internal" href="#id459" id="id1286">Overview:</a></li>
+<li><a class="reference internal" href="#id460" id="id1287">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-ctpop-intrinsic" id="id1288">‘<tt class="docutils literal"><span class="pre">llvm.ctpop.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id461" id="id1289">Syntax:</a></li>
+<li><a class="reference internal" href="#id462" id="id1290">Overview:</a></li>
+<li><a class="reference internal" href="#id463" id="id1291">Arguments:</a></li>
+<li><a class="reference internal" href="#id464" id="id1292">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-ctlz-intrinsic" id="id1293">‘<tt class="docutils literal"><span class="pre">llvm.ctlz.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id465" id="id1294">Syntax:</a></li>
+<li><a class="reference internal" href="#id466" id="id1295">Overview:</a></li>
+<li><a class="reference internal" href="#id467" id="id1296">Arguments:</a></li>
+<li><a class="reference internal" href="#id468" id="id1297">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-cttz-intrinsic" id="id1298">‘<tt class="docutils literal"><span class="pre">llvm.cttz.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id469" id="id1299">Syntax:</a></li>
+<li><a class="reference internal" href="#id470" id="id1300">Overview:</a></li>
+<li><a class="reference internal" href="#id471" id="id1301">Arguments:</a></li>
+<li><a class="reference internal" href="#id472" id="id1302">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#arithmetic-with-overflow-intrinsics" id="id1303">Arithmetic with Overflow Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-sadd-with-overflow-intrinsics" id="id1304">‘<tt class="docutils literal"><span class="pre">llvm.sadd.with.overflow.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id473" id="id1305">Syntax:</a></li>
+<li><a class="reference internal" href="#id474" id="id1306">Overview:</a></li>
+<li><a class="reference internal" href="#id475" id="id1307">Arguments:</a></li>
+<li><a class="reference internal" href="#id476" id="id1308">Semantics:</a></li>
+<li><a class="reference internal" href="#id477" id="id1309">Examples:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-uadd-with-overflow-intrinsics" id="id1310">‘<tt class="docutils literal"><span class="pre">llvm.uadd.with.overflow.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id478" id="id1311">Syntax:</a></li>
+<li><a class="reference internal" href="#id479" id="id1312">Overview:</a></li>
+<li><a class="reference internal" href="#id480" id="id1313">Arguments:</a></li>
+<li><a class="reference internal" href="#id481" id="id1314">Semantics:</a></li>
+<li><a class="reference internal" href="#id482" id="id1315">Examples:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-ssub-with-overflow-intrinsics" id="id1316">‘<tt class="docutils literal"><span class="pre">llvm.ssub.with.overflow.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id483" id="id1317">Syntax:</a></li>
+<li><a class="reference internal" href="#id484" id="id1318">Overview:</a></li>
+<li><a class="reference internal" href="#id485" id="id1319">Arguments:</a></li>
+<li><a class="reference internal" href="#id486" id="id1320">Semantics:</a></li>
+<li><a class="reference internal" href="#id487" id="id1321">Examples:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-usub-with-overflow-intrinsics" id="id1322">‘<tt class="docutils literal"><span class="pre">llvm.usub.with.overflow.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id488" id="id1323">Syntax:</a></li>
+<li><a class="reference internal" href="#id489" id="id1324">Overview:</a></li>
+<li><a class="reference internal" href="#id490" id="id1325">Arguments:</a></li>
+<li><a class="reference internal" href="#id491" id="id1326">Semantics:</a></li>
+<li><a class="reference internal" href="#id492" id="id1327">Examples:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-smul-with-overflow-intrinsics" id="id1328">‘<tt class="docutils literal"><span class="pre">llvm.smul.with.overflow.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id493" id="id1329">Syntax:</a></li>
+<li><a class="reference internal" href="#id494" id="id1330">Overview:</a></li>
+<li><a class="reference internal" href="#id495" id="id1331">Arguments:</a></li>
+<li><a class="reference internal" href="#id496" id="id1332">Semantics:</a></li>
+<li><a class="reference internal" href="#id497" id="id1333">Examples:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-umul-with-overflow-intrinsics" id="id1334">‘<tt class="docutils literal"><span class="pre">llvm.umul.with.overflow.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id498" id="id1335">Syntax:</a></li>
+<li><a class="reference internal" href="#id499" id="id1336">Overview:</a></li>
+<li><a class="reference internal" href="#id500" id="id1337">Arguments:</a></li>
+<li><a class="reference internal" href="#id501" id="id1338">Semantics:</a></li>
+<li><a class="reference internal" href="#id502" id="id1339">Examples:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#specialised-arithmetic-intrinsics" id="id1340">Specialised Arithmetic Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-canonicalize-intrinsic" id="id1341">‘<tt class="docutils literal"><span class="pre">llvm.canonicalize.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id503" id="id1342">Syntax:</a></li>
+<li><a class="reference internal" href="#id504" id="id1343">Overview:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-fmuladd-intrinsic" id="id1344">‘<tt class="docutils literal"><span class="pre">llvm.fmuladd.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id505" id="id1345">Syntax:</a></li>
+<li><a class="reference internal" href="#id506" id="id1346">Overview:</a></li>
+<li><a class="reference internal" href="#id507" id="id1347">Arguments:</a></li>
+<li><a class="reference internal" href="#id508" id="id1348">Semantics:</a></li>
+<li><a class="reference internal" href="#id509" id="id1349">Examples:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#half-precision-floating-point-intrinsics" id="id1350">Half Precision Floating Point Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-convert-to-fp16-intrinsic" id="id1351">‘<tt class="docutils literal"><span class="pre">llvm.convert.to.fp16</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id510" id="id1352">Syntax:</a></li>
+<li><a class="reference internal" href="#id511" id="id1353">Overview:</a></li>
+<li><a class="reference internal" href="#id512" id="id1354">Arguments:</a></li>
+<li><a class="reference internal" href="#id513" id="id1355">Semantics:</a></li>
+<li><a class="reference internal" href="#id514" id="id1356">Examples:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-convert-from-fp16-intrinsic" id="id1357">‘<tt class="docutils literal"><span class="pre">llvm.convert.from.fp16</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id515" id="id1358">Syntax:</a></li>
+<li><a class="reference internal" href="#id516" id="id1359">Overview:</a></li>
+<li><a class="reference internal" href="#id517" id="id1360">Arguments:</a></li>
+<li><a class="reference internal" href="#id518" id="id1361">Semantics:</a></li>
+<li><a class="reference internal" href="#id519" id="id1362">Examples:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#debugger-intrinsics" id="id1363">Debugger Intrinsics</a></li>
+<li><a class="reference internal" href="#exception-handling-intrinsics" id="id1364">Exception Handling Intrinsics</a></li>
+<li><a class="reference internal" href="#trampoline-intrinsics" id="id1365">Trampoline Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-init-trampoline-intrinsic" id="id1366">‘<tt class="docutils literal"><span class="pre">llvm.init.trampoline</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id520" id="id1367">Syntax:</a></li>
+<li><a class="reference internal" href="#id521" id="id1368">Overview:</a></li>
+<li><a class="reference internal" href="#id522" id="id1369">Arguments:</a></li>
+<li><a class="reference internal" href="#id523" id="id1370">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-adjust-trampoline-intrinsic" id="id1371">‘<tt class="docutils literal"><span class="pre">llvm.adjust.trampoline</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id524" id="id1372">Syntax:</a></li>
+<li><a class="reference internal" href="#id525" id="id1373">Overview:</a></li>
+<li><a class="reference internal" href="#id526" id="id1374">Arguments:</a></li>
+<li><a class="reference internal" href="#id527" id="id1375">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#masked-vector-load-and-store-intrinsics" id="id1376">Masked Vector Load and Store Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-masked-load-intrinsics" id="id1377">‘<tt class="docutils literal"><span class="pre">llvm.masked.load.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id528" id="id1378">Syntax:</a></li>
+<li><a class="reference internal" href="#id529" id="id1379">Overview:</a></li>
+<li><a class="reference internal" href="#id530" id="id1380">Arguments:</a></li>
+<li><a class="reference internal" href="#id531" id="id1381">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-masked-store-intrinsics" id="id1382">‘<tt class="docutils literal"><span class="pre">llvm.masked.store.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id532" id="id1383">Syntax:</a></li>
+<li><a class="reference internal" href="#id533" id="id1384">Overview:</a></li>
+<li><a class="reference internal" href="#id534" id="id1385">Arguments:</a></li>
+<li><a class="reference internal" href="#id535" id="id1386">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#masked-vector-gather-and-scatter-intrinsics" id="id1387">Masked Vector Gather and Scatter Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-masked-gather-intrinsics" id="id1388">‘<tt class="docutils literal"><span class="pre">llvm.masked.gather.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id536" id="id1389">Syntax:</a></li>
+<li><a class="reference internal" href="#id537" id="id1390">Overview:</a></li>
+<li><a class="reference internal" href="#id538" id="id1391">Arguments:</a></li>
+<li><a class="reference internal" href="#id539" id="id1392">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-masked-scatter-intrinsics" id="id1393">‘<tt class="docutils literal"><span class="pre">llvm.masked.scatter.*</span></tt>‘ Intrinsics</a><ul>
+<li><a class="reference internal" href="#id540" id="id1394">Syntax:</a></li>
+<li><a class="reference internal" href="#id541" id="id1395">Overview:</a></li>
+<li><a class="reference internal" href="#id542" id="id1396">Arguments:</a></li>
+<li><a class="reference internal" href="#id543" id="id1397">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#memory-use-markers" id="id1398">Memory Use Markers</a><ul>
+<li><a class="reference internal" href="#llvm-lifetime-start-intrinsic" id="id1399">‘<tt class="docutils literal"><span class="pre">llvm.lifetime.start</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id544" id="id1400">Syntax:</a></li>
+<li><a class="reference internal" href="#id545" id="id1401">Overview:</a></li>
+<li><a class="reference internal" href="#id546" id="id1402">Arguments:</a></li>
+<li><a class="reference internal" href="#id547" id="id1403">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-lifetime-end-intrinsic" id="id1404">‘<tt class="docutils literal"><span class="pre">llvm.lifetime.end</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id548" id="id1405">Syntax:</a></li>
+<li><a class="reference internal" href="#id549" id="id1406">Overview:</a></li>
+<li><a class="reference internal" href="#id550" id="id1407">Arguments:</a></li>
+<li><a class="reference internal" href="#id551" id="id1408">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-invariant-start-intrinsic" id="id1409">‘<tt class="docutils literal"><span class="pre">llvm.invariant.start</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id552" id="id1410">Syntax:</a></li>
+<li><a class="reference internal" href="#id553" id="id1411">Overview:</a></li>
+<li><a class="reference internal" href="#id554" id="id1412">Arguments:</a></li>
+<li><a class="reference internal" href="#id555" id="id1413">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-invariant-end-intrinsic" id="id1414">‘<tt class="docutils literal"><span class="pre">llvm.invariant.end</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id556" id="id1415">Syntax:</a></li>
+<li><a class="reference internal" href="#id557" id="id1416">Overview:</a></li>
+<li><a class="reference internal" href="#id558" id="id1417">Arguments:</a></li>
+<li><a class="reference internal" href="#id559" id="id1418">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#general-intrinsics" id="id1419">General Intrinsics</a><ul>
+<li><a class="reference internal" href="#llvm-var-annotation-intrinsic" id="id1420">‘<tt class="docutils literal"><span class="pre">llvm.var.annotation</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id560" id="id1421">Syntax:</a></li>
+<li><a class="reference internal" href="#id561" id="id1422">Overview:</a></li>
+<li><a class="reference internal" href="#id562" id="id1423">Arguments:</a></li>
+<li><a class="reference internal" href="#id563" id="id1424">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-ptr-annotation-intrinsic" id="id1425">‘<tt class="docutils literal"><span class="pre">llvm.ptr.annotation.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id564" id="id1426">Syntax:</a></li>
+<li><a class="reference internal" href="#id565" id="id1427">Overview:</a></li>
+<li><a class="reference internal" href="#id566" id="id1428">Arguments:</a></li>
+<li><a class="reference internal" href="#id567" id="id1429">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-annotation-intrinsic" id="id1430">‘<tt class="docutils literal"><span class="pre">llvm.annotation.*</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id568" id="id1431">Syntax:</a></li>
+<li><a class="reference internal" href="#id569" id="id1432">Overview:</a></li>
+<li><a class="reference internal" href="#id570" id="id1433">Arguments:</a></li>
+<li><a class="reference internal" href="#id571" id="id1434">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-trap-intrinsic" id="id1435">‘<tt class="docutils literal"><span class="pre">llvm.trap</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id572" id="id1436">Syntax:</a></li>
+<li><a class="reference internal" href="#id573" id="id1437">Overview:</a></li>
+<li><a class="reference internal" href="#id574" id="id1438">Arguments:</a></li>
+<li><a class="reference internal" href="#id575" id="id1439">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-debugtrap-intrinsic" id="id1440">‘<tt class="docutils literal"><span class="pre">llvm.debugtrap</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id576" id="id1441">Syntax:</a></li>
+<li><a class="reference internal" href="#id577" id="id1442">Overview:</a></li>
+<li><a class="reference internal" href="#id578" id="id1443">Arguments:</a></li>
+<li><a class="reference internal" href="#id579" id="id1444">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-stackprotector-intrinsic" id="id1445">‘<tt class="docutils literal"><span class="pre">llvm.stackprotector</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id580" id="id1446">Syntax:</a></li>
+<li><a class="reference internal" href="#id581" id="id1447">Overview:</a></li>
+<li><a class="reference internal" href="#id582" id="id1448">Arguments:</a></li>
+<li><a class="reference internal" href="#id583" id="id1449">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-stackprotectorcheck-intrinsic" id="id1450">‘<tt class="docutils literal"><span class="pre">llvm.stackprotectorcheck</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id584" id="id1451">Syntax:</a></li>
+<li><a class="reference internal" href="#id585" id="id1452">Overview:</a></li>
+<li><a class="reference internal" href="#id586" id="id1453">Arguments:</a></li>
+<li><a class="reference internal" href="#id587" id="id1454">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-objectsize-intrinsic" id="id1455">‘<tt class="docutils literal"><span class="pre">llvm.objectsize</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id588" id="id1456">Syntax:</a></li>
+<li><a class="reference internal" href="#id589" id="id1457">Overview:</a></li>
+<li><a class="reference internal" href="#id590" id="id1458">Arguments:</a></li>
+<li><a class="reference internal" href="#id591" id="id1459">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-expect-intrinsic" id="id1460">‘<tt class="docutils literal"><span class="pre">llvm.expect</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id592" id="id1461">Syntax:</a></li>
+<li><a class="reference internal" href="#id593" id="id1462">Overview:</a></li>
+<li><a class="reference internal" href="#id594" id="id1463">Arguments:</a></li>
+<li><a class="reference internal" href="#id595" id="id1464">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-assume-intrinsic" id="id1465">‘<tt class="docutils literal"><span class="pre">llvm.assume</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id596" id="id1466">Syntax:</a></li>
+<li><a class="reference internal" href="#id597" id="id1467">Overview:</a></li>
+<li><a class="reference internal" href="#id598" id="id1468">Arguments:</a></li>
+<li><a class="reference internal" href="#id599" id="id1469">Semantics:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-bitset-test-intrinsic" id="id1470">‘<tt class="docutils literal"><span class="pre">llvm.bitset.test</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id600" id="id1471">Syntax:</a></li>
+<li><a class="reference internal" href="#id601" id="id1472">Arguments:</a></li>
+<li><a class="reference internal" href="#id602" id="id1473">Overview:</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#llvm-donothing-intrinsic" id="id1474">‘<tt class="docutils literal"><span class="pre">llvm.donothing</span></tt>‘ Intrinsic</a><ul>
+<li><a class="reference internal" href="#id603" id="id1475">Syntax:</a></li>
+<li><a class="reference internal" href="#id604" id="id1476">Overview:</a></li>
+<li><a class="reference internal" href="#id605" id="id1477">Arguments:</a></li>
+<li><a class="reference internal" href="#id606" id="id1478">Semantics:</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#stack-map-intrinsics" id="id1479">Stack Map Intrinsics</a></li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="section" id="abstract">
+<h2><a class="toc-backref" href="#id607">Abstract</a><a class="headerlink" href="#abstract" title="Permalink to this headline">¶</a></h2>
+<p>This document is a reference manual for the LLVM assembly language. LLVM
+is a Static Single Assignment (SSA) based representation that provides
+type safety, low-level operations, flexibility, and the capability of
+representing ‘all’ high-level languages cleanly. It is the common code
+representation used throughout all phases of the LLVM compilation
+strategy.</p>
+</div>
+<div class="section" id="introduction">
+<h2><a class="toc-backref" href="#id608">Introduction</a><a class="headerlink" href="#introduction" title="Permalink to this headline">¶</a></h2>
+<p>The LLVM code representation is designed to be used in three different
+forms: as an in-memory compiler IR, as an on-disk bitcode representation
+(suitable for fast loading by a Just-In-Time compiler), and as a human
+readable assembly language representation. This allows LLVM to provide a
+powerful intermediate representation for efficient compiler
+transformations and analysis, while providing a natural means to debug
+and visualize the transformations. The three different forms of LLVM are
+all equivalent. This document describes the human readable
+representation and notation.</p>
+<p>The LLVM representation aims to be light-weight and low-level while
+being expressive, typed, and extensible at the same time. It aims to be
+a “universal IR” of sorts, by being at a low enough level that
+high-level ideas may be cleanly mapped to it (similar to how
+microprocessors are “universal IR’s”, allowing many source languages to
+be mapped to them). By providing type information, LLVM can be used as
+the target of optimizations: for example, through pointer analysis, it
+can be proven that a C automatic variable is never accessed outside of
+the current function, allowing it to be promoted to a simple SSA value
+instead of a memory location.</p>
+<div class="section" id="well-formedness">
+<span id="wellformed"></span><h3><a class="toc-backref" href="#id609">Well-Formedness</a><a class="headerlink" href="#well-formedness" title="Permalink to this headline">¶</a></h3>
+<p>It is important to note that this document describes ‘well formed’ LLVM
+assembly language. There is a difference between what the parser accepts
+and what is considered ‘well formed’. For example, the following
+instruction is syntactically okay, but not well formed:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%x</span> <span class="p">=</span> <span class="k">add</span> <span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="nv">%x</span>
+</pre></div>
+</div>
+<p>because the definition of <tt class="docutils literal"><span class="pre">%x</span></tt> does not dominate all of its uses. The
+LLVM infrastructure provides a verification pass that may be used to
+verify that an LLVM module is well formed. This pass is automatically
+run by the parser after parsing input assembly and by the optimizer
+before it outputs bitcode. The violations pointed out by the verifier
+pass indicate bugs in transformation passes or input to the parser.</p>
+</div>
+</div>
+<div class="section" id="identifiers">
+<span id="id1"></span><h2><a class="toc-backref" href="#id610">Identifiers</a><a class="headerlink" href="#identifiers" title="Permalink to this headline">¶</a></h2>
+<p>LLVM identifiers come in two basic types: global and local. Global
+identifiers (functions, global variables) begin with the <tt class="docutils literal"><span class="pre">'@'</span></tt>
+character. Local identifiers (register names, types) begin with the
+<tt class="docutils literal"><span class="pre">'%'</span></tt> character. Additionally, there are three different formats for
+identifiers, for different purposes:</p>
+<ol class="arabic simple">
+<li>Named values are represented as a string of characters with their
+prefix. For example, <tt class="docutils literal"><span class="pre">%foo</span></tt>, <tt class="docutils literal"><span class="pre">@DivisionByZero</span></tt>,
+<tt class="docutils literal"><span class="pre">%a.really.long.identifier</span></tt>. The actual regular expression used is
+‘<tt class="docutils literal"><span class="pre">[%@][-a-zA-Z$._][-a-zA-Z$._0-9]*</span></tt>‘. Identifiers that require other
+characters in their names can be surrounded with quotes. Special
+characters may be escaped using <tt class="docutils literal"><span class="pre">"\xx"</span></tt> where <tt class="docutils literal"><span class="pre">xx</span></tt> is the ASCII
+code for the character in hexadecimal. In this way, any character can
+be used in a name value, even quotes themselves. The <tt class="docutils literal"><span class="pre">"\01"</span></tt> prefix
+can be used on global variables to suppress mangling.</li>
+<li>Unnamed values are represented as an unsigned numeric value with
+their prefix. For example, <tt class="docutils literal"><span class="pre">%12</span></tt>, <tt class="docutils literal"><span class="pre">@2</span></tt>, <tt class="docutils literal"><span class="pre">%44</span></tt>.</li>
+<li>Constants, which are described in the section <a class="reference internal" href="#constants">Constants</a> below.</li>
+</ol>
+<p>LLVM requires that values start with a prefix for two reasons: Compilers
+don’t need to worry about name clashes with reserved words, and the set
+of reserved words may be expanded in the future without penalty.
+Additionally, unnamed identifiers allow a compiler to quickly come up
+with a temporary variable without having to avoid symbol table
+conflicts.</p>
+<p>Reserved words in LLVM are very similar to reserved words in other
+languages. There are keywords for different opcodes (‘<tt class="docutils literal"><span class="pre">add</span></tt>‘,
+‘<tt class="docutils literal"><span class="pre">bitcast</span></tt>‘, ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘, etc...), for primitive type names (‘<tt class="docutils literal"><span class="pre">void</span></tt>‘,
+‘<tt class="docutils literal"><span class="pre">i32</span></tt>‘, etc...), and others. These reserved words cannot conflict
+with variable names, because none of them start with a prefix character
+(<tt class="docutils literal"><span class="pre">'%'</span></tt> or <tt class="docutils literal"><span class="pre">'@'</span></tt>).</p>
+<p>Here is an example of LLVM code to multiply the integer variable
+‘<tt class="docutils literal"><span class="pre">%X</span></tt>‘ by 8:</p>
+<p>The easy way:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%result</span> <span class="p">=</span> <span class="k">mul</span> <span class="k">i32</span> <span class="nv">%X</span><span class="p">,</span> <span class="m">8</span>
+</pre></div>
+</div>
+<p>After strength reduction:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%result</span> <span class="p">=</span> <span class="k">shl</span> <span class="k">i32</span> <span class="nv">%X</span><span class="p">,</span> <span class="m">3</span>
+</pre></div>
+</div>
+<p>And the hard way:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">%0</span> <span class="p">=</span> <span class="k">add</span> <span class="k">i32</span> <span class="nv">%X</span><span class="p">,</span> <span class="nv">%X</span> <span class="c">; yields i32:%0</span>
+<span class="nv-Anonymous">%1</span> <span class="p">=</span> <span class="k">add</span> <span class="k">i32</span> <span class="nv-Anonymous">%0</span><span class="p">,</span> <span class="nv-Anonymous">%0</span> <span class="c">; yields i32:%1</span>
+<span class="nv">%result</span> <span class="p">=</span> <span class="k">add</span> <span class="k">i32</span> <span class="nv-Anonymous">%1</span><span class="p">,</span> <span class="nv-Anonymous">%1</span>
+</pre></div>
+</div>
+<p>This last way of multiplying <tt class="docutils literal"><span class="pre">%X</span></tt> by 8 illustrates several important
+lexical features of LLVM:</p>
+<ol class="arabic simple">
+<li>Comments are delimited with a ‘<tt class="docutils literal"><span class="pre">;</span></tt>‘ and go until the end of line.</li>
+<li>Unnamed temporaries are created when the result of a computation is
+not assigned to a named value.</li>
+<li>Unnamed temporaries are numbered sequentially (using a per-function
+incrementing counter, starting with 0). Note that basic blocks and unnamed
+function parameters are included in this numbering. For example, if the
+entry basic block is not given a label name and all function parameters are
+named, then it will get number 0.</li>
+</ol>
+<p>It also shows a convention that we follow in this document. When
+demonstrating instructions, we will follow an instruction with a comment
+that defines the type and name of value produced.</p>
+</div>
+<div class="section" id="high-level-structure">
+<h2><a class="toc-backref" href="#id611">High Level Structure</a><a class="headerlink" href="#high-level-structure" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="module-structure">
+<h3><a class="toc-backref" href="#id612">Module Structure</a><a class="headerlink" href="#module-structure" title="Permalink to this headline">¶</a></h3>
+<p>LLVM programs are composed of <tt class="docutils literal"><span class="pre">Module</span></tt>‘s, each of which is a
+translation unit of the input programs. Each module consists of
+functions, global variables, and symbol table entries. Modules may be
+combined together with the LLVM linker, which merges function (and
+global variable) definitions, resolves forward declarations, and merges
+symbol table entries. Here is an example of the “hello world” module:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>; Declare the string constant as a global constant.
+ at .str = private unnamed_addr constant [13 x i8] c"hello world\0A\00"
+
+; External declaration of the puts function
+declare i32 @puts(i8* nocapture) nounwind
+
+; Definition of main function
+define i32 @main() { ; i32()*
+ ; Convert [13 x i8]* to i8 *...
+ %cast210 = getelementptr [13 x i8], [13 x i8]* @.str, i64 0, i64 0
+
+ ; Call puts function to write out the string to stdout.
+ call i32 @puts(i8* %cast210)
+ ret i32 0
+}
+
+; Named metadata
+!0 = !{i32 42, null, !"string"}
+!foo = !{!0}
+</pre></div>
+</div>
+<p>This example is made up of a <a class="reference internal" href="#globalvars"><em>global variable</em></a> named
+“<tt class="docutils literal"><span class="pre">.str</span></tt>”, an external declaration of the “<tt class="docutils literal"><span class="pre">puts</span></tt>” function, a
+<a class="reference internal" href="#functionstructure"><em>function definition</em></a> for “<tt class="docutils literal"><span class="pre">main</span></tt>” and
+<a class="reference internal" href="#namedmetadatastructure"><em>named metadata</em></a> “<tt class="docutils literal"><span class="pre">foo</span></tt>”.</p>
+<p>In general, a module is made up of a list of global values (where both
+functions and global variables are global values). Global values are
+represented by a pointer to a memory location (in this case, a pointer
+to an array of char, and a pointer to a function), and have one of the
+following <a class="reference internal" href="#linkage"><em>linkage types</em></a>.</p>
+</div>
+<div class="section" id="linkage-types">
+<span id="linkage"></span><h3><a class="toc-backref" href="#id613">Linkage Types</a><a class="headerlink" href="#linkage-types" title="Permalink to this headline">¶</a></h3>
+<p>All Global Variables and Functions have one of the following types of
+linkage:</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">private</span></tt></dt>
+<dd>Global values with “<tt class="docutils literal"><span class="pre">private</span></tt>” linkage are only directly
+accessible by objects in the current module. In particular, linking
+code into a module with an private global value may cause the
+private to be renamed as necessary to avoid collisions. Because the
+symbol is private to the module, all references can be updated. This
+doesn’t show up in any symbol table in the object file.</dd>
+<dt><tt class="docutils literal"><span class="pre">internal</span></tt></dt>
+<dd>Similar to private, but the value shows as a local symbol
+(<tt class="docutils literal"><span class="pre">STB_LOCAL</span></tt> in the case of ELF) in the object file. This
+corresponds to the notion of the ‘<tt class="docutils literal"><span class="pre">static</span></tt>‘ keyword in C.</dd>
+<dt><tt class="docutils literal"><span class="pre">available_externally</span></tt></dt>
+<dd>Globals with “<tt class="docutils literal"><span class="pre">available_externally</span></tt>” linkage are never emitted
+into the object file corresponding to the LLVM module. They exist to
+allow inlining and other optimizations to take place given knowledge
+of the definition of the global, which is known to be somewhere
+outside the module. Globals with <tt class="docutils literal"><span class="pre">available_externally</span></tt> linkage
+are allowed to be discarded at will, and are otherwise the same as
+<tt class="docutils literal"><span class="pre">linkonce_odr</span></tt>. This linkage type is only allowed on definitions,
+not declarations.</dd>
+<dt><tt class="docutils literal"><span class="pre">linkonce</span></tt></dt>
+<dd>Globals with “<tt class="docutils literal"><span class="pre">linkonce</span></tt>” linkage are merged with other globals of
+the same name when linkage occurs. This can be used to implement
+some forms of inline functions, templates, or other code which must
+be generated in each translation unit that uses it, but where the
+body may be overridden with a more definitive definition later.
+Unreferenced <tt class="docutils literal"><span class="pre">linkonce</span></tt> globals are allowed to be discarded. Note
+that <tt class="docutils literal"><span class="pre">linkonce</span></tt> linkage does not actually allow the optimizer to
+inline the body of this function into callers because it doesn’t
+know if this definition of the function is the definitive definition
+within the program or whether it will be overridden by a stronger
+definition. To enable inlining and other optimizations, use
+“<tt class="docutils literal"><span class="pre">linkonce_odr</span></tt>” linkage.</dd>
+<dt><tt class="docutils literal"><span class="pre">weak</span></tt></dt>
+<dd>“<tt class="docutils literal"><span class="pre">weak</span></tt>” linkage has the same merging semantics as <tt class="docutils literal"><span class="pre">linkonce</span></tt>
+linkage, except that unreferenced globals with <tt class="docutils literal"><span class="pre">weak</span></tt> linkage may
+not be discarded. This is used for globals that are declared “weak”
+in C source code.</dd>
+<dt><tt class="docutils literal"><span class="pre">common</span></tt></dt>
+<dd>“<tt class="docutils literal"><span class="pre">common</span></tt>” linkage is most similar to “<tt class="docutils literal"><span class="pre">weak</span></tt>” linkage, but they
+are used for tentative definitions in C, such as “<tt class="docutils literal"><span class="pre">int</span> <span class="pre">X;</span></tt>” at
+global scope. Symbols with “<tt class="docutils literal"><span class="pre">common</span></tt>” linkage are merged in the
+same way as <tt class="docutils literal"><span class="pre">weak</span> <span class="pre">symbols</span></tt>, and they may not be deleted if
+unreferenced. <tt class="docutils literal"><span class="pre">common</span></tt> symbols may not have an explicit section,
+must have a zero initializer, and may not be marked
+‘<a class="reference internal" href="#globalvars"><em>constant</em></a>‘. Functions and aliases may not have
+common linkage.</dd>
+</dl>
+<dl class="docutils" id="linkage-appending">
+<dt><tt class="docutils literal"><span class="pre">appending</span></tt></dt>
+<dd>“<tt class="docutils literal"><span class="pre">appending</span></tt>” linkage may only be applied to global variables of
+pointer to array type. When two global variables with appending
+linkage are linked together, the two global arrays are appended
+together. This is the LLVM, typesafe, equivalent of having the
+system linker append together “sections” with identical names when
+.o files are linked.</dd>
+<dt><tt class="docutils literal"><span class="pre">extern_weak</span></tt></dt>
+<dd>The semantics of this linkage follow the ELF object file model: the
+symbol is weak until linked, if not linked, the symbol becomes null
+instead of being an undefined reference.</dd>
+<dt><tt class="docutils literal"><span class="pre">linkonce_odr</span></tt>, <tt class="docutils literal"><span class="pre">weak_odr</span></tt></dt>
+<dd>Some languages allow differing globals to be merged, such as two
+functions with different semantics. Other languages, such as
+<tt class="docutils literal"><span class="pre">C++</span></tt>, ensure that only equivalent globals are ever merged (the
+“one definition rule” — “ODR”). Such languages can use the
+<tt class="docutils literal"><span class="pre">linkonce_odr</span></tt> and <tt class="docutils literal"><span class="pre">weak_odr</span></tt> linkage types to indicate that the
+global will only be merged with equivalent globals. These linkage
+types are otherwise the same as their non-<tt class="docutils literal"><span class="pre">odr</span></tt> versions.</dd>
+<dt><tt class="docutils literal"><span class="pre">external</span></tt></dt>
+<dd>If none of the above identifiers are used, the global is externally
+visible, meaning that it participates in linkage and can be used to
+resolve external symbol references.</dd>
+</dl>
+<p>It is illegal for a function <em>declaration</em> to have any linkage type
+other than <tt class="docutils literal"><span class="pre">external</span></tt> or <tt class="docutils literal"><span class="pre">extern_weak</span></tt>.</p>
+</div>
+<div class="section" id="calling-conventions">
+<span id="callingconv"></span><h3><a class="toc-backref" href="#id614">Calling Conventions</a><a class="headerlink" href="#calling-conventions" title="Permalink to this headline">¶</a></h3>
+<p>LLVM <a class="reference internal" href="#functionstructure"><em>functions</em></a>, <a class="reference internal" href="#i-call"><em>calls</em></a> and
+<a class="reference internal" href="#i-invoke"><em>invokes</em></a> can all have an optional calling convention
+specified for the call. The calling convention of any pair of dynamic
+caller/callee must match, or the behavior of the program is undefined.
+The following calling conventions are supported by LLVM, and more may be
+added in the future:</p>
+<dl class="docutils">
+<dt>“<tt class="docutils literal"><span class="pre">ccc</span></tt>” - The C calling convention</dt>
+<dd>This calling convention (the default if no other calling convention
+is specified) matches the target C calling conventions. This calling
+convention supports varargs function calls and tolerates some
+mismatch in the declared prototype and implemented declaration of
+the function (as does normal C).</dd>
+<dt>“<tt class="docutils literal"><span class="pre">fastcc</span></tt>” - The fast calling convention</dt>
+<dd>This calling convention attempts to make calls as fast as possible
+(e.g. by passing things in registers). This calling convention
+allows the target to use whatever tricks it wants to produce fast
+code for the target, without having to conform to an externally
+specified ABI (Application Binary Interface). <a class="reference external" href="CodeGenerator.html#id80">Tail calls can only
+be optimized when this, the GHC or the HiPE convention is
+used.</a> This calling convention does not
+support varargs and requires the prototype of all callees to exactly
+match the prototype of the function definition.</dd>
+<dt>“<tt class="docutils literal"><span class="pre">coldcc</span></tt>” - The cold calling convention</dt>
+<dd>This calling convention attempts to make code in the caller as
+efficient as possible under the assumption that the call is not
+commonly executed. As such, these calls often preserve all registers
+so that the call does not break any live ranges in the caller side.
+This calling convention does not support varargs and requires the
+prototype of all callees to exactly match the prototype of the
+function definition. Furthermore the inliner doesn’t consider such function
+calls for inlining.</dd>
+<dt>“<tt class="docutils literal"><span class="pre">cc</span> <span class="pre">10</span></tt>” - GHC convention</dt>
+<dd><p class="first">This calling convention has been implemented specifically for use by
+the <a class="reference external" href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
+It passes everything in registers, going to extremes to achieve this
+by disabling callee save registers. This calling convention should
+not be used lightly but only for specific situations such as an
+alternative to the <em>register pinning</em> performance technique often
+used when implementing functional programming languages. At the
+moment only X86 supports this convention and it has the following
+limitations:</p>
+<ul class="simple">
+<li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
+floating point types are supported.</li>
+<li>On <em>X86-64</em> only supports up to 10 bit type parameters and 6
+floating point parameters.</li>
+</ul>
+<p class="last">This calling convention supports <a class="reference external" href="CodeGenerator.html#id80">tail call
+optimization</a> but requires both the
+caller and callee are using it.</p>
+</dd>
+<dt>“<tt class="docutils literal"><span class="pre">cc</span> <span class="pre">11</span></tt>” - The HiPE calling convention</dt>
+<dd>This calling convention has been implemented specifically for use by
+the <a class="reference external" href="http://www.it.uu.se/research/group/hipe/">High-Performance Erlang
+(HiPE)</a> compiler, <em>the</em>
+native code compiler of the <a class="reference external" href="http://www.erlang.org/download.shtml">Ericsson’s Open Source Erlang/OTP
+system</a>. It uses more
+registers for argument passing than the ordinary C calling
+convention and defines no callee-saved registers. The calling
+convention properly supports <a class="reference external" href="CodeGenerator.html#id80">tail call
+optimization</a> but requires that both the
+caller and the callee use it. It uses a <em>register pinning</em>
+mechanism, similar to GHC’s convention, for keeping frequently
+accessed runtime components pinned to specific hardware registers.
+At the moment only X86 supports this convention (both 32 and 64
+bit).</dd>
+<dt>“<tt class="docutils literal"><span class="pre">webkit_jscc</span></tt>” - WebKit’s JavaScript calling convention</dt>
+<dd>This calling convention has been implemented for <a class="reference external" href="https://trac.webkit.org/wiki/FTLJIT">WebKit FTL JIT</a>. It passes arguments on the
+stack right to left (as cdecl does), and returns a value in the
+platform’s customary return register.</dd>
+<dt>“<tt class="docutils literal"><span class="pre">anyregcc</span></tt>” - Dynamic calling convention for code patching</dt>
+<dd>This is a special convention that supports patching an arbitrary code
+sequence in place of a call site. This convention forces the call
+arguments into registers but allows them to be dynamically
+allocated. This can currently only be used with calls to
+llvm.experimental.patchpoint because only this intrinsic records
+the location of its arguments in a side table. See <a class="reference internal" href="StackMaps.html"><em>Stack maps and patch points in LLVM</em></a>.</dd>
+<dt>“<tt class="docutils literal"><span class="pre">preserve_mostcc</span></tt>” - The <cite>PreserveMost</cite> calling convention</dt>
+<dd><p class="first">This calling convention attempts to make the code in the caller as
+unintrusive as possible. This convention behaves identically to the <cite>C</cite>
+calling convention on how arguments and return values are passed, but it
+uses a different set of caller/callee-saved registers. This alleviates the
+burden of saving and recovering a large register set before and after the
+call in the caller. If the arguments are passed in callee-saved registers,
+then they will be preserved by the callee across the call. This doesn’t
+apply for values returned in callee-saved registers.</p>
+<ul class="simple">
+<li>On X86-64 the callee preserves all general purpose registers, except for
+R11. R11 can be used as a scratch register. Floating-point registers
+(XMMs/YMMs) are not preserved and need to be saved by the caller.</li>
+</ul>
+<p>The idea behind this convention is to support calls to runtime functions
+that have a hot path and a cold path. The hot path is usually a small piece
+of code that doesn’t use many registers. The cold path might need to call out to
+another function and therefore only needs to preserve the caller-saved
+registers, which haven’t already been saved by the caller. The
+<cite>PreserveMost</cite> calling convention is very similar to the <cite>cold</cite> calling
+convention in terms of caller/callee-saved registers, but they are used for
+different types of function calls. <cite>coldcc</cite> is for function calls that are
+rarely executed, whereas <cite>preserve_mostcc</cite> function calls are intended to be
+on the hot path and definitely executed a lot. Furthermore <cite>preserve_mostcc</cite>
+doesn’t prevent the inliner from inlining the function call.</p>
+<p class="last">This calling convention will be used by a future version of the ObjectiveC
+runtime and should therefore still be considered experimental at this time.
+Although this convention was created to optimize certain runtime calls to
+the ObjectiveC runtime, it is not limited to this runtime and might be used
+by other runtimes in the future too. The current implementation only
+supports X86-64, but the intention is to support more architectures in the
+future.</p>
+</dd>
+<dt>“<tt class="docutils literal"><span class="pre">preserve_allcc</span></tt>” - The <cite>PreserveAll</cite> calling convention</dt>
+<dd><p class="first">This calling convention attempts to make the code in the caller even less
+intrusive than the <cite>PreserveMost</cite> calling convention. This calling
+convention also behaves identical to the <cite>C</cite> calling convention on how
+arguments and return values are passed, but it uses a different set of
+caller/callee-saved registers. This removes the burden of saving and
+recovering a large register set before and after the call in the caller. If
+the arguments are passed in callee-saved registers, then they will be
+preserved by the callee across the call. This doesn’t apply for values
+returned in callee-saved registers.</p>
+<ul class="simple">
+<li>On X86-64 the callee preserves all general purpose registers, except for
+R11. R11 can be used as a scratch register. Furthermore it also preserves
+all floating-point registers (XMMs/YMMs).</li>
+</ul>
+<p>The idea behind this convention is to support calls to runtime functions
+that don’t need to call out to any other functions.</p>
+<p class="last">This calling convention, like the <cite>PreserveMost</cite> calling convention, will be
+used by a future version of the ObjectiveC runtime and should be considered
+experimental at this time.</p>
+</dd>
+<dt>“<tt class="docutils literal"><span class="pre">cc</span> <span class="pre"><n></span></tt>” - Numbered convention</dt>
+<dd>Any calling convention may be specified by number, allowing
+target-specific calling conventions to be used. Target specific
+calling conventions start at 64.</dd>
+</dl>
+<p>More calling conventions can be added/defined on an as-needed basis, to
+support Pascal conventions or any other well-known target-independent
+convention.</p>
+</div>
+<div class="section" id="visibility-styles">
+<span id="visibilitystyles"></span><h3><a class="toc-backref" href="#id615">Visibility Styles</a><a class="headerlink" href="#visibility-styles" title="Permalink to this headline">¶</a></h3>
+<p>All Global Variables and Functions have one of the following visibility
+styles:</p>
+<dl class="docutils">
+<dt>“<tt class="docutils literal"><span class="pre">default</span></tt>” - Default style</dt>
+<dd>On targets that use the ELF object file format, default visibility
+means that the declaration is visible to other modules and, in
+shared libraries, means that the declared entity may be overridden.
+On Darwin, default visibility means that the declaration is visible
+to other modules. Default visibility corresponds to “external
+linkage” in the language.</dd>
+<dt>“<tt class="docutils literal"><span class="pre">hidden</span></tt>” - Hidden style</dt>
+<dd>Two declarations of an object with hidden visibility refer to the
+same object if they are in the same shared object. Usually, hidden
+visibility indicates that the symbol will not be placed into the
+dynamic symbol table, so no other module (executable or shared
+library) can reference it directly.</dd>
+<dt>“<tt class="docutils literal"><span class="pre">protected</span></tt>” - Protected style</dt>
+<dd>On ELF, protected visibility indicates that the symbol will be
+placed in the dynamic symbol table, but that references within the
+defining module will bind to the local symbol. That is, the symbol
+cannot be overridden by another module.</dd>
+</dl>
+<p>A symbol with <tt class="docutils literal"><span class="pre">internal</span></tt> or <tt class="docutils literal"><span class="pre">private</span></tt> linkage must have <tt class="docutils literal"><span class="pre">default</span></tt>
+visibility.</p>
+</div>
+<div class="section" id="dll-storage-classes">
+<span id="dllstorageclass"></span><h3><a class="toc-backref" href="#id616">DLL Storage Classes</a><a class="headerlink" href="#dll-storage-classes" title="Permalink to this headline">¶</a></h3>
+<p>All Global Variables, Functions and Aliases can have one of the following
+DLL storage class:</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">dllimport</span></tt></dt>
+<dd>“<tt class="docutils literal"><span class="pre">dllimport</span></tt>” causes the compiler to reference a function or variable via
+a global pointer to a pointer that is set up by the DLL exporting the
+symbol. On Microsoft Windows targets, the pointer name is formed by
+combining <tt class="docutils literal"><span class="pre">__imp_</span></tt> and the function or variable name.</dd>
+<dt><tt class="docutils literal"><span class="pre">dllexport</span></tt></dt>
+<dd>“<tt class="docutils literal"><span class="pre">dllexport</span></tt>” causes the compiler to provide a global pointer to a pointer
+in a DLL, so that it can be referenced with the <tt class="docutils literal"><span class="pre">dllimport</span></tt> attribute. On
+Microsoft Windows targets, the pointer name is formed by combining
+<tt class="docutils literal"><span class="pre">__imp_</span></tt> and the function or variable name. Since this storage class
+exists for defining a dll interface, the compiler, assembler and linker know
+it is externally referenced and must refrain from deleting the symbol.</dd>
+</dl>
+</div>
+<div class="section" id="thread-local-storage-models">
+<span id="tls-model"></span><h3><a class="toc-backref" href="#id617">Thread Local Storage Models</a><a class="headerlink" href="#thread-local-storage-models" title="Permalink to this headline">¶</a></h3>
+<p>A variable may be defined as <tt class="docutils literal"><span class="pre">thread_local</span></tt>, which means that it will
+not be shared by threads (each thread will have a separated copy of the
+variable). Not all targets support thread-local variables. Optionally, a
+TLS model may be specified:</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">localdynamic</span></tt></dt>
+<dd>For variables that are only used within the current shared library.</dd>
+<dt><tt class="docutils literal"><span class="pre">initialexec</span></tt></dt>
+<dd>For variables in modules that will not be loaded dynamically.</dd>
+<dt><tt class="docutils literal"><span class="pre">localexec</span></tt></dt>
+<dd>For variables defined in the executable and only used within it.</dd>
+</dl>
+<p>If no explicit model is given, the “general dynamic” model is used.</p>
+<p>The models correspond to the ELF TLS models; see <a class="reference external" href="http://people.redhat.com/drepper/tls.pdf">ELF Handling For
+Thread-Local Storage</a> for
+more information on under which circumstances the different models may
+be used. The target may choose a different TLS model if the specified
+model is not supported, or if a better choice of model can be made.</p>
+<p>A model can also be specified in a alias, but then it only governs how
+the alias is accessed. It will not have any effect in the aliasee.</p>
+</div>
+<div class="section" id="structure-types">
+<span id="namedtypes"></span><h3><a class="toc-backref" href="#id618">Structure Types</a><a class="headerlink" href="#structure-types" title="Permalink to this headline">¶</a></h3>
+<p>LLVM IR allows you to specify both “identified” and “literal” <a class="reference internal" href="#t-struct"><em>structure
+types</em></a>. Literal types are uniqued structurally, but identified types
+are never uniqued. An <a class="reference internal" href="#t-opaque"><em>opaque structural type</em></a> can also be used
+to forward declare a type that is not yet available.</p>
+<p>An example of a identified structure specification is:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%mytype</span> <span class="p">=</span> <span class="k">type</span> <span class="p">{</span> <span class="nv">%mytype</span><span class="p">*,</span> <span class="k">i32</span> <span class="p">}</span>
+</pre></div>
+</div>
+<p>Prior to the LLVM 3.0 release, identified types were structurally uniqued. Only
+literal types are uniqued in recent versions of LLVM.</p>
+</div>
+<div class="section" id="global-variables">
+<span id="globalvars"></span><h3><a class="toc-backref" href="#id619">Global Variables</a><a class="headerlink" href="#global-variables" title="Permalink to this headline">¶</a></h3>
+<p>Global variables define regions of memory allocated at compilation time
+instead of run-time.</p>
+<p>Global variable definitions must be initialized.</p>
+<p>Global variables in other translation units can also be declared, in which
+case they don’t have an initializer.</p>
+<p>Either global variable definitions or declarations may have an explicit section
+to be placed in and may have an optional explicit alignment specified.</p>
+<p>A variable may be defined as a global <tt class="docutils literal"><span class="pre">constant</span></tt>, which indicates that
+the contents of the variable will <strong>never</strong> be modified (enabling better
+optimization, allowing the global data to be placed in the read-only
+section of an executable, etc). Note that variables that need runtime
+initialization cannot be marked <tt class="docutils literal"><span class="pre">constant</span></tt> as there is a store to the
+variable.</p>
+<p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
+constant, even if the final definition of the global is not. This
+capability can be used to enable slightly better optimization of the
+program, but requires the language definition to guarantee that
+optimizations based on the ‘constantness’ are valid for the translation
+units that do not include the definition.</p>
+<p>As SSA values, global variables define pointer values that are in scope
+(i.e. they dominate) all basic blocks in the program. Global variables
+always define a pointer to their “content” type because they describe a
+region of memory, and all memory objects in LLVM are accessed through
+pointers.</p>
+<p>Global variables can be marked with <tt class="docutils literal"><span class="pre">unnamed_addr</span></tt> which indicates
+that the address is not significant, only the content. Constants marked
+like this can be merged with other constants if they have the same
+initializer. Note that a constant with significant address <em>can</em> be
+merged with a <tt class="docutils literal"><span class="pre">unnamed_addr</span></tt> constant, the result being a constant
+whose address is significant.</p>
+<p>A global variable may be declared to reside in a target-specific
+numbered address space. For targets that support them, address spaces
+may affect how optimizations are performed and/or what target
+instructions are used to access the variable. The default address space
+is zero. The address space qualifier must precede any other attributes.</p>
+<p>LLVM allows an explicit section to be specified for globals. If the
+target supports it, it will emit globals to the section specified.
+Additionally, the global can placed in a comdat if the target has the necessary
+support.</p>
+<p>By default, global initializers are optimized by assuming that global
+variables defined within the module are not modified from their
+initial values before the start of the global initializer. This is
+true even for variables potentially accessible from outside the
+module, including those with external linkage or appearing in
+<tt class="docutils literal"><span class="pre">@llvm.used</span></tt> or dllexported variables. This assumption may be suppressed
+by marking the variable with <tt class="docutils literal"><span class="pre">externally_initialized</span></tt>.</p>
+<p>An explicit alignment may be specified for a global, which must be a
+power of 2. If not present, or if the alignment is set to zero, the
+alignment of the global is set by the target to whatever it feels
+convenient. If an explicit alignment is specified, the global is forced
+to have exactly that alignment. Targets and optimizers are not allowed
+to over-align the global if the global has an assigned section. In this
+case, the extra alignment could be observable: for example, code could
+assume that the globals are densely packed in their section and try to
+iterate over them as an array, alignment padding would break this
+iteration. The maximum alignment is <tt class="docutils literal"><span class="pre">1</span> <span class="pre"><<</span> <span class="pre">29</span></tt>.</p>
+<p>Globals can also have a <a class="reference internal" href="#dllstorageclass"><em>DLL storage class</em></a>.</p>
+<p>Variables and aliases can have a
+<a class="reference internal" href="#tls-model"><em>Thread Local Storage Model</em></a>.</p>
+<p>Syntax:</p>
+<div class="highlight-python"><div class="highlight"><pre>[@<GlobalVarName> =] [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal]
+ [unnamed_addr] [AddrSpace] [ExternallyInitialized]
+ <global | constant> <Type> [<InitializerConstant>]
+ [, section "name"] [, comdat [($name)]]
+ [, align <Alignment>]
+</pre></div>
+</div>
+<p>For example, the following defines a global in a numbered address space
+with an initializer, section, and alignment:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="vg">@G</span> <span class="p">=</span> <span class="k">addrspace</span><span class="p">(</span><span class="m">5</span><span class="p">)</span> <span class="k">constant</span> <span class="kt">float</span> <span class="m">1.0</span><span class="p">,</span> <span class="k">section</span> <span class="s">"foo"</span><span class="p">,</span> <span class="k">align</span> <span class="m">4</span>
+</pre></div>
+</div>
+<p>The following example just declares a global variable</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="vg">@G</span> <span class="p">=</span> <span class="k">external</span> <span class="k">global</span> <span class="k">i32</span>
+</pre></div>
+</div>
+<p>The following example defines a thread-local global with the
+<tt class="docutils literal"><span class="pre">initialexec</span></tt> TLS model:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>@G = thread_local(initialexec) global i32 0, align 4
+</pre></div>
+</div>
+</div>
+<div class="section" id="functions">
+<span id="functionstructure"></span><h3><a class="toc-backref" href="#id620">Functions</a><a class="headerlink" href="#functions" title="Permalink to this headline">¶</a></h3>
+<p>LLVM function definitions consist of the “<tt class="docutils literal"><span class="pre">define</span></tt>” keyword, an
+optional <a class="reference internal" href="#linkage"><em>linkage type</em></a>, an optional <a class="reference internal" href="BitCodeFormat.html#visibility"><em>visibility
+style</em></a>, an optional <a class="reference internal" href="#dllstorageclass"><em>DLL storage class</em></a>,
+an optional <a class="reference internal" href="#callingconv"><em>calling convention</em></a>,
+an optional <tt class="docutils literal"><span class="pre">unnamed_addr</span></tt> attribute, a return type, an optional
+<a class="reference internal" href="#paramattrs"><em>parameter attribute</em></a> for the return type, a function
+name, a (possibly empty) argument list (each with optional <a class="reference internal" href="#paramattrs"><em>parameter
+attributes</em></a>), optional <a class="reference internal" href="#fnattrs"><em>function attributes</em></a>,
+an optional section, an optional alignment,
+an optional <a class="reference internal" href="#langref-comdats"><em>comdat</em></a>,
+an optional <a class="reference internal" href="#gc"><em>garbage collector name</em></a>, an optional <a class="reference internal" href="#prefixdata"><em>prefix</em></a>,
+an optional <a class="reference internal" href="#prologuedata"><em>prologue</em></a>,
+an optional <a class="reference internal" href="#personalityfn"><em>personality</em></a>,
+an opening curly brace, a list of basic blocks, and a closing curly brace.</p>
+<p>LLVM function declarations consist of the “<tt class="docutils literal"><span class="pre">declare</span></tt>” keyword, an
+optional <a class="reference internal" href="#linkage"><em>linkage type</em></a>, an optional <a class="reference internal" href="BitCodeFormat.html#visibility"><em>visibility
+style</em></a>, an optional <a class="reference internal" href="#dllstorageclass"><em>DLL storage class</em></a>,
+an optional <a class="reference internal" href="#callingconv"><em>calling convention</em></a>,
+an optional <tt class="docutils literal"><span class="pre">unnamed_addr</span></tt> attribute, a return type, an optional
+<a class="reference internal" href="#paramattrs"><em>parameter attribute</em></a> for the return type, a function
+name, a possibly empty list of arguments, an optional alignment, an optional
+<a class="reference internal" href="#gc"><em>garbage collector name</em></a>, an optional <a class="reference internal" href="#prefixdata"><em>prefix</em></a>,
+and an optional <a class="reference internal" href="#prologuedata"><em>prologue</em></a>.</p>
+<p>A function definition contains a list of basic blocks, forming the CFG (Control
+Flow Graph) for the function. Each basic block may optionally start with a label
+(giving the basic block a symbol table entry), contains a list of instructions,
+and ends with a <a class="reference internal" href="#terminators"><em>terminator</em></a> instruction (such as a branch or
+function return). If an explicit label is not provided, a block is assigned an
+implicit numbered label, using the next value from the same counter as used for
+unnamed temporaries (<a class="reference internal" href="#identifiers"><em>see above</em></a>). For example, if a function
+entry block does not have an explicit label, it will be assigned label “%0”,
+then the first unnamed temporary in that block will be “%1”, etc.</p>
+<p>The first basic block in a function is special in two ways: it is
+immediately executed on entrance to the function, and it is not allowed
+to have predecessor basic blocks (i.e. there can not be any branches to
+the entry block of a function). Because the block can have no
+predecessors, it also cannot have any <a class="reference internal" href="#i-phi"><em>PHI nodes</em></a>.</p>
+<p>LLVM allows an explicit section to be specified for functions. If the
+target supports it, it will emit functions to the section specified.
+Additionally, the function can be placed in a COMDAT.</p>
+<p>An explicit alignment may be specified for a function. If not present,
+or if the alignment is set to zero, the alignment of the function is set
+by the target to whatever it feels convenient. If an explicit alignment
+is specified, the function is forced to have at least that much
+alignment. All alignments must be a power of 2.</p>
+<p>If the <tt class="docutils literal"><span class="pre">unnamed_addr</span></tt> attribute is given, the address is known to not
+be significant and two identical functions can be merged.</p>
+<p>Syntax:</p>
+<div class="highlight-python"><div class="highlight"><pre>define [linkage] [visibility] [DLLStorageClass]
+ [cconv] [ret attrs]
+ <ResultType> @<FunctionName> ([argument list])
+ [unnamed_addr] [fn Attrs] [section "name"] [comdat [($name)]]
+ [align N] [gc] [prefix Constant] [prologue Constant]
+ [personality Constant] { ... }
+</pre></div>
+</div>
+<p>The argument list is a comma seperated sequence of arguments where each
+argument is of the following form</p>
+<p>Syntax:</p>
+<div class="highlight-python"><div class="highlight"><pre><type> [parameter Attrs] [name]
+</pre></div>
+</div>
+</div>
+<div class="section" id="aliases">
+<span id="langref-aliases"></span><h3><a class="toc-backref" href="#id621">Aliases</a><a class="headerlink" href="#aliases" title="Permalink to this headline">¶</a></h3>
+<p>Aliases, unlike function or variables, don’t create any new data. They
+are just a new symbol and metadata for an existing position.</p>
+<p>Aliases have a name and an aliasee that is either a global value or a
+constant expression.</p>
+<p>Aliases may have an optional <a class="reference internal" href="#linkage"><em>linkage type</em></a>, an optional
+<a class="reference internal" href="BitCodeFormat.html#visibility"><em>visibility style</em></a>, an optional <a class="reference internal" href="#dllstorageclass"><em>DLL storage class</em></a> and an optional <a class="reference internal" href="#tls-model"><em>tls model</em></a>.</p>
+<p>Syntax:</p>
+<div class="highlight-python"><div class="highlight"><pre>@<Name> = [Linkage] [Visibility] [DLLStorageClass] [ThreadLocal] [unnamed_addr] alias <AliaseeTy> @<Aliasee>
+</pre></div>
+</div>
+<p>The linkage must be one of <tt class="docutils literal"><span class="pre">private</span></tt>, <tt class="docutils literal"><span class="pre">internal</span></tt>, <tt class="docutils literal"><span class="pre">linkonce</span></tt>, <tt class="docutils literal"><span class="pre">weak</span></tt>,
+<tt class="docutils literal"><span class="pre">linkonce_odr</span></tt>, <tt class="docutils literal"><span class="pre">weak_odr</span></tt>, <tt class="docutils literal"><span class="pre">external</span></tt>. Note that some system linkers
+might not correctly handle dropping a weak symbol that is aliased.</p>
+<p>Aliases that are not <tt class="docutils literal"><span class="pre">unnamed_addr</span></tt> are guaranteed to have the same address as
+the aliasee expression. <tt class="docutils literal"><span class="pre">unnamed_addr</span></tt> ones are only guaranteed to point
+to the same content.</p>
+<p>Since aliases are only a second name, some restrictions apply, of which
+some can only be checked when producing an object file:</p>
+<ul class="simple">
+<li>The expression defining the aliasee must be computable at assembly
+time. Since it is just a name, no relocations can be used.</li>
+<li>No alias in the expression can be weak as the possibility of the
+intermediate alias being overridden cannot be represented in an
+object file.</li>
+<li>No global value in the expression can be a declaration, since that
+would require a relocation, which is not possible.</li>
+</ul>
+</div>
+<div class="section" id="comdats">
+<span id="langref-comdats"></span><h3><a class="toc-backref" href="#id622">Comdats</a><a class="headerlink" href="#comdats" title="Permalink to this headline">¶</a></h3>
+<p>Comdat IR provides access to COFF and ELF object file COMDAT functionality.</p>
+<p>Comdats have a name which represents the COMDAT key. All global objects that
+specify this key will only end up in the final object file if the linker chooses
+that key over some other key. Aliases are placed in the same COMDAT that their
+aliasee computes to, if any.</p>
+<p>Comdats have a selection kind to provide input on how the linker should
+choose between keys in two different object files.</p>
+<p>Syntax:</p>
+<div class="highlight-python"><div class="highlight"><pre>$<Name> = comdat SelectionKind
+</pre></div>
+</div>
+<p>The selection kind must be one of the following:</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">any</span></tt></dt>
+<dd>The linker may choose any COMDAT key, the choice is arbitrary.</dd>
+<dt><tt class="docutils literal"><span class="pre">exactmatch</span></tt></dt>
+<dd>The linker may choose any COMDAT key but the sections must contain the
+same data.</dd>
+<dt><tt class="docutils literal"><span class="pre">largest</span></tt></dt>
+<dd>The linker will choose the section containing the largest COMDAT key.</dd>
+<dt><tt class="docutils literal"><span class="pre">noduplicates</span></tt></dt>
+<dd>The linker requires that only section with this COMDAT key exist.</dd>
+<dt><tt class="docutils literal"><span class="pre">samesize</span></tt></dt>
+<dd>The linker may choose any COMDAT key but the sections must contain the
+same amount of data.</dd>
+</dl>
+<p>Note that the Mach-O platform doesn’t support COMDATs and ELF only supports
+<tt class="docutils literal"><span class="pre">any</span></tt> as a selection kind.</p>
+<p>Here is an example of a COMDAT group where a function will only be selected if
+the COMDAT key’s section is the largest:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>$foo = comdat largest
+ at foo = global i32 2, comdat($foo)
+
+define void @bar() comdat($foo) {
+ ret void
+}
+</pre></div>
+</div>
+<p>As a syntactic sugar the <tt class="docutils literal"><span class="pre">$name</span></tt> can be omitted if the name is the same as
+the global name:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>$foo = comdat any
+ at foo = global i32 2, comdat
+</pre></div>
+</div>
+<p>In a COFF object file, this will create a COMDAT section with selection kind
+<tt class="docutils literal"><span class="pre">IMAGE_COMDAT_SELECT_LARGEST</span></tt> containing the contents of the <tt class="docutils literal"><span class="pre">@foo</span></tt> symbol
+and another COMDAT section with selection kind
+<tt class="docutils literal"><span class="pre">IMAGE_COMDAT_SELECT_ASSOCIATIVE</span></tt> which is associated with the first COMDAT
+section and contains the contents of the <tt class="docutils literal"><span class="pre">@bar</span></tt> symbol.</p>
+<p>There are some restrictions on the properties of the global object.
+It, or an alias to it, must have the same name as the COMDAT group when
+targeting COFF.
+The contents and size of this object may be used during link-time to determine
+which COMDAT groups get selected depending on the selection kind.
+Because the name of the object must match the name of the COMDAT group, the
+linkage of the global object must not be local; local symbols can get renamed
+if a collision occurs in the symbol table.</p>
+<p>The combined use of COMDATS and section attributes may yield surprising results.
+For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>$foo = comdat any
+$bar = comdat any
+ at g1 = global i32 42, section "sec", comdat($foo)
+ at g2 = global i32 42, section "sec", comdat($bar)
+</pre></div>
+</div>
+<p>From the object file perspective, this requires the creation of two sections
+with the same name. This is necessary because both globals belong to different
+COMDAT groups and COMDATs, at the object file level, are represented by
+sections.</p>
+<p>Note that certain IR constructs like global variables and functions may
+create COMDATs in the object file in addition to any which are specified using
+COMDAT IR. This arises when the code generator is configured to emit globals
+in individual sections (e.g. when <cite>-data-sections</cite> or <cite>-function-sections</cite>
+is supplied to <cite>llc</cite>).</p>
+</div>
+<div class="section" id="named-metadata">
+<span id="namedmetadatastructure"></span><h3><a class="toc-backref" href="#id623">Named Metadata</a><a class="headerlink" href="#named-metadata" title="Permalink to this headline">¶</a></h3>
+<p>Named metadata is a collection of metadata. <a class="reference internal" href="#metadata"><em>Metadata
+nodes</em></a> (but not metadata strings) are the only valid
+operands for a named metadata.</p>
+<ol class="arabic simple">
+<li>Named metadata are represented as a string of characters with the
+metadata prefix. The rules for metadata names are the same as for
+identifiers, but quoted names are not allowed. <tt class="docutils literal"><span class="pre">"\xx"</span></tt> type escapes
+are still valid, which allows any character to be part of a name.</li>
+</ol>
+<p>Syntax:</p>
+<div class="highlight-python"><div class="highlight"><pre>; Some unnamed metadata nodes, which are referenced by the named metadata.
+!0 = !{!"zero"}
+!1 = !{!"one"}
+!2 = !{!"two"}
+; A named metadata.
+!name = !{!0, !1, !2}
+</pre></div>
+</div>
+</div>
+<div class="section" id="parameter-attributes">
+<span id="paramattrs"></span><h3><a class="toc-backref" href="#id624">Parameter Attributes</a><a class="headerlink" href="#parameter-attributes" title="Permalink to this headline">¶</a></h3>
+<p>The return type and each parameter of a function type may have a set of
+<em>parameter attributes</em> associated with them. Parameter attributes are
+used to communicate additional information about the result or
+parameters of a function. Parameter attributes are considered to be part
+of the function, not of the function type, so functions with different
+parameter attributes can have the same function type.</p>
+<p>Parameter attributes are simple keywords that follow the type specified.
+If multiple parameter attributes are needed, they are space separated.
+For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">declare</span> <span class="k">i32</span> <span class="vg">@printf</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="k">noalias</span> <span class="k">nocapture</span><span class="p">,</span> <span class="p">...)</span>
+<span class="k">declare</span> <span class="k">i32</span> <span class="vg">@atoi</span><span class="p">(</span><span class="k">i8</span> <span class="k">zeroext</span><span class="p">)</span>
+<span class="k">declare</span> <span class="k">signext</span> <span class="k">i8</span> <span class="vg">@returns_signed_char</span><span class="p">()</span>
+</pre></div>
+</div>
+<p>Note that any attributes for the function result (<tt class="docutils literal"><span class="pre">nounwind</span></tt>,
+<tt class="docutils literal"><span class="pre">readonly</span></tt>) come immediately after the argument list.</p>
+<p>Currently, only the following parameter attributes are defined:</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">zeroext</span></tt></dt>
+<dd>This indicates to the code generator that the parameter or return
+value should be zero-extended to the extent required by the target’s
+ABI (which is usually 32-bits, but is 8-bits for a i1 on x86-64) by
+the caller (for a parameter) or the callee (for a return value).</dd>
+<dt><tt class="docutils literal"><span class="pre">signext</span></tt></dt>
+<dd>This indicates to the code generator that the parameter or return
+value should be sign-extended to the extent required by the target’s
+ABI (which is usually 32-bits) by the caller (for a parameter) or
+the callee (for a return value).</dd>
+<dt><tt class="docutils literal"><span class="pre">inreg</span></tt></dt>
+<dd>This indicates that this parameter or return value should be treated
+in a special target-dependent fashion during while emitting code for
+a function call or return (usually, by putting it in a register as
+opposed to memory, though some targets use it to distinguish between
+two different kinds of registers). Use of this attribute is
+target-specific.</dd>
+<dt><tt class="docutils literal"><span class="pre">byval</span></tt></dt>
+<dd><p class="first">This indicates that the pointer parameter should really be passed by
+value to the function. The attribute implies that a hidden copy of
+the pointee is made between the caller and the callee, so the callee
+is unable to modify the value in the caller. This attribute is only
+valid on LLVM pointer arguments. It is generally used to pass
+structs and arrays by value, but is also valid on pointers to
+scalars. The copy is considered to belong to the caller not the
+callee (for example, <tt class="docutils literal"><span class="pre">readonly</span></tt> functions should not write to
+<tt class="docutils literal"><span class="pre">byval</span></tt> parameters). This is not a valid attribute for return
+values.</p>
+<p class="last">The byval attribute also supports specifying an alignment with the
+align attribute. It indicates the alignment of the stack slot to
+form and the known alignment of the pointer specified to the call
+site. If the alignment is not specified, then the code generator
+makes a target-specific assumption.</p>
+</dd>
+</dl>
+<p id="attr-inalloca"><tt class="docutils literal"><span class="pre">inalloca</span></tt></p>
+<blockquote>
+<div><p>The <tt class="docutils literal"><span class="pre">inalloca</span></tt> argument attribute allows the caller to take the
+address of outgoing stack arguments. An <tt class="docutils literal"><span class="pre">inalloca</span></tt> argument must
+be a pointer to stack memory produced by an <tt class="docutils literal"><span class="pre">alloca</span></tt> instruction.
+The alloca, or argument allocation, must also be tagged with the
+inalloca keyword. Only the last argument may have the <tt class="docutils literal"><span class="pre">inalloca</span></tt>
+attribute, and that argument is guaranteed to be passed in memory.</p>
+<p>An argument allocation may be used by a call at most once because
+the call may deallocate it. The <tt class="docutils literal"><span class="pre">inalloca</span></tt> attribute cannot be
+used in conjunction with other attributes that affect argument
+storage, like <tt class="docutils literal"><span class="pre">inreg</span></tt>, <tt class="docutils literal"><span class="pre">nest</span></tt>, <tt class="docutils literal"><span class="pre">sret</span></tt>, or <tt class="docutils literal"><span class="pre">byval</span></tt>. The
+<tt class="docutils literal"><span class="pre">inalloca</span></tt> attribute also disables LLVM’s implicit lowering of
+large aggregate return values, which means that frontend authors
+must lower them with <tt class="docutils literal"><span class="pre">sret</span></tt> pointers.</p>
+<p>When the call site is reached, the argument allocation must have
+been the most recent stack allocation that is still live, or the
+results are undefined. It is possible to allocate additional stack
+space after an argument allocation and before its call site, but it
+must be cleared off with <a class="reference internal" href="#int-stackrestore"><em>llvm.stackrestore</em></a>.</p>
+<p>See <a class="reference internal" href="InAlloca.html"><em>Design and Usage of the InAlloca Attribute</em></a> for more information on how to use this
+attribute.</p>
+</div></blockquote>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">sret</span></tt></dt>
+<dd>This indicates that the pointer parameter specifies the address of a
+structure that is the return value of the function in the source
+program. This pointer must be guaranteed by the caller to be valid:
+loads and stores to the structure may be assumed by the callee
+not to trap and to be properly aligned. This may only be applied to
+the first parameter. This is not a valid attribute for return
+values.</dd>
+<dt><tt class="docutils literal"><span class="pre">align</span> <span class="pre"><n></span></tt></dt>
+<dd><p class="first">This indicates that the pointer value may be assumed by the optimizer to
+have the specified alignment.</p>
+<p class="last">Note that this attribute has additional semantics when combined with the
+<tt class="docutils literal"><span class="pre">byval</span></tt> attribute.</p>
+</dd>
+</dl>
+<dl class="docutils" id="noalias">
+<dt><tt class="docutils literal"><span class="pre">noalias</span></tt></dt>
+<dd><p class="first">This indicates that objects accessed via pointer values
+<a class="reference internal" href="#pointeraliasing"><em>based</em></a> on the argument or return value are not also
+accessed, during the execution of the function, via pointer values not
+<em>based</em> on the argument or return value. The attribute on a return value
+also has additional semantics described below. The caller shares the
+responsibility with the callee for ensuring that these requirements are met.
+For further details, please see the discussion of the NoAlias response in
+<a class="reference internal" href="AliasAnalysis.html#must-may-or-no"><em>alias analysis</em></a>.</p>
+<p>Note that this definition of <tt class="docutils literal"><span class="pre">noalias</span></tt> is intentionally similar
+to the definition of <tt class="docutils literal"><span class="pre">restrict</span></tt> in C99 for function arguments.</p>
+<p class="last">For function return values, C99’s <tt class="docutils literal"><span class="pre">restrict</span></tt> is not meaningful,
+while LLVM’s <tt class="docutils literal"><span class="pre">noalias</span></tt> is. Furthermore, the semantics of the <tt class="docutils literal"><span class="pre">noalias</span></tt>
+attribute on return values are stronger than the semantics of the attribute
+when used on function arguments. On function return values, the <tt class="docutils literal"><span class="pre">noalias</span></tt>
+attribute indicates that the function acts like a system memory allocation
+function, returning a pointer to allocated storage disjoint from the
+storage for any other object accessible to the caller.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">nocapture</span></tt></dt>
+<dd>This indicates that the callee does not make any copies of the
+pointer that outlive the callee itself. This is not a valid
+attribute for return values.</dd>
+</dl>
+<dl class="docutils" id="nest">
+<dt><tt class="docutils literal"><span class="pre">nest</span></tt></dt>
+<dd>This indicates that the pointer parameter can be excised using the
+<a class="reference internal" href="#int-trampoline"><em>trampoline intrinsics</em></a>. This is not a valid
+attribute for return values and can only be applied to one parameter.</dd>
+<dt><tt class="docutils literal"><span class="pre">returned</span></tt></dt>
+<dd>This indicates that the function always returns the argument as its return
+value. This is an optimization hint to the code generator when generating
+the caller, allowing tail call optimization and omission of register saves
+and restores in some cases; it is not checked or enforced when generating
+the callee. The parameter and the function return type must be valid
+operands for the <a class="reference internal" href="#i-bitcast"><em>bitcast instruction</em></a>. This is not a
+valid attribute for return values and can only be applied to one parameter.</dd>
+<dt><tt class="docutils literal"><span class="pre">nonnull</span></tt></dt>
+<dd>This indicates that the parameter or return pointer is not null. This
+attribute may only be applied to pointer typed parameters. This is not
+checked or enforced by LLVM, the caller must ensure that the pointer
+passed in is non-null, or the callee must ensure that the returned pointer
+is non-null.</dd>
+<dt><tt class="docutils literal"><span class="pre">dereferenceable(<n>)</span></tt></dt>
+<dd>This indicates that the parameter or return pointer is dereferenceable. This
+attribute may only be applied to pointer typed parameters. A pointer that
+is dereferenceable can be loaded from speculatively without a risk of
+trapping. The number of bytes known to be dereferenceable must be provided
+in parentheses. It is legal for the number of bytes to be less than the
+size of the pointee type. The <tt class="docutils literal"><span class="pre">nonnull</span></tt> attribute does not imply
+dereferenceability (consider a pointer to one element past the end of an
+array), however <tt class="docutils literal"><span class="pre">dereferenceable(<n>)</span></tt> does imply <tt class="docutils literal"><span class="pre">nonnull</span></tt> in
+<tt class="docutils literal"><span class="pre">addrspace(0)</span></tt> (which is the default address space).</dd>
+<dt><tt class="docutils literal"><span class="pre">dereferenceable_or_null(<n>)</span></tt></dt>
+<dd>This indicates that the parameter or return value isn’t both
+non-null and non-dereferenceable (up to <tt class="docutils literal"><span class="pre"><n></span></tt> bytes) at the same
+time. All non-null pointers tagged with
+<tt class="docutils literal"><span class="pre">dereferenceable_or_null(<n>)</span></tt> are <tt class="docutils literal"><span class="pre">dereferenceable(<n>)</span></tt>.
+For address space 0 <tt class="docutils literal"><span class="pre">dereferenceable_or_null(<n>)</span></tt> implies that
+a pointer is exactly one of <tt class="docutils literal"><span class="pre">dereferenceable(<n>)</span></tt> or <tt class="docutils literal"><span class="pre">null</span></tt>,
+and in other address spaces <tt class="docutils literal"><span class="pre">dereferenceable_or_null(<n>)</span></tt>
+implies that a pointer is at least one of <tt class="docutils literal"><span class="pre">dereferenceable(<n>)</span></tt>
+or <tt class="docutils literal"><span class="pre">null</span></tt> (i.e. it may be both <tt class="docutils literal"><span class="pre">null</span></tt> and
+<tt class="docutils literal"><span class="pre">dereferenceable(<n>)</span></tt>). This attribute may only be applied to
+pointer typed parameters.</dd>
+</dl>
+</div>
+<div class="section" id="garbage-collector-strategy-names">
+<span id="gc"></span><h3><a class="toc-backref" href="#id625">Garbage Collector Strategy Names</a><a class="headerlink" href="#garbage-collector-strategy-names" title="Permalink to this headline">¶</a></h3>
+<p>Each function may specify a garbage collector strategy name, which is simply a
+string:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">define</span> <span class="kt">void</span> <span class="vg">@f</span><span class="p">()</span> <span class="k">gc</span> <span class="s">"name"</span> <span class="p">{</span> <span class="p">...</span> <span class="p">}</span>
+</pre></div>
+</div>
+<p>The supported values of <em>name</em> includes those <a class="reference internal" href="GarbageCollection.html#builtin-gc-strategies"><em>built in to LLVM</em></a> and any provided by loaded plugins. Specifying a GC
+strategy will cause the compiler to alter its output in order to support the
+named garbage collection algorithm. Note that LLVM itself does not contain a
+garbage collector, this functionality is restricted to generating machine code
+which can interoperate with a collector provided externally.</p>
+</div>
+<div class="section" id="prefix-data">
+<span id="prefixdata"></span><h3><a class="toc-backref" href="#id626">Prefix Data</a><a class="headerlink" href="#prefix-data" title="Permalink to this headline">¶</a></h3>
+<p>Prefix data is data associated with a function which the code
+generator will emit immediately before the function’s entrypoint.
+The purpose of this feature is to allow frontends to associate
+language-specific runtime metadata with specific functions and make it
+available through the function pointer while still allowing the
+function pointer to be called.</p>
+<p>To access the data for a given function, a program may bitcast the
+function pointer to a pointer to the constant’s type and dereference
+index -1. This implies that the IR symbol points just past the end of
+the prefix data. For instance, take the example of a function annotated
+with a single <tt class="docutils literal"><span class="pre">i32</span></tt>,</p>
+<div class="highlight-llvm"><div class="highlight"><pre>define void @f() prefix i32 123 { ... }
+</pre></div>
+</div>
+<p>The prefix data can be referenced as,</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">%0</span> <span class="p">=</span> <span class="k">bitcast</span> <span class="kt">void</span><span class="p">*</span> <span class="p">()</span> <span class="vg">@f</span> <span class="k">to</span> <span class="k">i32</span><span class="p">*</span>
+<span class="nv">%a</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="k">inbounds</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv-Anonymous">%0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">-1</span>
+<span class="nv">%b</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%a</span>
+</pre></div>
+</div>
+<p>Prefix data is laid out as if it were an initializer for a global variable
+of the prefix data’s type. The function will be placed such that the
+beginning of the prefix data is aligned. This means that if the size
+of the prefix data is not a multiple of the alignment size, the
+function’s entrypoint will not be aligned. If alignment of the
+function’s entrypoint is desired, padding must be added to the prefix
+data.</p>
+<p>A function may have prefix data but no body. This has similar semantics
+to the <tt class="docutils literal"><span class="pre">available_externally</span></tt> linkage in that the data may be used by the
+optimizers but will not be emitted in the object file.</p>
+</div>
+<div class="section" id="prologue-data">
+<span id="prologuedata"></span><h3><a class="toc-backref" href="#id627">Prologue Data</a><a class="headerlink" href="#prologue-data" title="Permalink to this headline">¶</a></h3>
+<p>The <tt class="docutils literal"><span class="pre">prologue</span></tt> attribute allows arbitrary code (encoded as bytes) to
+be inserted prior to the function body. This can be used for enabling
+function hot-patching and instrumentation.</p>
+<p>To maintain the semantics of ordinary function calls, the prologue data must
+have a particular format. Specifically, it must begin with a sequence of
+bytes which decode to a sequence of machine instructions, valid for the
+module’s target, which transfer control to the point immediately succeeding
+the prologue data, without performing any other visible action. This allows
+the inliner and other passes to reason about the semantics of the function
+definition without needing to reason about the prologue data. Obviously this
+makes the format of the prologue data highly target dependent.</p>
+<p>A trivial example of valid prologue data for the x86 architecture is <tt class="docutils literal"><span class="pre">i8</span> <span class="pre">144</span></tt>,
+which encodes the <tt class="docutils literal"><span class="pre">nop</span></tt> instruction:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>define void @f() prologue i8 144 { ... }
+</pre></div>
+</div>
+<p>Generally prologue data can be formed by encoding a relative branch instruction
+which skips the metadata, as in this example of valid prologue data for the
+x86_64 architecture, where the first two bytes encode <tt class="docutils literal"><span class="pre">jmp</span> <span class="pre">.+10</span></tt>:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>%0 = type <{ i8, i8, i8* }>
+
+define void @f() prologue %0 <{ i8 235, i8 8, i8* @md}> { ... }
+</pre></div>
+</div>
+<p>A function may have prologue data but no body. This has similar semantics
+to the <tt class="docutils literal"><span class="pre">available_externally</span></tt> linkage in that the data may be used by the
+optimizers but will not be emitted in the object file.</p>
+</div>
+<div class="section" id="personality-function">
+<span id="personalityfn"></span><h3><a class="toc-backref" href="#id628">Personality Function</a><a class="headerlink" href="#personality-function" title="Permalink to this headline">¶</a></h3>
+<p>The <tt class="docutils literal"><span class="pre">personality</span></tt> attribute permits functions to specify what function
+to use for exception handling.</p>
+</div>
+<div class="section" id="attribute-groups">
+<span id="attrgrp"></span><h3><a class="toc-backref" href="#id629">Attribute Groups</a><a class="headerlink" href="#attribute-groups" title="Permalink to this headline">¶</a></h3>
+<p>Attribute groups are groups of attributes that are referenced by objects within
+the IR. They are important for keeping <tt class="docutils literal"><span class="pre">.ll</span></tt> files readable, because a lot of
+functions will use the same set of attributes. In the degenerative case of a
+<tt class="docutils literal"><span class="pre">.ll</span></tt> file that corresponds to a single <tt class="docutils literal"><span class="pre">.c</span></tt> file, the single attribute
+group will capture the important command line flags used to build that file.</p>
+<p>An attribute group is a module-level object. To use an attribute group, an
+object references the attribute group’s ID (e.g. <tt class="docutils literal"><span class="pre">#37</span></tt>). An object may refer
+to more than one attribute group. In that situation, the attributes from the
+different groups are merged.</p>
+<p>Here is an example of attribute groups for a function that should always be
+inlined, has a stack alignment of 4, and which shouldn’t use SSE instructions:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>; Target-independent attributes:
+attributes #0 = { alwaysinline alignstack=4 }
+
+; Target-dependent attributes:
+attributes #1 = { "no-sse" }
+
+; Function @f has attributes: alwaysinline, alignstack=4, and "no-sse".
+define void @f() #0 #1 { ... }
+</pre></div>
+</div>
+</div>
+<div class="section" id="function-attributes">
+<span id="fnattrs"></span><h3><a class="toc-backref" href="#id630">Function Attributes</a><a class="headerlink" href="#function-attributes" title="Permalink to this headline">¶</a></h3>
+<p>Function attributes are set to communicate additional information about
+a function. Function attributes are considered to be part of the
+function, not of the function type, so functions with different function
+attributes can have the same function type.</p>
+<p>Function attributes are simple keywords that follow the type specified.
+If multiple attributes are needed, they are space separated. For
+example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">define</span> <span class="kt">void</span> <span class="vg">@f</span><span class="p">()</span> <span class="k">noinline</span> <span class="p">{</span> <span class="p">...</span> <span class="p">}</span>
+<span class="k">define</span> <span class="kt">void</span> <span class="vg">@f</span><span class="p">()</span> <span class="k">alwaysinline</span> <span class="p">{</span> <span class="p">...</span> <span class="p">}</span>
+<span class="k">define</span> <span class="kt">void</span> <span class="vg">@f</span><span class="p">()</span> <span class="k">alwaysinline</span> <span class="k">optsize</span> <span class="p">{</span> <span class="p">...</span> <span class="p">}</span>
+<span class="k">define</span> <span class="kt">void</span> <span class="vg">@f</span><span class="p">()</span> <span class="k">optsize</span> <span class="p">{</span> <span class="p">...</span> <span class="p">}</span>
+</pre></div>
+</div>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">alignstack(<n>)</span></tt></dt>
+<dd>This attribute indicates that, when emitting the prologue and
+epilogue, the backend should forcibly align the stack pointer.
+Specify the desired alignment, which must be a power of two, in
+parentheses.</dd>
+<dt><tt class="docutils literal"><span class="pre">alwaysinline</span></tt></dt>
+<dd>This attribute indicates that the inliner should attempt to inline
+this function into callers whenever possible, ignoring any active
+inlining size threshold for this caller.</dd>
+<dt><tt class="docutils literal"><span class="pre">builtin</span></tt></dt>
+<dd>This indicates that the callee function at a call site should be
+recognized as a built-in function, even though the function’s declaration
+uses the <tt class="docutils literal"><span class="pre">nobuiltin</span></tt> attribute. This is only valid at call sites for
+direct calls to functions that are declared with the <tt class="docutils literal"><span class="pre">nobuiltin</span></tt>
+attribute.</dd>
+<dt><tt class="docutils literal"><span class="pre">cold</span></tt></dt>
+<dd>This attribute indicates that this function is rarely called. When
+computing edge weights, basic blocks post-dominated by a cold
+function call are also considered to be cold; and, thus, given low
+weight.</dd>
+<dt><tt class="docutils literal"><span class="pre">convergent</span></tt></dt>
+<dd>This attribute indicates that the callee is dependent on a convergent
+thread execution pattern under certain parallel execution models.
+Transformations that are execution model agnostic may only move or
+tranform this call if the final location is control equivalent to its
+original position in the program, where control equivalence is defined as
+A dominates B and B post-dominates A, or vice versa.</dd>
+<dt><tt class="docutils literal"><span class="pre">inlinehint</span></tt></dt>
+<dd>This attribute indicates that the source code contained a hint that
+inlining this function is desirable (such as the “inline” keyword in
+C/C++). It is just a hint; it imposes no requirements on the
+inliner.</dd>
+<dt><tt class="docutils literal"><span class="pre">jumptable</span></tt></dt>
+<dd>This attribute indicates that the function should be added to a
+jump-instruction table at code-generation time, and that all address-taken
+references to this function should be replaced with a reference to the
+appropriate jump-instruction-table function pointer. Note that this creates
+a new pointer for the original function, which means that code that depends
+on function-pointer identity can break. So, any function annotated with
+<tt class="docutils literal"><span class="pre">jumptable</span></tt> must also be <tt class="docutils literal"><span class="pre">unnamed_addr</span></tt>.</dd>
+<dt><tt class="docutils literal"><span class="pre">minsize</span></tt></dt>
+<dd>This attribute suggests that optimization passes and code generator
+passes make choices that keep the code size of this function as small
+as possible and perform optimizations that may sacrifice runtime
+performance in order to minimize the size of the generated code.</dd>
+<dt><tt class="docutils literal"><span class="pre">naked</span></tt></dt>
+<dd>This attribute disables prologue / epilogue emission for the
+function. This can have very system-specific consequences.</dd>
+<dt><tt class="docutils literal"><span class="pre">nobuiltin</span></tt></dt>
+<dd>This indicates that the callee function at a call site is not recognized as
+a built-in function. LLVM will retain the original call and not replace it
+with equivalent code based on the semantics of the built-in function, unless
+the call site uses the <tt class="docutils literal"><span class="pre">builtin</span></tt> attribute. This is valid at call sites
+and on function declarations and definitions.</dd>
+<dt><tt class="docutils literal"><span class="pre">noduplicate</span></tt></dt>
+<dd><p class="first">This attribute indicates that calls to the function cannot be
+duplicated. A call to a <tt class="docutils literal"><span class="pre">noduplicate</span></tt> function may be moved
+within its parent function, but may not be duplicated within
+its parent function.</p>
+<p class="last">A function containing a <tt class="docutils literal"><span class="pre">noduplicate</span></tt> call may still
+be an inlining candidate, provided that the call is not
+duplicated by inlining. That implies that the function has
+internal linkage and only has one call site, so the original
+call is dead after inlining.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">noimplicitfloat</span></tt></dt>
+<dd>This attributes disables implicit floating point instructions.</dd>
+<dt><tt class="docutils literal"><span class="pre">noinline</span></tt></dt>
+<dd>This attribute indicates that the inliner should never inline this
+function in any situation. This attribute may not be used together
+with the <tt class="docutils literal"><span class="pre">alwaysinline</span></tt> attribute.</dd>
+<dt><tt class="docutils literal"><span class="pre">nonlazybind</span></tt></dt>
+<dd>This attribute suppresses lazy symbol binding for the function. This
+may make calls to the function faster, at the cost of extra program
+startup time if the function is not called during program startup.</dd>
+<dt><tt class="docutils literal"><span class="pre">noredzone</span></tt></dt>
+<dd>This attribute indicates that the code generator should not use a
+red zone, even if the target-specific ABI normally permits it.</dd>
+<dt><tt class="docutils literal"><span class="pre">noreturn</span></tt></dt>
+<dd>This function attribute indicates that the function never returns
+normally. This produces undefined behavior at runtime if the
+function ever does dynamically return.</dd>
+<dt><tt class="docutils literal"><span class="pre">nounwind</span></tt></dt>
+<dd>This function attribute indicates that the function never raises an
+exception. If the function does raise an exception, its runtime
+behavior is undefined. However, functions marked nounwind may still
+trap or generate asynchronous exceptions. Exception handling schemes
+that are recognized by LLVM to handle asynchronous exceptions, such
+as SEH, will still provide their implementation defined semantics.</dd>
+<dt><tt class="docutils literal"><span class="pre">optnone</span></tt></dt>
+<dd><p class="first">This function attribute indicates that the function is not optimized
+by any optimization or code generator passes with the
+exception of interprocedural optimization passes.
+This attribute cannot be used together with the <tt class="docutils literal"><span class="pre">alwaysinline</span></tt>
+attribute; this attribute is also incompatible
+with the <tt class="docutils literal"><span class="pre">minsize</span></tt> attribute and the <tt class="docutils literal"><span class="pre">optsize</span></tt> attribute.</p>
+<p class="last">This attribute requires the <tt class="docutils literal"><span class="pre">noinline</span></tt> attribute to be specified on
+the function as well, so the function is never inlined into any caller.
+Only functions with the <tt class="docutils literal"><span class="pre">alwaysinline</span></tt> attribute are valid
+candidates for inlining into the body of this function.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">optsize</span></tt></dt>
+<dd>This attribute suggests that optimization passes and code generator
+passes make choices that keep the code size of this function low,
+and otherwise do optimizations specifically to reduce code size as
+long as they do not significantly impact runtime performance.</dd>
+<dt><tt class="docutils literal"><span class="pre">readnone</span></tt></dt>
+<dd><p class="first">On a function, this attribute indicates that the function computes its
+result (or decides to unwind an exception) based strictly on its arguments,
+without dereferencing any pointer arguments or otherwise accessing
+any mutable state (e.g. memory, control registers, etc) visible to
+caller functions. It does not write through any pointer arguments
+(including <tt class="docutils literal"><span class="pre">byval</span></tt> arguments) and never changes any state visible
+to callers. This means that it cannot unwind exceptions by calling
+the <tt class="docutils literal"><span class="pre">C++</span></tt> exception throwing methods.</p>
+<p class="last">On an argument, this attribute indicates that the function does not
+dereference that pointer argument, even though it may read or write the
+memory that the pointer points to if accessed through other pointers.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">readonly</span></tt></dt>
+<dd><p class="first">On a function, this attribute indicates that the function does not write
+through any pointer arguments (including <tt class="docutils literal"><span class="pre">byval</span></tt> arguments) or otherwise
+modify any state (e.g. memory, control registers, etc) visible to
+caller functions. It may dereference pointer arguments and read
+state that may be set in the caller. A readonly function always
+returns the same value (or unwinds an exception identically) when
+called with the same set of arguments and global state. It cannot
+unwind an exception by calling the <tt class="docutils literal"><span class="pre">C++</span></tt> exception throwing
+methods.</p>
+<p class="last">On an argument, this attribute indicates that the function does not write
+through this pointer argument, even though it may write to the memory that
+the pointer points to.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">argmemonly</span></tt></dt>
+<dd>This attribute indicates that the only memory accesses inside function are
+loads and stores from objects pointed to by its pointer-typed arguments,
+with arbitrary offsets. Or in other words, all memory operations in the
+function can refer to memory only using pointers based on its function
+arguments.
+Note that <tt class="docutils literal"><span class="pre">argmemonly</span></tt> can be used together with <tt class="docutils literal"><span class="pre">readonly</span></tt> attribute
+in order to specify that function reads only from its arguments.</dd>
+<dt><tt class="docutils literal"><span class="pre">returns_twice</span></tt></dt>
+<dd>This attribute indicates that this function can return twice. The C
+<tt class="docutils literal"><span class="pre">setjmp</span></tt> is an example of such a function. The compiler disables
+some optimizations (like tail calls) in the caller of these
+functions.</dd>
+<dt><tt class="docutils literal"><span class="pre">safestack</span></tt></dt>
+<dd><p class="first">This attribute indicates that
+<a class="reference external" href="http://clang.llvm.org/docs/SafeStack.html">SafeStack</a>
+protection is enabled for this function.</p>
+<p class="last">If a function that has a <tt class="docutils literal"><span class="pre">safestack</span></tt> attribute is inlined into a
+function that doesn’t have a <tt class="docutils literal"><span class="pre">safestack</span></tt> attribute or which has an
+<tt class="docutils literal"><span class="pre">ssp</span></tt>, <tt class="docutils literal"><span class="pre">sspstrong</span></tt> or <tt class="docutils literal"><span class="pre">sspreq</span></tt> attribute, then the resulting
+function will have a <tt class="docutils literal"><span class="pre">safestack</span></tt> attribute.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">sanitize_address</span></tt></dt>
+<dd>This attribute indicates that AddressSanitizer checks
+(dynamic address safety analysis) are enabled for this function.</dd>
+<dt><tt class="docutils literal"><span class="pre">sanitize_memory</span></tt></dt>
+<dd>This attribute indicates that MemorySanitizer checks (dynamic detection
+of accesses to uninitialized memory) are enabled for this function.</dd>
+<dt><tt class="docutils literal"><span class="pre">sanitize_thread</span></tt></dt>
+<dd>This attribute indicates that ThreadSanitizer checks
+(dynamic thread safety analysis) are enabled for this function.</dd>
+<dt><tt class="docutils literal"><span class="pre">ssp</span></tt></dt>
+<dd><p class="first">This attribute indicates that the function should emit a stack
+smashing protector. It is in the form of a “canary” — a random value
+placed on the stack before the local variables that’s checked upon
+return from the function to see if it has been overwritten. A
+heuristic is used to determine if a function needs stack protectors
+or not. The heuristic used will enable protectors for functions with:</p>
+<ul class="simple">
+<li>Character arrays larger than <tt class="docutils literal"><span class="pre">ssp-buffer-size</span></tt> (default 8).</li>
+<li>Aggregates containing character arrays larger than <tt class="docutils literal"><span class="pre">ssp-buffer-size</span></tt>.</li>
+<li>Calls to alloca() with variable sizes or constant sizes greater than
+<tt class="docutils literal"><span class="pre">ssp-buffer-size</span></tt>.</li>
+</ul>
+<p>Variables that are identified as requiring a protector will be arranged
+on the stack such that they are adjacent to the stack protector guard.</p>
+<p class="last">If a function that has an <tt class="docutils literal"><span class="pre">ssp</span></tt> attribute is inlined into a
+function that doesn’t have an <tt class="docutils literal"><span class="pre">ssp</span></tt> attribute, then the resulting
+function will have an <tt class="docutils literal"><span class="pre">ssp</span></tt> attribute.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">sspreq</span></tt></dt>
+<dd><p class="first">This attribute indicates that the function should <em>always</em> emit a
+stack smashing protector. This overrides the <tt class="docutils literal"><span class="pre">ssp</span></tt> function
+attribute.</p>
+<p>Variables that are identified as requiring a protector will be arranged
+on the stack such that they are adjacent to the stack protector guard.
+The specific layout rules are:</p>
+<ol class="arabic simple">
+<li>Large arrays and structures containing large arrays
+(<tt class="docutils literal"><span class="pre">>=</span> <span class="pre">ssp-buffer-size</span></tt>) are closest to the stack protector.</li>
+<li>Small arrays and structures containing small arrays
+(<tt class="docutils literal"><span class="pre"><</span> <span class="pre">ssp-buffer-size</span></tt>) are 2nd closest to the protector.</li>
+<li>Variables that have had their address taken are 3rd closest to the
+protector.</li>
+</ol>
+<p class="last">If a function that has an <tt class="docutils literal"><span class="pre">sspreq</span></tt> attribute is inlined into a
+function that doesn’t have an <tt class="docutils literal"><span class="pre">sspreq</span></tt> attribute or which has an
+<tt class="docutils literal"><span class="pre">ssp</span></tt> or <tt class="docutils literal"><span class="pre">sspstrong</span></tt> attribute, then the resulting function will have
+an <tt class="docutils literal"><span class="pre">sspreq</span></tt> attribute.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">sspstrong</span></tt></dt>
+<dd><p class="first">This attribute indicates that the function should emit a stack smashing
+protector. This attribute causes a strong heuristic to be used when
+determining if a function needs stack protectors. The strong heuristic
+will enable protectors for functions with:</p>
+<ul class="simple">
+<li>Arrays of any size and type</li>
+<li>Aggregates containing an array of any size and type.</li>
+<li>Calls to alloca().</li>
+<li>Local variables that have had their address taken.</li>
+</ul>
+<p>Variables that are identified as requiring a protector will be arranged
+on the stack such that they are adjacent to the stack protector guard.
+The specific layout rules are:</p>
+<ol class="arabic simple">
+<li>Large arrays and structures containing large arrays
+(<tt class="docutils literal"><span class="pre">>=</span> <span class="pre">ssp-buffer-size</span></tt>) are closest to the stack protector.</li>
+<li>Small arrays and structures containing small arrays
+(<tt class="docutils literal"><span class="pre"><</span> <span class="pre">ssp-buffer-size</span></tt>) are 2nd closest to the protector.</li>
+<li>Variables that have had their address taken are 3rd closest to the
+protector.</li>
+</ol>
+<p>This overrides the <tt class="docutils literal"><span class="pre">ssp</span></tt> function attribute.</p>
+<p class="last">If a function that has an <tt class="docutils literal"><span class="pre">sspstrong</span></tt> attribute is inlined into a
+function that doesn’t have an <tt class="docutils literal"><span class="pre">sspstrong</span></tt> attribute, then the
+resulting function will have an <tt class="docutils literal"><span class="pre">sspstrong</span></tt> attribute.</p>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">"thunk"</span></tt></dt>
+<dd>This attribute indicates that the function will delegate to some other
+function with a tail call. The prototype of a thunk should not be used for
+optimization purposes. The caller is expected to cast the thunk prototype to
+match the thunk target prototype.</dd>
+<dt><tt class="docutils literal"><span class="pre">uwtable</span></tt></dt>
+<dd>This attribute indicates that the ABI being targeted requires that
+an unwind table entry be produce for this function even if we can
+show that no exceptions passes by it. This is normally the case for
+the ELF x86-64 abi, but it can be disabled for some compilation
+units.</dd>
+</dl>
+</div>
+<div class="section" id="module-level-inline-assembly">
+<span id="moduleasm"></span><h3><a class="toc-backref" href="#id631">Module-Level Inline Assembly</a><a class="headerlink" href="#module-level-inline-assembly" title="Permalink to this headline">¶</a></h3>
+<p>Modules may contain “module-level inline asm” blocks, which corresponds
+to the GCC “file scope inline asm” blocks. These blocks are internally
+concatenated by LLVM and treated as a single unit, but may be separated
+in the <tt class="docutils literal"><span class="pre">.ll</span></tt> file if desired. The syntax is very simple:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">module</span> <span class="k">asm</span> <span class="s">"inline asm code goes here"</span>
+<span class="k">module</span> <span class="k">asm</span> <span class="s">"more can go here"</span>
+</pre></div>
+</div>
+<p>The strings can contain any character by escaping non-printable
+characters. The escape sequence used is simply “\xx” where “xx” is the
+two digit hex code for the number.</p>
+<p>Note that the assembly string <em>must</em> be parseable by LLVM’s integrated assembler
+(unless it is disabled), even when emitting a <tt class="docutils literal"><span class="pre">.s</span></tt> file.</p>
+</div>
+<div class="section" id="data-layout">
+<span id="langref-datalayout"></span><h3><a class="toc-backref" href="#id632">Data Layout</a><a class="headerlink" href="#data-layout" title="Permalink to this headline">¶</a></h3>
+<p>A module may specify a target specific data layout string that specifies
+how data is to be laid out in memory. The syntax for the data layout is
+simply:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">target</span> <span class="k">datalayout</span> <span class="p">=</span> <span class="s">"layout specification"</span>
+</pre></div>
+</div>
+<p>The <em>layout specification</em> consists of a list of specifications
+separated by the minus sign character (‘-‘). Each specification starts
+with a letter and may include other information after the letter to
+define some aspect of the data layout. The specifications accepted are
+as follows:</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">E</span></tt></dt>
+<dd>Specifies that the target lays out data in big-endian form. That is,
+the bits with the most significance have the lowest address
+location.</dd>
+<dt><tt class="docutils literal"><span class="pre">e</span></tt></dt>
+<dd>Specifies that the target lays out data in little-endian form. That
+is, the bits with the least significance have the lowest address
+location.</dd>
+<dt><tt class="docutils literal"><span class="pre">S<size></span></tt></dt>
+<dd>Specifies the natural alignment of the stack in bits. Alignment
+promotion of stack variables is limited to the natural stack
+alignment to avoid dynamic stack realignment. The stack alignment
+must be a multiple of 8-bits. If omitted, the natural stack
+alignment defaults to “unspecified”, which does not prevent any
+alignment promotions.</dd>
+<dt><tt class="docutils literal"><span class="pre">p[n]:<size>:<abi>:<pref></span></tt></dt>
+<dd>This specifies the <em>size</em> of a pointer and its <tt class="docutils literal"><span class="pre"><abi></span></tt> and
+<tt class="docutils literal"><span class="pre"><pref></span></tt>erred alignments for address space <tt class="docutils literal"><span class="pre">n</span></tt>. All sizes are in
+bits. The address space, <tt class="docutils literal"><span class="pre">n</span></tt> is optional, and if not specified,
+denotes the default address space 0. The value of <tt class="docutils literal"><span class="pre">n</span></tt> must be
+in the range [1,2^23).</dd>
+<dt><tt class="docutils literal"><span class="pre">i<size>:<abi>:<pref></span></tt></dt>
+<dd>This specifies the alignment for an integer type of a given bit
+<tt class="docutils literal"><span class="pre"><size></span></tt>. The value of <tt class="docutils literal"><span class="pre"><size></span></tt> must be in the range [1,2^23).</dd>
+<dt><tt class="docutils literal"><span class="pre">v<size>:<abi>:<pref></span></tt></dt>
+<dd>This specifies the alignment for a vector type of a given bit
+<tt class="docutils literal"><span class="pre"><size></span></tt>.</dd>
+<dt><tt class="docutils literal"><span class="pre">f<size>:<abi>:<pref></span></tt></dt>
+<dd>This specifies the alignment for a floating point type of a given bit
+<tt class="docutils literal"><span class="pre"><size></span></tt>. Only values of <tt class="docutils literal"><span class="pre"><size></span></tt> that are supported by the target
+will work. 32 (float) and 64 (double) are supported on all targets; 80
+or 128 (different flavors of long double) are also supported on some
+targets.</dd>
+<dt><tt class="docutils literal"><span class="pre">a:<abi>:<pref></span></tt></dt>
+<dd>This specifies the alignment for an object of aggregate type.</dd>
+<dt><tt class="docutils literal"><span class="pre">m:<mangling></span></tt></dt>
+<dd><p class="first">If present, specifies that llvm names are mangled in the output. The
+options are</p>
+<ul class="last simple">
+<li><tt class="docutils literal"><span class="pre">e</span></tt>: ELF mangling: Private symbols get a <tt class="docutils literal"><span class="pre">.L</span></tt> prefix.</li>
+<li><tt class="docutils literal"><span class="pre">m</span></tt>: Mips mangling: Private symbols get a <tt class="docutils literal"><span class="pre">$</span></tt> prefix.</li>
+<li><tt class="docutils literal"><span class="pre">o</span></tt>: Mach-O mangling: Private symbols get <tt class="docutils literal"><span class="pre">L</span></tt> prefix. Other
+symbols get a <tt class="docutils literal"><span class="pre">_</span></tt> prefix.</li>
+<li><tt class="docutils literal"><span class="pre">w</span></tt>: Windows COFF prefix: Similar to Mach-O, but stdcall and fastcall
+functions also get a suffix based on the frame size.</li>
+</ul>
+</dd>
+<dt><tt class="docutils literal"><span class="pre">n<size1>:<size2>:<size3>...</span></tt></dt>
+<dd>This specifies a set of native integer widths for the target CPU in
+bits. For example, it might contain <tt class="docutils literal"><span class="pre">n32</span></tt> for 32-bit PowerPC,
+<tt class="docutils literal"><span class="pre">n32:64</span></tt> for PowerPC 64, or <tt class="docutils literal"><span class="pre">n8:16:32:64</span></tt> for X86-64. Elements of
+this set are considered to support most general arithmetic operations
+efficiently.</dd>
+</dl>
+<p>On every specification that takes a <tt class="docutils literal"><span class="pre"><abi>:<pref></span></tt>, specifying the
+<tt class="docutils literal"><span class="pre"><pref></span></tt> alignment is optional. If omitted, the preceding <tt class="docutils literal"><span class="pre">:</span></tt>
+should be omitted too and <tt class="docutils literal"><span class="pre"><pref></span></tt> will be equal to <tt class="docutils literal"><span class="pre"><abi></span></tt>.</p>
+<p>When constructing the data layout for a given target, LLVM starts with a
+default set of specifications which are then (possibly) overridden by
+the specifications in the <tt class="docutils literal"><span class="pre">datalayout</span></tt> keyword. The default
+specifications are given in this list:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">E</span></tt> - big endian</li>
+<li><tt class="docutils literal"><span class="pre">p:64:64:64</span></tt> - 64-bit pointers with 64-bit alignment.</li>
+<li><tt class="docutils literal"><span class="pre">p[n]:64:64:64</span></tt> - Other address spaces are assumed to be the
+same as the default address space.</li>
+<li><tt class="docutils literal"><span class="pre">S0</span></tt> - natural stack alignment is unspecified</li>
+<li><tt class="docutils literal"><span class="pre">i1:8:8</span></tt> - i1 is 8-bit (byte) aligned</li>
+<li><tt class="docutils literal"><span class="pre">i8:8:8</span></tt> - i8 is 8-bit (byte) aligned</li>
+<li><tt class="docutils literal"><span class="pre">i16:16:16</span></tt> - i16 is 16-bit aligned</li>
+<li><tt class="docutils literal"><span class="pre">i32:32:32</span></tt> - i32 is 32-bit aligned</li>
+<li><tt class="docutils literal"><span class="pre">i64:32:64</span></tt> - i64 has ABI alignment of 32-bits but preferred
+alignment of 64-bits</li>
+<li><tt class="docutils literal"><span class="pre">f16:16:16</span></tt> - half is 16-bit aligned</li>
+<li><tt class="docutils literal"><span class="pre">f32:32:32</span></tt> - float is 32-bit aligned</li>
+<li><tt class="docutils literal"><span class="pre">f64:64:64</span></tt> - double is 64-bit aligned</li>
+<li><tt class="docutils literal"><span class="pre">f128:128:128</span></tt> - quad is 128-bit aligned</li>
+<li><tt class="docutils literal"><span class="pre">v64:64:64</span></tt> - 64-bit vector is 64-bit aligned</li>
+<li><tt class="docutils literal"><span class="pre">v128:128:128</span></tt> - 128-bit vector is 128-bit aligned</li>
+<li><tt class="docutils literal"><span class="pre">a:0:64</span></tt> - aggregates are 64-bit aligned</li>
+</ul>
+<p>When LLVM is determining the alignment for a given type, it uses the
+following rules:</p>
+<ol class="arabic simple">
+<li>If the type sought is an exact match for one of the specifications,
+that specification is used.</li>
+<li>If no match is found, and the type sought is an integer type, then
+the smallest integer type that is larger than the bitwidth of the
+sought type is used. If none of the specifications are larger than
+the bitwidth then the largest integer type is used. For example,
+given the default specifications above, the i7 type will use the
+alignment of i8 (next largest) while both i65 and i256 will use the
+alignment of i64 (largest specified).</li>
+<li>If no match is found, and the type sought is a vector type, then the
+largest vector type that is smaller than the sought vector type will
+be used as a fall back. This happens because <128 x double> can be
+implemented in terms of 64 <2 x double>, for example.</li>
+</ol>
+<p>The function of the data layout string may not be what you expect.
+Notably, this is not a specification from the frontend of what alignment
+the code generator should use.</p>
+<p>Instead, if specified, the target data layout is required to match what
+the ultimate <em>code generator</em> expects. This string is used by the
+mid-level optimizers to improve code, and this only works if it matches
+what the ultimate code generator uses. There is no way to generate IR
+that does not embed this target-specific detail into the IR. If you
+don’t specify the string, the default specifications will be used to
+generate a Data Layout and the optimization phases will operate
+accordingly and introduce target specificity into the IR with respect to
+these default specifications.</p>
+</div>
+<div class="section" id="target-triple">
+<span id="langref-triple"></span><h3><a class="toc-backref" href="#id633">Target Triple</a><a class="headerlink" href="#target-triple" title="Permalink to this headline">¶</a></h3>
+<p>A module may specify a target triple string that describes the target
+host. The syntax for the target triple is simply:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">target</span> <span class="k">triple</span> <span class="p">=</span> <span class="s">"x86_64-apple-macosx10.7.0"</span>
+</pre></div>
+</div>
+<p>The <em>target triple</em> string consists of a series of identifiers delimited
+by the minus sign character (‘-‘). The canonical forms are:</p>
+<div class="highlight-python"><div class="highlight"><pre><span class="n">ARCHITECTURE</span><span class="o">-</span><span class="n">VENDOR</span><span class="o">-</span><span class="n">OPERATING_SYSTEM</span>
+<span class="n">ARCHITECTURE</span><span class="o">-</span><span class="n">VENDOR</span><span class="o">-</span><span class="n">OPERATING_SYSTEM</span><span class="o">-</span><span class="n">ENVIRONMENT</span>
+</pre></div>
+</div>
+<p>This information is passed along to the backend so that it generates
+code for the proper architecture. It’s possible to override this on the
+command line with the <tt class="docutils literal"><span class="pre">-mtriple</span></tt> command line option.</p>
+</div>
+<div class="section" id="pointer-aliasing-rules">
+<span id="pointeraliasing"></span><h3><a class="toc-backref" href="#id634">Pointer Aliasing Rules</a><a class="headerlink" href="#pointer-aliasing-rules" title="Permalink to this headline">¶</a></h3>
+<p>Any memory access must be done through a pointer value associated with
+an address range of the memory access, otherwise the behavior is
+undefined. Pointer values are associated with address ranges according
+to the following rules:</p>
+<ul class="simple">
+<li>A pointer value is associated with the addresses associated with any
+value it is <em>based</em> on.</li>
+<li>An address of a global variable is associated with the address range
+of the variable’s storage.</li>
+<li>The result value of an allocation instruction is associated with the
+address range of the allocated storage.</li>
+<li>A null pointer in the default address-space is associated with no
+address.</li>
+<li>An integer constant other than zero or a pointer value returned from
+a function not defined within LLVM may be associated with address
+ranges allocated through mechanisms other than those provided by
+LLVM. Such ranges shall not overlap with any ranges of addresses
+allocated by mechanisms provided by LLVM.</li>
+</ul>
+<p>A pointer value is <em>based</em> on another pointer value according to the
+following rules:</p>
+<ul class="simple">
+<li>A pointer value formed from a <tt class="docutils literal"><span class="pre">getelementptr</span></tt> operation is <em>based</em>
+on the first value operand of the <tt class="docutils literal"><span class="pre">getelementptr</span></tt>.</li>
+<li>The result value of a <tt class="docutils literal"><span class="pre">bitcast</span></tt> is <em>based</em> on the operand of the
+<tt class="docutils literal"><span class="pre">bitcast</span></tt>.</li>
+<li>A pointer value formed by an <tt class="docutils literal"><span class="pre">inttoptr</span></tt> is <em>based</em> on all pointer
+values that contribute (directly or indirectly) to the computation of
+the pointer’s value.</li>
+<li>The “<em>based</em> on” relationship is transitive.</li>
+</ul>
+<p>Note that this definition of <em>“based”</em> is intentionally similar to the
+definition of <em>“based”</em> in C99, though it is slightly weaker.</p>
+<p>LLVM IR does not associate types with memory. The result type of a
+<tt class="docutils literal"><span class="pre">load</span></tt> merely indicates the size and alignment of the memory from
+which to load, as well as the interpretation of the value. The first
+operand type of a <tt class="docutils literal"><span class="pre">store</span></tt> similarly only indicates the size and
+alignment of the store.</p>
+<p>Consequently, type-based alias analysis, aka TBAA, aka
+<tt class="docutils literal"><span class="pre">-fstrict-aliasing</span></tt>, is not applicable to general unadorned LLVM IR.
+<a class="reference internal" href="#metadata"><em>Metadata</em></a> may be used to encode additional information
+which specialized optimization passes may use to implement type-based
+alias analysis.</p>
+</div>
+<div class="section" id="volatile-memory-accesses">
+<span id="volatile"></span><h3><a class="toc-backref" href="#id635">Volatile Memory Accesses</a><a class="headerlink" href="#volatile-memory-accesses" title="Permalink to this headline">¶</a></h3>
+<p>Certain memory accesses, such as <a class="reference internal" href="#i-load"><em>load</em></a>‘s,
+<a class="reference internal" href="#i-store"><em>store</em></a>‘s, and <a class="reference internal" href="#int-memcpy"><em>llvm.memcpy</em></a>‘s may be
+marked <tt class="docutils literal"><span class="pre">volatile</span></tt>. The optimizers must not change the number of
+volatile operations or change their order of execution relative to other
+volatile operations. The optimizers <em>may</em> change the order of volatile
+operations relative to non-volatile operations. This is not Java’s
+“volatile” and has no cross-thread synchronization behavior.</p>
+<p>IR-level volatile loads and stores cannot safely be optimized into
+llvm.memcpy or llvm.memmove intrinsics even when those intrinsics are
+flagged volatile. Likewise, the backend should never split or merge
+target-legal volatile load/store instructions.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Platforms may rely on volatile loads and stores of natively supported
+data width to be executed as single instruction. For example, in C
+this holds for an l-value of volatile primitive type with native
+hardware support, but not necessarily for aggregate types. The
+frontend upholds these expectations, which are intentionally
+unspecified in the IR. The rules above ensure that IR transformation
+do not violate the frontend’s contract with the language.</p>
+</div>
+</div>
+<div class="section" id="memory-model-for-concurrent-operations">
+<span id="memmodel"></span><h3><a class="toc-backref" href="#id636">Memory Model for Concurrent Operations</a><a class="headerlink" href="#memory-model-for-concurrent-operations" title="Permalink to this headline">¶</a></h3>
+<p>The LLVM IR does not define any way to start parallel threads of
+execution or to register signal handlers. Nonetheless, there are
+platform-specific ways to create them, and we define LLVM IR’s behavior
+in their presence. This model is inspired by the C++0x memory model.</p>
+<p>For a more informal introduction to this model, see the <a class="reference internal" href="Atomics.html"><em>LLVM Atomic Instructions and Concurrency Guide</em></a>.</p>
+<p>We define a <em>happens-before</em> partial order as the least partial order
+that</p>
+<ul class="simple">
+<li>Is a superset of single-thread program order, and</li>
+<li>When a <em>synchronizes-with</em> <tt class="docutils literal"><span class="pre">b</span></tt>, includes an edge from <tt class="docutils literal"><span class="pre">a</span></tt> to
+<tt class="docutils literal"><span class="pre">b</span></tt>. <em>Synchronizes-with</em> pairs are introduced by platform-specific
+techniques, like pthread locks, thread creation, thread joining,
+etc., and by atomic instructions. (See also <a class="reference internal" href="#ordering"><em>Atomic Memory Ordering
+Constraints</em></a>).</li>
+</ul>
+<p>Note that program order does not introduce <em>happens-before</em> edges
+between a thread and signals executing inside that thread.</p>
+<p>Every (defined) read operation (load instructions, memcpy, atomic
+loads/read-modify-writes, etc.) R reads a series of bytes written by
+(defined) write operations (store instructions, atomic
+stores/read-modify-writes, memcpy, etc.). For the purposes of this
+section, initialized globals are considered to have a write of the
+initializer which is atomic and happens before any other read or write
+of the memory in question. For each byte of a read R, R<sub>byte</sub>
+may see any write to the same byte, except:</p>
+<ul class="simple">
+<li>If write<sub>1</sub> happens before write<sub>2</sub>, and
+write<sub>2</sub> happens before R<sub>byte</sub>, then
+R<sub>byte</sub> does not see write<sub>1</sub>.</li>
+<li>If R<sub>byte</sub> happens before write<sub>3</sub>, then
+R<sub>byte</sub> does not see write<sub>3</sub>.</li>
+</ul>
+<p>Given that definition, R<sub>byte</sub> is defined as follows:</p>
+<ul class="simple">
+<li>If R is volatile, the result is target-dependent. (Volatile is
+supposed to give guarantees which can support <tt class="docutils literal"><span class="pre">sig_atomic_t</span></tt> in
+C/C++, and may be used for accesses to addresses that do not behave
+like normal memory. It does not generally provide cross-thread
+synchronization.)</li>
+<li>Otherwise, if there is no write to the same byte that happens before
+R<sub>byte</sub>, R<sub>byte</sub> returns <tt class="docutils literal"><span class="pre">undef</span></tt> for that byte.</li>
+<li>Otherwise, if R<sub>byte</sub> may see exactly one write,
+R<sub>byte</sub> returns the value written by that write.</li>
+<li>Otherwise, if R is atomic, and all the writes R<sub>byte</sub> may
+see are atomic, it chooses one of the values written. See the <a class="reference internal" href="#ordering"><em>Atomic
+Memory Ordering Constraints</em></a> section for additional
+constraints on how the choice is made.</li>
+<li>Otherwise R<sub>byte</sub> returns <tt class="docutils literal"><span class="pre">undef</span></tt>.</li>
+</ul>
+<p>R returns the value composed of the series of bytes it read. This
+implies that some bytes within the value may be <tt class="docutils literal"><span class="pre">undef</span></tt> <strong>without</strong>
+the entire value being <tt class="docutils literal"><span class="pre">undef</span></tt>. Note that this only defines the
+semantics of the operation; it doesn’t mean that targets will emit more
+than one instruction to read the series of bytes.</p>
+<p>Note that in cases where none of the atomic intrinsics are used, this
+model places only one restriction on IR transformations on top of what
+is required for single-threaded execution: introducing a store to a byte
+which might not otherwise be stored is not allowed in general.
+(Specifically, in the case where another thread might write to and read
+from an address, introducing a store can change a load that may see
+exactly one write into a load that may see multiple writes.)</p>
+</div>
+<div class="section" id="atomic-memory-ordering-constraints">
+<span id="ordering"></span><h3><a class="toc-backref" href="#id637">Atomic Memory Ordering Constraints</a><a class="headerlink" href="#atomic-memory-ordering-constraints" title="Permalink to this headline">¶</a></h3>
+<p>Atomic instructions (<a class="reference internal" href="#i-cmpxchg"><em>cmpxchg</em></a>,
+<a class="reference internal" href="#i-atomicrmw"><em>atomicrmw</em></a>, <a class="reference internal" href="#i-fence"><em>fence</em></a>,
+<a class="reference internal" href="#i-load"><em>atomic load</em></a>, and <a class="reference internal" href="#i-store"><em>atomic store</em></a>) take
+ordering parameters that determine which other atomic instructions on
+the same address they <em>synchronize with</em>. These semantics are borrowed
+from Java and C++0x, but are somewhat more colloquial. If these
+descriptions aren’t precise enough, check those specs (see spec
+references in the <a class="reference internal" href="Atomics.html"><em>atomics guide</em></a>).
+<a class="reference internal" href="#i-fence"><em>fence</em></a> instructions treat these orderings somewhat
+differently since they don’t take an address. See that instruction’s
+documentation for details.</p>
+<p>For a simpler introduction to the ordering constraints, see the
+<a class="reference internal" href="Atomics.html"><em>LLVM Atomic Instructions and Concurrency Guide</em></a>.</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">unordered</span></tt></dt>
+<dd>The set of values that can be read is governed by the happens-before
+partial order. A value cannot be read unless some operation wrote
+it. This is intended to provide a guarantee strong enough to model
+Java’s non-volatile shared variables. This ordering cannot be
+specified for read-modify-write operations; it is not strong enough
+to make them atomic in any interesting way.</dd>
+<dt><tt class="docutils literal"><span class="pre">monotonic</span></tt></dt>
+<dd>In addition to the guarantees of <tt class="docutils literal"><span class="pre">unordered</span></tt>, there is a single
+total order for modifications by <tt class="docutils literal"><span class="pre">monotonic</span></tt> operations on each
+address. All modification orders must be compatible with the
+happens-before order. There is no guarantee that the modification
+orders can be combined to a global total order for the whole program
+(and this often will not be possible). The read in an atomic
+read-modify-write operation (<a class="reference internal" href="#i-cmpxchg"><em>cmpxchg</em></a> and
+<a class="reference internal" href="#i-atomicrmw"><em>atomicrmw</em></a>) reads the value in the modification
+order immediately before the value it writes. If one atomic read
+happens before another atomic read of the same address, the later
+read must see the same value or a later value in the address’s
+modification order. This disallows reordering of <tt class="docutils literal"><span class="pre">monotonic</span></tt> (or
+stronger) operations on the same address. If an address is written
+<tt class="docutils literal"><span class="pre">monotonic</span></tt>-ally by one thread, and other threads <tt class="docutils literal"><span class="pre">monotonic</span></tt>-ally
+read that address repeatedly, the other threads must eventually see
+the write. This corresponds to the C++0x/C1x
+<tt class="docutils literal"><span class="pre">memory_order_relaxed</span></tt>.</dd>
+<dt><tt class="docutils literal"><span class="pre">acquire</span></tt></dt>
+<dd>In addition to the guarantees of <tt class="docutils literal"><span class="pre">monotonic</span></tt>, a
+<em>synchronizes-with</em> edge may be formed with a <tt class="docutils literal"><span class="pre">release</span></tt> operation.
+This is intended to model C++’s <tt class="docutils literal"><span class="pre">memory_order_acquire</span></tt>.</dd>
+<dt><tt class="docutils literal"><span class="pre">release</span></tt></dt>
+<dd>In addition to the guarantees of <tt class="docutils literal"><span class="pre">monotonic</span></tt>, if this operation
+writes a value which is subsequently read by an <tt class="docutils literal"><span class="pre">acquire</span></tt>
+operation, it <em>synchronizes-with</em> that operation. (This isn’t a
+complete description; see the C++0x definition of a release
+sequence.) This corresponds to the C++0x/C1x
+<tt class="docutils literal"><span class="pre">memory_order_release</span></tt>.</dd>
+<dt><tt class="docutils literal"><span class="pre">acq_rel</span></tt> (acquire+release)</dt>
+<dd>Acts as both an <tt class="docutils literal"><span class="pre">acquire</span></tt> and <tt class="docutils literal"><span class="pre">release</span></tt> operation on its
+address. This corresponds to the C++0x/C1x <tt class="docutils literal"><span class="pre">memory_order_acq_rel</span></tt>.</dd>
+<dt><tt class="docutils literal"><span class="pre">seq_cst</span></tt> (sequentially consistent)</dt>
+<dd>In addition to the guarantees of <tt class="docutils literal"><span class="pre">acq_rel</span></tt> (<tt class="docutils literal"><span class="pre">acquire</span></tt> for an
+operation that only reads, <tt class="docutils literal"><span class="pre">release</span></tt> for an operation that only
+writes), there is a global total order on all
+sequentially-consistent operations on all addresses, which is
+consistent with the <em>happens-before</em> partial order and with the
+modification orders of all the affected addresses. Each
+sequentially-consistent read sees the last preceding write to the
+same address in this global order. This corresponds to the C++0x/C1x
+<tt class="docutils literal"><span class="pre">memory_order_seq_cst</span></tt> and Java volatile.</dd>
+</dl>
+<p id="singlethread">If an atomic operation is marked <tt class="docutils literal"><span class="pre">singlethread</span></tt>, it only <em>synchronizes
+with</em> or participates in modification and seq_cst total orderings with
+other operations running in the same thread (for example, in signal
+handlers).</p>
+</div>
+<div class="section" id="fast-math-flags">
+<span id="fastmath"></span><h3><a class="toc-backref" href="#id638">Fast-Math Flags</a><a class="headerlink" href="#fast-math-flags" title="Permalink to this headline">¶</a></h3>
+<p>LLVM IR floating-point binary ops (<a class="reference internal" href="#i-fadd"><em>fadd</em></a>,
+<a class="reference internal" href="#i-fsub"><em>fsub</em></a>, <a class="reference internal" href="#i-fmul"><em>fmul</em></a>, <a class="reference internal" href="#i-fdiv"><em>fdiv</em></a>,
+<a class="reference internal" href="#i-frem"><em>frem</em></a>, <a class="reference internal" href="#i-fcmp"><em>fcmp</em></a>) have the following flags that can
+be set to enable otherwise unsafe floating point operations</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">nnan</span></tt></dt>
+<dd>No NaNs - Allow optimizations to assume the arguments and result are not
+NaN. Such optimizations are required to retain defined behavior over
+NaNs, but the value of the result is undefined.</dd>
+<dt><tt class="docutils literal"><span class="pre">ninf</span></tt></dt>
+<dd>No Infs - Allow optimizations to assume the arguments and result are not
++/-Inf. Such optimizations are required to retain defined behavior over
++/-Inf, but the value of the result is undefined.</dd>
+<dt><tt class="docutils literal"><span class="pre">nsz</span></tt></dt>
+<dd>No Signed Zeros - Allow optimizations to treat the sign of a zero
+argument or result as insignificant.</dd>
+<dt><tt class="docutils literal"><span class="pre">arcp</span></tt></dt>
+<dd>Allow Reciprocal - Allow optimizations to use the reciprocal of an
+argument rather than perform division.</dd>
+<dt><tt class="docutils literal"><span class="pre">fast</span></tt></dt>
+<dd>Fast - Allow algebraically equivalent transformations that may
+dramatically change results in floating point (e.g. reassociate). This
+flag implies all the others.</dd>
+</dl>
+</div>
+<div class="section" id="use-list-order-directives">
+<span id="uselistorder"></span><h3><a class="toc-backref" href="#id639">Use-list Order Directives</a><a class="headerlink" href="#use-list-order-directives" title="Permalink to this headline">¶</a></h3>
+<p>Use-list directives encode the in-memory order of each use-list, allowing the
+order to be recreated. <tt class="docutils literal"><span class="pre"><order-indexes></span></tt> is a comma-separated list of
+indexes that are assigned to the referenced value’s uses. The referenced
+value’s use-list is immediately sorted by these indexes.</p>
+<p>Use-list directives may appear at function scope or global scope. They are not
+instructions, and have no effect on the semantics of the IR. When they’re at
+function scope, they must appear after the terminator of the final basic block.</p>
+<p>If basic blocks have their address taken via <tt class="docutils literal"><span class="pre">blockaddress()</span></tt> expressions,
+<tt class="docutils literal"><span class="pre">uselistorder_bb</span></tt> can be used to reorder their use-lists from outside their
+function’s scope.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre>uselistorder <ty> <value>, { <order-indexes> }
+uselistorder_bb @function, %block { <order-indexes> }
+</pre></div>
+</div>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Examples:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre>define void @foo(i32 %arg1, i32 %arg2) {
+entry:
+ ; ... instructions ...
+bb:
+ ; ... instructions ...
+
+ ; At function scope.
+ uselistorder i32 %arg1, { 1, 0, 2 }
+ uselistorder label %bb, { 1, 0 }
+}
+
+; At global scope.
+uselistorder i32* @global, { 1, 2, 0 }
+uselistorder i32 7, { 1, 0 }
+uselistorder i32 (i32) @bar, { 1, 0 }
+uselistorder_bb @foo, %bb, { 5, 1, 3, 2, 0, 4 }
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="type-system">
+<span id="typesystem"></span><h2><a class="toc-backref" href="#id640">Type System</a><a class="headerlink" href="#type-system" title="Permalink to this headline">¶</a></h2>
+<p>The LLVM type system is one of the most important features of the
+intermediate representation. Being typed enables a number of
+optimizations to be performed on the intermediate representation
+directly, without having to do extra analyses on the side before the
+transformation. A strong type system makes it easier to read the
+generated code and enables novel analyses and transformations that are
+not feasible to perform on normal three address code representations.</p>
+<div class="section" id="void-type">
+<span id="t-void"></span><h3><a class="toc-backref" href="#id641">Void Type</a><a class="headerlink" href="#void-type" title="Permalink to this headline">¶</a></h3>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The void type does not represent any value and has no size.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre><span class="n">void</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="function-type">
+<span id="t-function"></span><h3><a class="toc-backref" href="#id642">Function Type</a><a class="headerlink" href="#function-type" title="Permalink to this headline">¶</a></h3>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The function type can be thought of as a function signature. It consists of a
+return type and a list of formal parameter types. The return type of a function
+type is a void type or first class type — except for <a class="reference internal" href="#t-label"><em>label</em></a>
+and <a class="reference internal" href="#t-metadata"><em>metadata</em></a> types.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre><returntype> (<parameter list>)
+</pre></div>
+</div>
+<p>...where ‘<tt class="docutils literal"><span class="pre"><parameter</span> <span class="pre">list></span></tt>‘ is a comma-separated list of type
+specifiers. Optionally, the parameter list may include a type <tt class="docutils literal"><span class="pre">...</span></tt>, which
+indicates that the function takes a variable number of arguments. Variable
+argument functions can access their arguments with the <a class="reference internal" href="#int-varargs"><em>variable argument
+handling intrinsic</em></a> functions. ‘<tt class="docutils literal"><span class="pre"><returntype></span></tt>‘ is any type
+except <a class="reference internal" href="#t-label"><em>label</em></a> and <a class="reference internal" href="#t-metadata"><em>metadata</em></a>.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Examples:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<table border="1" class="docutils">
+<colgroup>
+<col width="17%" />
+<col width="83%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">i32</span> <span class="pre">(i32)</span></tt></td>
+<td>function taking an <tt class="docutils literal"><span class="pre">i32</span></tt>, returning an <tt class="docutils literal"><span class="pre">i32</span></tt></td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">float</span> <span class="pre">(i16,</span> <span class="pre">i32</span> <span class="pre">*)</span> <span class="pre">*</span></tt></td>
+<td><a class="reference internal" href="#t-pointer"><em>Pointer</em></a> to a function that takes an <tt class="docutils literal"><span class="pre">i16</span></tt> and a <a class="reference internal" href="#t-pointer"><em>pointer</em></a> to <tt class="docutils literal"><span class="pre">i32</span></tt>, returning <tt class="docutils literal"><span class="pre">float</span></tt>.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">i32</span> <span class="pre">(i8*,</span> <span class="pre">...)</span></tt></td>
+<td>A vararg function that takes at least one <a class="reference internal" href="#t-pointer"><em>pointer</em></a> to <tt class="docutils literal"><span class="pre">i8</span></tt> (char in C), which returns an integer. This is the signature for <tt class="docutils literal"><span class="pre">printf</span></tt> in LLVM.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">{i32,</span> <span class="pre">i32}</span> <span class="pre">(i32)</span></tt></td>
+<td>A function taking an <tt class="docutils literal"><span class="pre">i32</span></tt>, returning a <a class="reference internal" href="#t-struct"><em>structure</em></a> containing two <tt class="docutils literal"><span class="pre">i32</span></tt> values</td>
+</tr>
+</tbody>
+</table>
+</div>
+<div class="section" id="first-class-types">
+<span id="t-firstclass"></span><h3><a class="toc-backref" href="#id643">First Class Types</a><a class="headerlink" href="#first-class-types" title="Permalink to this headline">¶</a></h3>
+<p>The <a class="reference internal" href="#t-firstclass"><em>first class</em></a> types are perhaps the most important.
+Values of these types are the only ones which can be produced by
+instructions.</p>
+<div class="section" id="single-value-types">
+<span id="t-single-value"></span><h4><a class="toc-backref" href="#id644">Single Value Types</a><a class="headerlink" href="#single-value-types" title="Permalink to this headline">¶</a></h4>
+<p>These are the types that are valid in registers from CodeGen’s perspective.</p>
+<div class="section" id="integer-type">
+<span id="t-integer"></span><h5><a class="toc-backref" href="#id645">Integer Type</a><a class="headerlink" href="#integer-type" title="Permalink to this headline">¶</a></h5>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The integer type is a very simple type that simply specifies an
+arbitrary bit width for the integer type desired. Any bit width from 1
+bit to 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre><span class="n">iN</span>
+</pre></div>
+</div>
+<p>The number of bits the integer will occupy is specified by the <tt class="docutils literal"><span class="pre">N</span></tt>
+value.</p>
+<div class="section" id="examples">
+<h6>Examples:<a class="headerlink" href="#examples" title="Permalink to this headline">¶</a></h6>
+<table border="1" class="docutils">
+<colgroup>
+<col width="25%" />
+<col width="75%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">i1</span></tt></td>
+<td>a single-bit integer.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">i32</span></tt></td>
+<td>a 32-bit integer.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">i1942652</span></tt></td>
+<td>a really big integer of over 1 million bits.</td>
+</tr>
+</tbody>
+</table>
+</div>
+</div>
+<div class="section" id="floating-point-types">
+<span id="t-floating"></span><h5><a class="toc-backref" href="#id646">Floating Point Types</a><a class="headerlink" href="#floating-point-types" title="Permalink to this headline">¶</a></h5>
+<table border="1" class="docutils">
+<colgroup>
+<col width="50%" />
+<col width="50%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">Type</th>
+<th class="head">Description</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">half</span></tt></td>
+<td>16-bit floating point value</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">float</span></tt></td>
+<td>32-bit floating point value</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">double</span></tt></td>
+<td>64-bit floating point value</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">fp128</span></tt></td>
+<td>128-bit floating point value (112-bit mantissa)</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">x86_fp80</span></tt></td>
+<td>80-bit floating point value (X87)</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">ppc_fp128</span></tt></td>
+<td>128-bit floating point value (two 64-bits)</td>
+</tr>
+</tbody>
+</table>
+</div>
+<div class="section" id="x86-mmx-type">
+<h5><a class="toc-backref" href="#id647">X86_mmx Type</a><a class="headerlink" href="#x86-mmx-type" title="Permalink to this headline">¶</a></h5>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The x86_mmx type represents a value held in an MMX register on an x86
+machine. The operations allowed on it are quite limited: parameters and
+return values, load and store, and bitcast. User-specified MMX
+instructions are represented as intrinsic or asm calls with arguments
+and/or results of this type. There are no arrays, vectors or constants
+of this type.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre><span class="n">x86_mmx</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="pointer-type">
+<span id="t-pointer"></span><h5><a class="toc-backref" href="#id648">Pointer Type</a><a class="headerlink" href="#pointer-type" title="Permalink to this headline">¶</a></h5>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The pointer type is used to specify memory locations. Pointers are
+commonly used to reference objects in memory.</p>
+<p>Pointer types may have an optional address space attribute defining the
+numbered address space where the pointed-to object resides. The default
+address space is number zero. The semantics of non-zero address spaces
+are target-specific.</p>
+<p>Note that LLVM does not permit pointers to void (<tt class="docutils literal"><span class="pre">void*</span></tt>) nor does it
+permit pointers to labels (<tt class="docutils literal"><span class="pre">label*</span></tt>). Use <tt class="docutils literal"><span class="pre">i8*</span></tt> instead.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre><type> *
+</pre></div>
+</div>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Examples:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<table border="1" class="docutils">
+<colgroup>
+<col width="19%" />
+<col width="81%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">[4</span> <span class="pre">x</span> <span class="pre">i32]*</span></tt></td>
+<td>A <a class="reference internal" href="#t-pointer"><em>pointer</em></a> to <a class="reference internal" href="#t-array"><em>array</em></a> of four <tt class="docutils literal"><span class="pre">i32</span></tt> values.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">i32</span> <span class="pre">(i32*)</span> <span class="pre">*</span></tt></td>
+<td>A <a class="reference internal" href="#t-pointer"><em>pointer</em></a> to a <a class="reference internal" href="#t-function"><em>function</em></a> that takes an <tt class="docutils literal"><span class="pre">i32*</span></tt>, returning an <tt class="docutils literal"><span class="pre">i32</span></tt>.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">i32</span> <span class="pre">addrspace(5)*</span></tt></td>
+<td>A <a class="reference internal" href="#t-pointer"><em>pointer</em></a> to an <tt class="docutils literal"><span class="pre">i32</span></tt> value that resides in address space #5.</td>
+</tr>
+</tbody>
+</table>
+</div>
+<div class="section" id="vector-type">
+<span id="t-vector"></span><h5><a class="toc-backref" href="#id649">Vector Type</a><a class="headerlink" href="#vector-type" title="Permalink to this headline">¶</a></h5>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>A vector type is a simple derived type that represents a vector of
+elements. Vector types are used when multiple primitive data are
+operated in parallel using a single instruction (SIMD). A vector type
+requires a size (number of elements) and an underlying primitive data
+type. Vector types are considered <a class="reference internal" href="#t-firstclass"><em>first class</em></a>.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre>< <# elements> x <elementtype> >
+</pre></div>
+</div>
+<p>The number of elements is a constant integer value larger than 0;
+elementtype may be any integer, floating point or pointer type. Vectors
+of size zero are not allowed.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Examples:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<table border="1" class="docutils">
+<colgroup>
+<col width="28%" />
+<col width="72%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre"><4</span> <span class="pre">x</span> <span class="pre">i32></span></tt></td>
+<td>Vector of 4 32-bit integer values.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre"><8</span> <span class="pre">x</span> <span class="pre">float></span></tt></td>
+<td>Vector of 8 32-bit floating-point values.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre"><2</span> <span class="pre">x</span> <span class="pre">i64></span></tt></td>
+<td>Vector of 2 64-bit integer values.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre"><4</span> <span class="pre">x</span> <span class="pre">i64*></span></tt></td>
+<td>Vector of 4 pointers to 64-bit integer values.</td>
+</tr>
+</tbody>
+</table>
+</div>
+</div>
+<div class="section" id="label-type">
+<span id="t-label"></span><h4><a class="toc-backref" href="#id650">Label Type</a><a class="headerlink" href="#label-type" title="Permalink to this headline">¶</a></h4>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The label type represents code labels.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre><span class="n">label</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="metadata-type">
+<span id="t-metadata"></span><h4><a class="toc-backref" href="#id651">Metadata Type</a><a class="headerlink" href="#metadata-type" title="Permalink to this headline">¶</a></h4>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The metadata type represents embedded metadata. No derived types may be
+created from metadata except for <a class="reference internal" href="#t-function"><em>function</em></a> arguments.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre><span class="n">metadata</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="aggregate-types">
+<span id="t-aggregate"></span><h4><a class="toc-backref" href="#id652">Aggregate Types</a><a class="headerlink" href="#aggregate-types" title="Permalink to this headline">¶</a></h4>
+<p>Aggregate Types are a subset of derived types that can contain multiple
+member types. <a class="reference internal" href="#t-array"><em>Arrays</em></a> and <a class="reference internal" href="#t-struct"><em>structs</em></a> are
+aggregate types. <a class="reference internal" href="#t-vector"><em>Vectors</em></a> are not considered to be
+aggregate types.</p>
+<div class="section" id="array-type">
+<span id="t-array"></span><h5><a class="toc-backref" href="#id653">Array Type</a><a class="headerlink" href="#array-type" title="Permalink to this headline">¶</a></h5>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The array type is a very simple derived type that arranges elements
+sequentially in memory. The array type requires a size (number of
+elements) and an underlying data type.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre>[<# elements> x <elementtype>]
+</pre></div>
+</div>
+<p>The number of elements is a constant integer value; <tt class="docutils literal"><span class="pre">elementtype</span></tt> may
+be any type with a size.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Examples:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<table border="1" class="docutils">
+<colgroup>
+<col width="32%" />
+<col width="68%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">[40</span> <span class="pre">x</span> <span class="pre">i32]</span></tt></td>
+<td>Array of 40 32-bit integer values.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">[41</span> <span class="pre">x</span> <span class="pre">i32]</span></tt></td>
+<td>Array of 41 32-bit integer values.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">[4</span> <span class="pre">x</span> <span class="pre">i8]</span></tt></td>
+<td>Array of 4 8-bit integer values.</td>
+</tr>
+</tbody>
+</table>
+<p>Here are some examples of multidimensional arrays:</p>
+<table border="1" class="docutils">
+<colgroup>
+<col width="33%" />
+<col width="67%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">[3</span> <span class="pre">x</span> <span class="pre">[4</span> <span class="pre">x</span> <span class="pre">i32]]</span></tt></td>
+<td>3x4 array of 32-bit integer values.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">[12</span> <span class="pre">x</span> <span class="pre">[10</span> <span class="pre">x</span> <span class="pre">float]]</span></tt></td>
+<td>12x10 array of single precision floating point values.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">[2</span> <span class="pre">x</span> <span class="pre">[3</span> <span class="pre">x</span> <span class="pre">[4</span> <span class="pre">x</span> <span class="pre">i16]]]</span></tt></td>
+<td>2x3x4 array of 16-bit integer values.</td>
+</tr>
+</tbody>
+</table>
+<p>There is no restriction on indexing beyond the end of the array implied
+by a static type (though there are restrictions on indexing beyond the
+bounds of an allocated object in some cases). This means that
+single-dimension ‘variable sized array’ addressing can be implemented in
+LLVM with a zero length array type. An implementation of ‘pascal style
+arrays’ in LLVM could use the type “<tt class="docutils literal"><span class="pre">{</span> <span class="pre">i32,</span> <span class="pre">[0</span> <span class="pre">x</span> <span class="pre">float]}</span></tt>”, for
+example.</p>
+</div>
+<div class="section" id="structure-type">
+<span id="t-struct"></span><h5><a class="toc-backref" href="#id654">Structure Type</a><a class="headerlink" href="#structure-type" title="Permalink to this headline">¶</a></h5>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>The structure type is used to represent a collection of data members
+together in memory. The elements of a structure may be any type that has
+a size.</p>
+<p>Structures in memory are accessed using ‘<tt class="docutils literal"><span class="pre">load</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">store</span></tt>‘ by
+getting a pointer to a field with the ‘<tt class="docutils literal"><span class="pre">getelementptr</span></tt>‘ instruction.
+Structures in registers are accessed using the ‘<tt class="docutils literal"><span class="pre">extractvalue</span></tt>‘ and
+‘<tt class="docutils literal"><span class="pre">insertvalue</span></tt>‘ instructions.</p>
+<p>Structures may optionally be “packed” structures, which indicate that
+the alignment of the struct is one byte, and that there is no padding
+between the elements. In non-packed structs, padding between field types
+is inserted as defined by the DataLayout string in the module, which is
+required to match what the underlying code generator expects.</p>
+<p>Structures can either be “literal” or “identified”. A literal structure
+is defined inline with other types (e.g. <tt class="docutils literal"><span class="pre">{i32,</span> <span class="pre">i32}*</span></tt>) whereas
+identified types are always defined at the top level with a name.
+Literal types are uniqued by their contents and can never be recursive
+or opaque since there is no way to write one. Identified types can be
+recursive, can be opaqued, and are never uniqued.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre>%T1 = type { <type list> } ; Identified normal struct type
+%T2 = type <{ <type list> }> ; Identified packed struct type
+</pre></div>
+</div>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Examples:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<table border="1" class="docutils">
+<colgroup>
+<col width="14%" />
+<col width="86%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">{</span> <span class="pre">i32,</span> <span class="pre">i32,</span> <span class="pre">i32</span> <span class="pre">}</span></tt></td>
+<td>A triple of three <tt class="docutils literal"><span class="pre">i32</span></tt> values</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">{</span> <span class="pre">float,</span> <span class="pre">i32</span> <span class="pre">(i32)</span> <span class="pre">*</span> <span class="pre">}</span></tt></td>
+<td>A pair, where the first element is a <tt class="docutils literal"><span class="pre">float</span></tt> and the second element is a <a class="reference internal" href="#t-pointer"><em>pointer</em></a> to a <a class="reference internal" href="#t-function"><em>function</em></a> that takes an <tt class="docutils literal"><span class="pre">i32</span></tt>, returning an <tt class="docutils literal"><span class="pre">i32</span></tt>.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre"><{</span> <span class="pre">i8,</span> <span class="pre">i32</span> <span class="pre">}></span></tt></td>
+<td>A packed struct known to be 5 bytes in size.</td>
+</tr>
+</tbody>
+</table>
+</div>
+<div class="section" id="opaque-structure-types">
+<span id="t-opaque"></span><h5><a class="toc-backref" href="#id655">Opaque Structure Types</a><a class="headerlink" href="#opaque-structure-types" title="Permalink to this headline">¶</a></h5>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Overview:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<p>Opaque structure types are used to represent named structure types that
+do not have a body specified. This corresponds (for example) to the C
+notion of a forward declared structure.</p>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Syntax:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<div class="highlight-python"><div class="highlight"><pre>%X = type opaque
+%52 = type opaque
+</pre></div>
+</div>
+<table class="docutils field-list" frame="void" rules="none">
+<col class="field-name" />
+<col class="field-body" />
+<tbody valign="top">
+<tr class="field-odd field"><th class="field-name">Examples:</th><td class="field-body"></td>
+</tr>
+</tbody>
+</table>
+<table border="1" class="docutils">
+<colgroup>
+<col width="42%" />
+<col width="58%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">opaque</span></tt></td>
+<td>An opaque type.</td>
+</tr>
+</tbody>
+</table>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="constants">
+<span id="id3"></span><h2><a class="toc-backref" href="#id656">Constants</a><a class="headerlink" href="#constants" title="Permalink to this headline">¶</a></h2>
+<p>LLVM has several different basic types of constants. This section
+describes them all and their syntax.</p>
+<div class="section" id="simple-constants">
+<h3><a class="toc-backref" href="#id657">Simple Constants</a><a class="headerlink" href="#simple-constants" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>Boolean constants</strong></dt>
+<dd>The two strings ‘<tt class="docutils literal"><span class="pre">true</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">false</span></tt>‘ are both valid constants
+of the <tt class="docutils literal"><span class="pre">i1</span></tt> type.</dd>
+<dt><strong>Integer constants</strong></dt>
+<dd>Standard integers (such as ‘4’) are constants of the
+<a class="reference internal" href="#t-integer"><em>integer</em></a> type. Negative numbers may be used with
+integer types.</dd>
+<dt><strong>Floating point constants</strong></dt>
+<dd>Floating point constants use standard decimal notation (e.g.
+123.421), exponential notation (e.g. 1.23421e+2), or a more precise
+hexadecimal notation (see below). The assembler requires the exact
+decimal value of a floating-point constant. For example, the
+assembler accepts 1.25 but rejects 1.3 because 1.3 is a repeating
+decimal in binary. Floating point constants must have a <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> type.</dd>
+<dt><strong>Null pointer constants</strong></dt>
+<dd>The identifier ‘<tt class="docutils literal"><span class="pre">null</span></tt>‘ is recognized as a null pointer constant
+and must be of <a class="reference internal" href="#t-pointer"><em>pointer type</em></a>.</dd>
+</dl>
+<p>The one non-intuitive notation for constants is the hexadecimal form of
+floating point constants. For example, the form
+‘<tt class="docutils literal"><span class="pre">double</span> <span class="pre">0x432ff973cafa8000</span></tt>‘ is equivalent to (but harder to read
+than) ‘<tt class="docutils literal"><span class="pre">double</span> <span class="pre">4.5e+15</span></tt>‘. The only time hexadecimal floating point
+constants are required (and the only time that they are generated by the
+disassembler) is when a floating point constant must be emitted but it
+cannot be represented as a decimal floating point number in a reasonable
+number of digits. For example, NaN’s, infinities, and other special
+values are represented in their IEEE hexadecimal format so that assembly
+and disassembly do not cause any bits to change in the constants.</p>
+<p>When using the hexadecimal form, constants of types half, float, and
+double are represented using the 16-digit form shown above (which
+matches the IEEE754 representation for double); half and float values
+must, however, be exactly representable as IEEE 754 half and single
+precision, respectively. Hexadecimal format is always used for long
+double, and there are three forms of long double. The 80-bit format used
+by x86 is represented as <tt class="docutils literal"><span class="pre">0xK</span></tt> followed by 20 hexadecimal digits. The
+128-bit format used by PowerPC (two adjacent doubles) is represented by
+<tt class="docutils literal"><span class="pre">0xM</span></tt> followed by 32 hexadecimal digits. The IEEE 128-bit format is
+represented by <tt class="docutils literal"><span class="pre">0xL</span></tt> followed by 32 hexadecimal digits. Long doubles
+will only work if they match the long double format on your target.
+The IEEE 16-bit format (half precision) is represented by <tt class="docutils literal"><span class="pre">0xH</span></tt>
+followed by 4 hexadecimal digits. All hexadecimal formats are big-endian
+(sign bit at the left).</p>
+<p>There are no constants of type x86_mmx.</p>
+</div>
+<div class="section" id="complex-constants">
+<span id="complexconstants"></span><h3><a class="toc-backref" href="#id658">Complex Constants</a><a class="headerlink" href="#complex-constants" title="Permalink to this headline">¶</a></h3>
+<p>Complex constants are a (potentially recursive) combination of simple
+constants and smaller complex constants.</p>
+<dl class="docutils">
+<dt><strong>Structure constants</strong></dt>
+<dd>Structure constants are represented with notation similar to
+structure type definitions (a comma separated list of elements,
+surrounded by braces (<tt class="docutils literal"><span class="pre">{}</span></tt>)). For example:
+“<tt class="docutils literal"><span class="pre">{</span> <span class="pre">i32</span> <span class="pre">4,</span> <span class="pre">float</span> <span class="pre">17.0,</span> <span class="pre">i32*</span> <span class="pre">@G</span> <span class="pre">}</span></tt>”, where “<tt class="docutils literal"><span class="pre">@G</span></tt>” is declared as
+“<tt class="docutils literal"><span class="pre">@G</span> <span class="pre">=</span> <span class="pre">external</span> <span class="pre">global</span> <span class="pre">i32</span></tt>”. Structure constants must have
+<a class="reference internal" href="#t-struct"><em>structure type</em></a>, and the number and types of elements
+must match those specified by the type.</dd>
+<dt><strong>Array constants</strong></dt>
+<dd>Array constants are represented with notation similar to array type
+definitions (a comma separated list of elements, surrounded by
+square brackets (<tt class="docutils literal"><span class="pre">[]</span></tt>)). For example:
+“<tt class="docutils literal"><span class="pre">[</span> <span class="pre">i32</span> <span class="pre">42,</span> <span class="pre">i32</span> <span class="pre">11,</span> <span class="pre">i32</span> <span class="pre">74</span> <span class="pre">]</span></tt>”. Array constants must have
+<a class="reference internal" href="#t-array"><em>array type</em></a>, and the number and types of elements must
+match those specified by the type. As a special case, character array
+constants may also be represented as a double-quoted string using the <tt class="docutils literal"><span class="pre">c</span></tt>
+prefix. For example: “<tt class="docutils literal"><span class="pre">c"Hello</span> <span class="pre">World\0A\00"</span></tt>”.</dd>
+<dt><strong>Vector constants</strong></dt>
+<dd>Vector constants are represented with notation similar to vector
+type definitions (a comma separated list of elements, surrounded by
+less-than/greater-than’s (<tt class="docutils literal"><span class="pre"><></span></tt>)). For example:
+“<tt class="docutils literal"><span class="pre"><</span> <span class="pre">i32</span> <span class="pre">42,</span> <span class="pre">i32</span> <span class="pre">11,</span> <span class="pre">i32</span> <span class="pre">74,</span> <span class="pre">i32</span> <span class="pre">100</span> <span class="pre">></span></tt>”. Vector constants
+must have <a class="reference internal" href="#t-vector"><em>vector type</em></a>, and the number and types of
+elements must match those specified by the type.</dd>
+<dt><strong>Zero initialization</strong></dt>
+<dd>The string ‘<tt class="docutils literal"><span class="pre">zeroinitializer</span></tt>‘ can be used to zero initialize a
+value to zero of <em>any</em> type, including scalar and
+<a class="reference internal" href="#t-aggregate"><em>aggregate</em></a> types. This is often used to avoid
+having to print large zero initializers (e.g. for large arrays) and
+is always exactly equivalent to using explicit zero initializers.</dd>
+<dt><strong>Metadata node</strong></dt>
+<dd>A metadata node is a constant tuple without types. For example:
+“<tt class="docutils literal"><span class="pre">!{!0,</span> <span class="pre">!{!2,</span> <span class="pre">!0},</span> <span class="pre">!"test"}</span></tt>”. Metadata can reference constant values,
+for example: “<tt class="docutils literal"><span class="pre">!{!0,</span> <span class="pre">i32</span> <span class="pre">0,</span> <span class="pre">i8*</span> <span class="pre">@global,</span> <span class="pre">i64</span> <span class="pre">(i64)*</span> <span class="pre">@function,</span> <span class="pre">!"str"}</span></tt>”.
+Unlike other typed constants that are meant to be interpreted as part of
+the instruction stream, metadata is a place to attach additional
+information such as debug info.</dd>
+</dl>
+</div>
+<div class="section" id="global-variable-and-function-addresses">
+<h3><a class="toc-backref" href="#id659">Global Variable and Function Addresses</a><a class="headerlink" href="#global-variable-and-function-addresses" title="Permalink to this headline">¶</a></h3>
+<p>The addresses of <a class="reference internal" href="#globalvars"><em>global variables</em></a> and
+<a class="reference internal" href="#functionstructure"><em>functions</em></a> are always implicitly valid
+(link-time) constants. These constants are explicitly referenced when
+the <a class="reference internal" href="#identifiers"><em>identifier for the global</em></a> is used and always have
+<a class="reference internal" href="#t-pointer"><em>pointer</em></a> type. For example, the following is a legal LLVM
+file:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="vg">@X</span> <span class="p">=</span> <span class="k">global</span> <span class="k">i32</span> <span class="m">17</span>
+<span class="vg">@Y</span> <span class="p">=</span> <span class="k">global</span> <span class="k">i32</span> <span class="m">42</span>
+<span class="vg">@Z</span> <span class="p">=</span> <span class="k">global</span> <span class="p">[</span><span class="m">2</span> <span class="k">x</span> <span class="k">i32</span><span class="p">*]</span> <span class="p">[</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@X</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@Y</span> <span class="p">]</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="undefined-values">
+<span id="undefvalues"></span><h3><a class="toc-backref" href="#id660">Undefined Values</a><a class="headerlink" href="#undefined-values" title="Permalink to this headline">¶</a></h3>
+<p>The string ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘ can be used anywhere a constant is expected, and
+indicates that the user of the value may receive an unspecified
+bit-pattern. Undefined values may be of any type (other than ‘<tt class="docutils literal"><span class="pre">label</span></tt>‘
+or ‘<tt class="docutils literal"><span class="pre">void</span></tt>‘) and be used anywhere a constant is permitted.</p>
+<p>Undefined values are useful because they indicate to the compiler that
+the program is well defined no matter what value is used. This gives the
+compiler more freedom to optimize. Here are some examples of
+(potentially surprising) transformations that are valid (in pseudo IR):</p>
+<div class="highlight-llvm"><div class="highlight"><pre> <span class="nv">%A</span> <span class="p">=</span> <span class="k">add</span> <span class="nv">%X</span><span class="p">,</span> <span class="k">undef</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="k">sub</span> <span class="nv">%X</span><span class="p">,</span> <span class="k">undef</span>
+ <span class="nv">%C</span> <span class="p">=</span> <span class="k">xor</span> <span class="nv">%X</span><span class="p">,</span> <span class="k">undef</span>
+<span class="nl">Safe:</span>
+ <span class="nv">%A</span> <span class="p">=</span> <span class="k">undef</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="k">undef</span>
+ <span class="nv">%C</span> <span class="p">=</span> <span class="k">undef</span>
+</pre></div>
+</div>
+<p>This is safe because all of the output bits are affected by the undef
+bits. Any output bit can have a zero or one depending on the input bits.</p>
+<div class="highlight-llvm"><div class="highlight"><pre> <span class="nv">%A</span> <span class="p">=</span> <span class="k">or</span> <span class="nv">%X</span><span class="p">,</span> <span class="k">undef</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="k">and</span> <span class="nv">%X</span><span class="p">,</span> <span class="k">undef</span>
+<span class="nl">Safe:</span>
+ <span class="nv">%A</span> <span class="p">=</span> <span class="m">-1</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="m">0</span>
+<span class="nl">Unsafe:</span>
+ <span class="nv">%A</span> <span class="p">=</span> <span class="k">undef</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="k">undef</span>
+</pre></div>
+</div>
+<p>These logical operations have bits that are not always affected by the
+input. For example, if <tt class="docutils literal"><span class="pre">%X</span></tt> has a zero bit, then the output of the
+‘<tt class="docutils literal"><span class="pre">and</span></tt>‘ operation will always be a zero for that bit, no matter what
+the corresponding bit from the ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘ is. As such, it is unsafe to
+optimize or assume that the result of the ‘<tt class="docutils literal"><span class="pre">and</span></tt>‘ is ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘.
+However, it is safe to assume that all bits of the ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘ could be
+0, and optimize the ‘<tt class="docutils literal"><span class="pre">and</span></tt>‘ to 0. Likewise, it is safe to assume that
+all the bits of the ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘ operand to the ‘<tt class="docutils literal"><span class="pre">or</span></tt>‘ could be set,
+allowing the ‘<tt class="docutils literal"><span class="pre">or</span></tt>‘ to be folded to -1.</p>
+<div class="highlight-llvm"><div class="highlight"><pre> <span class="nv">%A</span> <span class="p">=</span> <span class="k">select</span> <span class="k">undef</span><span class="p">,</span> <span class="nv">%X</span><span class="p">,</span> <span class="nv">%Y</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="k">select</span> <span class="k">undef</span><span class="p">,</span> <span class="m">42</span><span class="p">,</span> <span class="nv">%Y</span>
+ <span class="nv">%C</span> <span class="p">=</span> <span class="k">select</span> <span class="nv">%X</span><span class="p">,</span> <span class="nv">%Y</span><span class="p">,</span> <span class="k">undef</span>
+<span class="nl">Safe:</span>
+ <span class="nv">%A</span> <span class="p">=</span> <span class="nv">%X</span> <span class="p">(</span><span class="k">or</span> <span class="nv">%Y</span><span class="p">)</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="m">42</span> <span class="p">(</span><span class="k">or</span> <span class="nv">%Y</span><span class="p">)</span>
+ <span class="nv">%C</span> <span class="p">=</span> <span class="nv">%Y</span>
+<span class="nl">Unsafe:</span>
+ <span class="nv">%A</span> <span class="p">=</span> <span class="k">undef</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="k">undef</span>
+ <span class="nv">%C</span> <span class="p">=</span> <span class="k">undef</span>
+</pre></div>
+</div>
+<p>This set of examples shows that undefined ‘<tt class="docutils literal"><span class="pre">select</span></tt>‘ (and conditional
+branch) conditions can go <em>either way</em>, but they have to come from one
+of the two operands. In the <tt class="docutils literal"><span class="pre">%A</span></tt> example, if <tt class="docutils literal"><span class="pre">%X</span></tt> and <tt class="docutils literal"><span class="pre">%Y</span></tt> were
+both known to have a clear low bit, then <tt class="docutils literal"><span class="pre">%A</span></tt> would have to have a
+cleared low bit. However, in the <tt class="docutils literal"><span class="pre">%C</span></tt> example, the optimizer is
+allowed to assume that the ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘ operand could be the same as
+<tt class="docutils literal"><span class="pre">%Y</span></tt>, allowing the whole ‘<tt class="docutils literal"><span class="pre">select</span></tt>‘ to be eliminated.</p>
+<div class="highlight-llvm"><div class="highlight"><pre> %A = xor undef, undef
+
+ %B = undef
+ %C = xor %B, %B
+
+ %D = undef
+ %E = icmp slt %D, 4
+ %F = icmp gte %D, 4
+
+Safe:
+ %A = undef
+ %B = undef
+ %C = undef
+ %D = undef
+ %E = undef
+ %F = undef
+</pre></div>
+</div>
+<p>This example points out that two ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘ operands are not
+necessarily the same. This can be surprising to people (and also matches
+C semantics) where they assume that “<tt class="docutils literal"><span class="pre">X^X</span></tt>” is always zero, even if
+<tt class="docutils literal"><span class="pre">X</span></tt> is undefined. This isn’t true for a number of reasons, but the
+short answer is that an ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘ “variable” can arbitrarily change
+its value over its “live range”. This is true because the variable
+doesn’t actually <em>have a live range</em>. Instead, the value is logically
+read from arbitrary registers that happen to be around when needed, so
+the value is not necessarily consistent over time. In fact, <tt class="docutils literal"><span class="pre">%A</span></tt> and
+<tt class="docutils literal"><span class="pre">%C</span></tt> need to have the same semantics or the core LLVM “replace all
+uses with” concept would not hold.</p>
+<div class="highlight-llvm"><div class="highlight"><pre> <span class="nv">%A</span> <span class="p">=</span> <span class="k">fdiv</span> <span class="k">undef</span><span class="p">,</span> <span class="nv">%X</span>
+ <span class="nv">%B</span> <span class="p">=</span> <span class="k">fdiv</span> <span class="nv">%X</span><span class="p">,</span> <span class="k">undef</span>
+<span class="nl">Safe:</span>
+ <span class="nv">%A</span> <span class="p">=</span> <span class="k">undef</span>
+<span class="nl">b:</span> <span class="k">unreachable</span>
+</pre></div>
+</div>
+<p>These examples show the crucial difference between an <em>undefined value</em>
+and <em>undefined behavior</em>. An undefined value (like ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘) is
+allowed to have an arbitrary bit-pattern. This means that the <tt class="docutils literal"><span class="pre">%A</span></tt>
+operation can be constant folded to ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘, because the ‘<tt class="docutils literal"><span class="pre">undef</span></tt>‘
+could be an SNaN, and <tt class="docutils literal"><span class="pre">fdiv</span></tt> is not (currently) defined on SNaN’s.
+However, in the second example, we can make a more aggressive
+assumption: because the <tt class="docutils literal"><span class="pre">undef</span></tt> is allowed to be an arbitrary value,
+we are allowed to assume that it could be zero. Since a divide by zero
+has <em>undefined behavior</em>, we are allowed to assume that the operation
+does not execute at all. This allows us to delete the divide and all
+code after it. Because the undefined operation “can’t happen”, the
+optimizer can assume that it occurs in dead code.</p>
+<div class="highlight-llvm"><div class="highlight"><pre>a: store undef -> %X
+b: store %X -> undef
+Safe:
+a: <deleted>
+b: unreachable
+</pre></div>
+</div>
+<p>These examples reiterate the <tt class="docutils literal"><span class="pre">fdiv</span></tt> example: a store <em>of</em> an undefined
+value can be assumed to not have any effect; we can assume that the
+value is overwritten with bits that happen to match what was already
+there. However, a store <em>to</em> an undefined location could clobber
+arbitrary memory, therefore, it has undefined behavior.</p>
+</div>
+<div class="section" id="poison-values">
+<span id="poisonvalues"></span><h3><a class="toc-backref" href="#id661">Poison Values</a><a class="headerlink" href="#poison-values" title="Permalink to this headline">¶</a></h3>
+<p>Poison values are similar to <a class="reference internal" href="#undefvalues"><em>undef values</em></a>, however
+they also represent the fact that an instruction or constant expression
+that cannot evoke side effects has nevertheless detected a condition
+that results in undefined behavior.</p>
+<p>There is currently no way of representing a poison value in the IR; they
+only exist when produced by operations such as <a class="reference internal" href="#i-add"><em>add</em></a> with
+the <tt class="docutils literal"><span class="pre">nsw</span></tt> flag.</p>
+<p>Poison value behavior is defined in terms of value <em>dependence</em>:</p>
+<ul class="simple">
+<li>Values other than <a class="reference internal" href="#i-phi"><em>phi</em></a> nodes depend on their operands.</li>
+<li><a class="reference internal" href="#i-phi"><em>Phi</em></a> nodes depend on the operand corresponding to
+their dynamic predecessor basic block.</li>
+<li>Function arguments depend on the corresponding actual argument values
+in the dynamic callers of their functions.</li>
+<li><a class="reference internal" href="#i-call"><em>Call</em></a> instructions depend on the <a class="reference internal" href="#i-ret"><em>ret</em></a>
+instructions that dynamically transfer control back to them.</li>
+<li><a class="reference internal" href="#i-invoke"><em>Invoke</em></a> instructions depend on the
+<a class="reference internal" href="#i-ret"><em>ret</em></a>, <a class="reference internal" href="#i-resume"><em>resume</em></a>, or exception-throwing
+call instructions that dynamically transfer control back to them.</li>
+<li>Non-volatile loads and stores depend on the most recent stores to all
+of the referenced memory addresses, following the order in the IR
+(including loads and stores implied by intrinsics such as
+<a class="reference internal" href="#int-memcpy"><em>@llvm.memcpy</em></a>.)</li>
+<li>An instruction with externally visible side effects depends on the
+most recent preceding instruction with externally visible side
+effects, following the order in the IR. (This includes <a class="reference internal" href="#volatile"><em>volatile
+operations</em></a>.)</li>
+<li>An instruction <em>control-depends</em> on a <a class="reference internal" href="#terminators"><em>terminator
+instruction</em></a> if the terminator instruction has
+multiple successors and the instruction is always executed when
+control transfers to one of the successors, and may not be executed
+when control is transferred to another.</li>
+<li>Additionally, an instruction also <em>control-depends</em> on a terminator
+instruction if the set of instructions it otherwise depends on would
+be different if the terminator had transferred control to a different
+successor.</li>
+<li>Dependence is transitive.</li>
+</ul>
+<p>Poison values have the same behavior as <a class="reference internal" href="#undefvalues"><em>undef values</em></a>,
+with the additional effect that any instruction that has a <em>dependence</em>
+on a poison value has undefined behavior.</p>
+<p>Here are some examples:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nl">entry:</span>
+ <span class="nv">%poison</span> <span class="p">=</span> <span class="k">sub</span> <span class="k">nuw</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="m">1</span> <span class="c">; Results in a poison value.</span>
+ <span class="nv">%still_poison</span> <span class="p">=</span> <span class="k">and</span> <span class="k">i32</span> <span class="nv">%poison</span><span class="p">,</span> <span class="m">0</span> <span class="c">; 0, but also poison.</span>
+ <span class="nv">%poison_yet_again</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@h</span><span class="p">,</span> <span class="k">i32</span> <span class="nv">%still_poison</span>
+ <span class="k">store</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%poison_yet_again</span> <span class="c">; memory at @h[0] is poisoned</span>
+
+ <span class="k">store</span> <span class="k">i32</span> <span class="nv">%poison</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@g</span> <span class="c">; Poison value stored to memory.</span>
+ <span class="nv">%poison2</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@g</span> <span class="c">; Poison value loaded back from memory.</span>
+
+ <span class="k">store</span> <span class="k">volatile</span> <span class="k">i32</span> <span class="nv">%poison</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@g</span> <span class="c">; External observation; undefined behavior.</span>
+
+ <span class="nv">%narrowaddr</span> <span class="p">=</span> <span class="k">bitcast</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@g</span> <span class="k">to</span> <span class="k">i16</span><span class="p">*</span>
+ <span class="nv">%wideaddr</span> <span class="p">=</span> <span class="k">bitcast</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@g</span> <span class="k">to</span> <span class="k">i64</span><span class="p">*</span>
+ <span class="nv">%poison3</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i16</span><span class="p">,</span> <span class="k">i16</span><span class="p">*</span> <span class="nv">%narrowaddr</span> <span class="c">; Returns a poison value.</span>
+ <span class="nv">%poison4</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i64</span><span class="p">,</span> <span class="k">i64</span><span class="p">*</span> <span class="nv">%wideaddr</span> <span class="c">; Returns a poison value.</span>
+
+ <span class="nv">%cmp</span> <span class="p">=</span> <span class="k">icmp</span> <span class="k">slt</span> <span class="k">i32</span> <span class="nv">%poison</span><span class="p">,</span> <span class="m">0</span> <span class="c">; Returns a poison value.</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%cmp</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%true</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%end</span> <span class="c">; Branch to either destination.</span>
+
+<span class="nl">true:</span>
+ <span class="k">store</span> <span class="k">volatile</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@g</span> <span class="c">; This is control-dependent on %cmp, so</span>
+ <span class="c">; it has undefined behavior.</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%end</span>
+
+<span class="nl">end:</span>
+ <span class="nv">%p</span> <span class="p">=</span> <span class="k">phi</span> <span class="k">i32</span> <span class="p">[</span> <span class="m">0</span><span class="p">,</span> <span class="nv">%entry</span> <span class="p">],</span> <span class="p">[</span> <span class="m">1</span><span class="p">,</span> <span class="nv">%true</span> <span class="p">]</span>
+ <span class="c">; Both edges into this PHI are</span>
+ <span class="c">; control-dependent on %cmp, so this</span>
+ <span class="c">; always results in a poison value.</span>
+
+ <span class="k">store</span> <span class="k">volatile</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@g</span> <span class="c">; This would depend on the store in %true</span>
+ <span class="c">; if %cmp is true, or the store in %entry</span>
+ <span class="c">; otherwise, so this is undefined behavior.</span>
+
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%cmp</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%second_true</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%second_end</span>
+ <span class="c">; The same branch again, but this time the</span>
+ <span class="c">; true block doesn't have side effects.</span>
+
+<span class="nl">second_true:</span>
+ <span class="c">; No side effects!</span>
+ <span class="k">ret</span> <span class="kt">void</span>
+
+<span class="nl">second_end:</span>
+ <span class="k">store</span> <span class="k">volatile</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@g</span> <span class="c">; This time, the instruction always depends</span>
+ <span class="c">; on the store in %end. Also, it is</span>
+ <span class="c">; control-equivalent to %end, so this is</span>
+ <span class="c">; well-defined (ignoring earlier undefined</span>
+ <span class="c">; behavior in this example).</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="addresses-of-basic-blocks">
+<span id="blockaddress"></span><h3><a class="toc-backref" href="#id662">Addresses of Basic Blocks</a><a class="headerlink" href="#addresses-of-basic-blocks" title="Permalink to this headline">¶</a></h3>
+<p><tt class="docutils literal"><span class="pre">blockaddress(@function,</span> <span class="pre">%block)</span></tt></p>
+<p>The ‘<tt class="docutils literal"><span class="pre">blockaddress</span></tt>‘ constant computes the address of the specified
+basic block in the specified function, and always has an <tt class="docutils literal"><span class="pre">i8*</span></tt> type.
+Taking the address of the entry block is illegal.</p>
+<p>This value only has defined behavior when used as an operand to the
+‘<a class="reference internal" href="#i-indirectbr"><em>indirectbr</em></a>‘ instruction, or for comparisons
+against null. Pointer equality tests between labels addresses results in
+undefined behavior — though, again, comparison against null is ok, and
+no label is equal to the null pointer. This may be passed around as an
+opaque pointer sized value as long as the bits are not inspected. This
+allows <tt class="docutils literal"><span class="pre">ptrtoint</span></tt> and arithmetic to be performed on these values so
+long as the original value is reconstituted before the <tt class="docutils literal"><span class="pre">indirectbr</span></tt>
+instruction.</p>
+<p>Finally, some targets may provide defined semantics when using the value
+as the operand to an inline assembly, but that is target specific.</p>
+</div>
+<div class="section" id="constant-expressions">
+<span id="constantexprs"></span><h3><a class="toc-backref" href="#id663">Constant Expressions</a><a class="headerlink" href="#constant-expressions" title="Permalink to this headline">¶</a></h3>
+<p>Constant expressions are used to allow expressions involving other
+constants to be used as constants. Constant expressions may be of any
+<a class="reference internal" href="#t-firstclass"><em>first class</em></a> type and may involve any LLVM operation
+that does not have side effects (e.g. load and call are not supported).
+The following is the syntax for constant expressions:</p>
+<dl class="docutils">
+<dt><tt class="docutils literal"><span class="pre">trunc</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Truncate a constant to another type. The bit size of CST must be
+larger than the bit size of TYPE. Both types must be integers.</dd>
+<dt><tt class="docutils literal"><span class="pre">zext</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Zero extend a constant to another type. The bit size of CST must be
+smaller than the bit size of TYPE. Both types must be integers.</dd>
+<dt><tt class="docutils literal"><span class="pre">sext</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Sign extend a constant to another type. The bit size of CST must be
+smaller than the bit size of TYPE. Both types must be integers.</dd>
+<dt><tt class="docutils literal"><span class="pre">fptrunc</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Truncate a floating point constant to another floating point type.
+The size of CST must be larger than the size of TYPE. Both types
+must be floating point.</dd>
+<dt><tt class="docutils literal"><span class="pre">fpext</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Floating point extend a constant to another type. The size of CST
+must be smaller or equal to the size of TYPE. Both types must be
+floating point.</dd>
+<dt><tt class="docutils literal"><span class="pre">fptoui</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Convert a floating point constant to the corresponding unsigned
+integer constant. TYPE must be a scalar or vector integer type. CST
+must be of scalar or vector floating point type. Both CST and TYPE
+must be scalars, or vectors of the same number of elements. If the
+value won’t fit in the integer type, the results are undefined.</dd>
+<dt><tt class="docutils literal"><span class="pre">fptosi</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Convert a floating point constant to the corresponding signed
+integer constant. TYPE must be a scalar or vector integer type. CST
+must be of scalar or vector floating point type. Both CST and TYPE
+must be scalars, or vectors of the same number of elements. If the
+value won’t fit in the integer type, the results are undefined.</dd>
+<dt><tt class="docutils literal"><span class="pre">uitofp</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Convert an unsigned integer constant to the corresponding floating
+point constant. TYPE must be a scalar or vector floating point type.
+CST must be of scalar or vector integer type. Both CST and TYPE must
+be scalars, or vectors of the same number of elements. If the value
+won’t fit in the floating point type, the results are undefined.</dd>
+<dt><tt class="docutils literal"><span class="pre">sitofp</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Convert a signed integer constant to the corresponding floating
+point constant. TYPE must be a scalar or vector floating point type.
+CST must be of scalar or vector integer type. Both CST and TYPE must
+be scalars, or vectors of the same number of elements. If the value
+won’t fit in the floating point type, the results are undefined.</dd>
+<dt><tt class="docutils literal"><span class="pre">ptrtoint</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Convert a pointer typed constant to the corresponding integer
+constant. <tt class="docutils literal"><span class="pre">TYPE</span></tt> must be an integer type. <tt class="docutils literal"><span class="pre">CST</span></tt> must be of
+pointer type. The <tt class="docutils literal"><span class="pre">CST</span></tt> value is zero extended, truncated, or
+unchanged to make it fit in <tt class="docutils literal"><span class="pre">TYPE</span></tt>.</dd>
+<dt><tt class="docutils literal"><span class="pre">inttoptr</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Convert an integer constant to a pointer constant. TYPE must be a
+pointer type. CST must be of integer type. The CST value is zero
+extended, truncated, or unchanged to make it fit in a pointer size.
+This one is <em>really</em> dangerous!</dd>
+<dt><tt class="docutils literal"><span class="pre">bitcast</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Convert a constant, CST, to another TYPE. The constraints of the
+operands are the same as those for the <a class="reference internal" href="#i-bitcast"><em>bitcast
+instruction</em></a>.</dd>
+<dt><tt class="docutils literal"><span class="pre">addrspacecast</span> <span class="pre">(CST</span> <span class="pre">to</span> <span class="pre">TYPE)</span></tt></dt>
+<dd>Convert a constant pointer or constant vector of pointer, CST, to another
+TYPE in a different address space. The constraints of the operands are the
+same as those for the <a class="reference internal" href="#i-addrspacecast"><em>addrspacecast instruction</em></a>.</dd>
+<dt><tt class="docutils literal"><span class="pre">getelementptr</span> <span class="pre">(TY,</span> <span class="pre">CSTPTR,</span> <span class="pre">IDX0,</span> <span class="pre">IDX1,</span> <span class="pre">...)</span></tt>, <tt class="docutils literal"><span class="pre">getelementptr</span> <span class="pre">inbounds</span> <span class="pre">(TY,</span> <span class="pre">CSTPTR,</span> <span class="pre">IDX0,</span> <span class="pre">IDX1,</span> <span class="pre">...)</span></tt></dt>
+<dd>Perform the <a class="reference internal" href="#i-getelementptr"><em>getelementptr operation</em></a> on
+constants. As with the <a class="reference internal" href="#i-getelementptr"><em>getelementptr</em></a>
+instruction, the index list may have zero or more indexes, which are
+required to make sense for the type of “pointer to TY”.</dd>
+<dt><tt class="docutils literal"><span class="pre">select</span> <span class="pre">(COND,</span> <span class="pre">VAL1,</span> <span class="pre">VAL2)</span></tt></dt>
+<dd>Perform the <a class="reference internal" href="#i-select"><em>select operation</em></a> on constants.</dd>
+<dt><tt class="docutils literal"><span class="pre">icmp</span> <span class="pre">COND</span> <span class="pre">(VAL1,</span> <span class="pre">VAL2)</span></tt></dt>
+<dd>Performs the <a class="reference internal" href="#i-icmp"><em>icmp operation</em></a> on constants.</dd>
+<dt><tt class="docutils literal"><span class="pre">fcmp</span> <span class="pre">COND</span> <span class="pre">(VAL1,</span> <span class="pre">VAL2)</span></tt></dt>
+<dd>Performs the <a class="reference internal" href="#i-fcmp"><em>fcmp operation</em></a> on constants.</dd>
+<dt><tt class="docutils literal"><span class="pre">extractelement</span> <span class="pre">(VAL,</span> <span class="pre">IDX)</span></tt></dt>
+<dd>Perform the <a class="reference internal" href="#i-extractelement"><em>extractelement operation</em></a> on
+constants.</dd>
+<dt><tt class="docutils literal"><span class="pre">insertelement</span> <span class="pre">(VAL,</span> <span class="pre">ELT,</span> <span class="pre">IDX)</span></tt></dt>
+<dd>Perform the <a class="reference internal" href="#i-insertelement"><em>insertelement operation</em></a> on
+constants.</dd>
+<dt><tt class="docutils literal"><span class="pre">shufflevector</span> <span class="pre">(VEC1,</span> <span class="pre">VEC2,</span> <span class="pre">IDXMASK)</span></tt></dt>
+<dd>Perform the <a class="reference internal" href="#i-shufflevector"><em>shufflevector operation</em></a> on
+constants.</dd>
+<dt><tt class="docutils literal"><span class="pre">extractvalue</span> <span class="pre">(VAL,</span> <span class="pre">IDX0,</span> <span class="pre">IDX1,</span> <span class="pre">...)</span></tt></dt>
+<dd>Perform the <a class="reference internal" href="#i-extractvalue"><em>extractvalue operation</em></a> on
+constants. The index list is interpreted in a similar manner as
+indices in a ‘<a class="reference internal" href="#i-getelementptr"><em>getelementptr</em></a>‘ operation. At
+least one index value must be specified.</dd>
+<dt><tt class="docutils literal"><span class="pre">insertvalue</span> <span class="pre">(VAL,</span> <span class="pre">ELT,</span> <span class="pre">IDX0,</span> <span class="pre">IDX1,</span> <span class="pre">...)</span></tt></dt>
+<dd>Perform the <a class="reference internal" href="#i-insertvalue"><em>insertvalue operation</em></a> on constants.
+The index list is interpreted in a similar manner as indices in a
+‘<a class="reference internal" href="#i-getelementptr"><em>getelementptr</em></a>‘ operation. At least one index
+value must be specified.</dd>
+<dt><tt class="docutils literal"><span class="pre">OPCODE</span> <span class="pre">(LHS,</span> <span class="pre">RHS)</span></tt></dt>
+<dd>Perform the specified operation of the LHS and RHS constants. OPCODE
+may be any of the <a class="reference internal" href="#binaryops"><em>binary</em></a> or <a class="reference internal" href="#bitwiseops"><em>bitwise
+binary</em></a> operations. The constraints on operands are
+the same as those for the corresponding instruction (e.g. no bitwise
+operations on floating point values are allowed).</dd>
+</dl>
+</div>
+</div>
+<div class="section" id="other-values">
+<h2><a class="toc-backref" href="#id664">Other Values</a><a class="headerlink" href="#other-values" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="inline-assembler-expressions">
+<span id="inlineasmexprs"></span><h3><a class="toc-backref" href="#id665">Inline Assembler Expressions</a><a class="headerlink" href="#inline-assembler-expressions" title="Permalink to this headline">¶</a></h3>
+<p>LLVM supports inline assembler expressions (as opposed to <a class="reference internal" href="#moduleasm"><em>Module-Level
+Inline Assembly</em></a>) through the use of a special value. This value
+represents the inline assembler as a template string (containing the
+instructions to emit), a list of operand constraints (stored as a string), a
+flag that indicates whether or not the inline asm expression has side effects,
+and a flag indicating whether the function containing the asm needs to align its
+stack conservatively.</p>
+<p>The template string supports argument substitution of the operands using “<tt class="docutils literal"><span class="pre">$</span></tt>”
+followed by a number, to indicate substitution of the given register/memory
+location, as specified by the constraint string. “<tt class="docutils literal"><span class="pre">${NUM:MODIFIER}</span></tt>” may also
+be used, where <tt class="docutils literal"><span class="pre">MODIFIER</span></tt> is a target-specific annotation for how to print the
+operand (See <a class="reference internal" href="#inline-asm-modifiers"><em>Asm template argument modifiers</em></a>).</p>
+<p>A literal “<tt class="docutils literal"><span class="pre">$</span></tt>” may be included by using “<tt class="docutils literal"><span class="pre">$$</span></tt>” in the template. To include
+other special characters into the output, the usual “<tt class="docutils literal"><span class="pre">\XX</span></tt>” escapes may be
+used, just as in other strings. Note that after template substitution, the
+resulting assembly string is parsed by LLVM’s integrated assembler unless it is
+disabled – even when emitting a <tt class="docutils literal"><span class="pre">.s</span></tt> file – and thus must contain assembly
+syntax known to LLVM.</p>
+<p>LLVM’s support for inline asm is modeled closely on the requirements of Clang’s
+GCC-compatible inline-asm support. Thus, the feature-set and the constraint and
+modifier codes listed here are similar or identical to those in GCC’s inline asm
+support. However, to be clear, the syntax of the template and constraint strings
+described here is <em>not</em> the same as the syntax accepted by GCC and Clang, and,
+while most constraint letters are passed through as-is by Clang, some get
+translated to other codes when converting from the C source to the LLVM
+assembly.</p>
+<p>An example inline assembler expression is:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">i32</span> <span class="p">(</span><span class="k">i32</span><span class="p">)</span> <span class="k">asm</span> <span class="s">"bswap $0"</span><span class="p">,</span> <span class="s">"=r,r"</span>
+</pre></div>
+</div>
+<p>Inline assembler expressions may <strong>only</strong> be used as the callee operand
+of a <a class="reference internal" href="#i-call"><em>call</em></a> or an <a class="reference internal" href="#i-invoke"><em>invoke</em></a> instruction.
+Thus, typically we have:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">call</span> <span class="k">i32</span> <span class="k">asm</span> <span class="s">"bswap $0"</span><span class="p">,</span> <span class="s">"=r,r"</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%Y</span><span class="p">)</span>
+</pre></div>
+</div>
+<p>Inline asms with side effects not visible in the constraint list must be
+marked as having side effects. This is done through the use of the
+‘<tt class="docutils literal"><span class="pre">sideeffect</span></tt>‘ keyword, like so:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">call</span> <span class="kt">void</span> <span class="k">asm</span> <span class="k">sideeffect</span> <span class="s">"eieio"</span><span class="p">,</span> <span class="s">""</span><span class="p">()</span>
+</pre></div>
+</div>
+<p>In some cases inline asms will contain code that will not work unless
+the stack is aligned in some way, such as calls or SSE instructions on
+x86, yet will not contain code that does that alignment within the asm.
+The compiler should make conservative assumptions about what the asm
+might contain and should generate its usual stack alignment code in the
+prologue if the ‘<tt class="docutils literal"><span class="pre">alignstack</span></tt>‘ keyword is present:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>call void asm alignstack "eieio", ""()
+</pre></div>
+</div>
+<p>Inline asms also support using non-standard assembly dialects. The
+assumed dialect is ATT. When the ‘<tt class="docutils literal"><span class="pre">inteldialect</span></tt>‘ keyword is present,
+the inline asm is using the Intel dialect. Currently, ATT and Intel are
+the only supported dialects. An example is:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>call void asm inteldialect "eieio", ""()
+</pre></div>
+</div>
+<p>If multiple keywords appear the ‘<tt class="docutils literal"><span class="pre">sideeffect</span></tt>‘ keyword must come
+first, the ‘<tt class="docutils literal"><span class="pre">alignstack</span></tt>‘ keyword second and the ‘<tt class="docutils literal"><span class="pre">inteldialect</span></tt>‘
+keyword last.</p>
+<div class="section" id="inline-asm-constraint-string">
+<h4><a class="toc-backref" href="#id666">Inline Asm Constraint String</a><a class="headerlink" href="#inline-asm-constraint-string" title="Permalink to this headline">¶</a></h4>
+<p>The constraint list is a comma-separated string, each element containing one or
+more constraint codes.</p>
+<p>For each element in the constraint list an appropriate register or memory
+operand will be chosen, and it will be made available to assembly template
+string expansion as <tt class="docutils literal"><span class="pre">$0</span></tt> for the first constraint in the list, <tt class="docutils literal"><span class="pre">$1</span></tt> for the
+second, etc.</p>
+<p>There are three different types of constraints, which are distinguished by a
+prefix symbol in front of the constraint code: Output, Input, and Clobber. The
+constraints must always be given in that order: outputs first, then inputs, then
+clobbers. They cannot be intermingled.</p>
+<p>There are also three different categories of constraint codes:</p>
+<ul class="simple">
+<li>Register constraint. This is either a register class, or a fixed physical
+register. This kind of constraint will allocate a register, and if necessary,
+bitcast the argument or result to the appropriate type.</li>
+<li>Memory constraint. This kind of constraint is for use with an instruction
+taking a memory operand. Different constraints allow for different addressing
+modes used by the target.</li>
+<li>Immediate value constraint. This kind of constraint is for an integer or other
+immediate value which can be rendered directly into an instruction. The
+various target-specific constraints allow the selection of a value in the
+proper range for the instruction you wish to use it with.</li>
+</ul>
+<div class="section" id="output-constraints">
+<h5><a class="toc-backref" href="#id667">Output constraints</a><a class="headerlink" href="#output-constraints" title="Permalink to this headline">¶</a></h5>
+<p>Output constraints are specified by an “<tt class="docutils literal"><span class="pre">=</span></tt>” prefix (e.g. “<tt class="docutils literal"><span class="pre">=r</span></tt>”). This
+indicates that the assembly will write to this operand, and the operand will
+then be made available as a return value of the <tt class="docutils literal"><span class="pre">asm</span></tt> expression. Output
+constraints do not consume an argument from the call instruction. (Except, see
+below about indirect outputs).</p>
+<p>Normally, it is expected that no output locations are written to by the assembly
+expression until <em>all</em> of the inputs have been read. As such, LLVM may assign
+the same register to an output and an input. If this is not safe (e.g. if the
+assembly contains two instructions, where the first writes to one output, and
+the second reads an input and writes to a second output), then the “<tt class="docutils literal"><span class="pre">&</span></tt>”
+modifier must be used (e.g. “<tt class="docutils literal"><span class="pre">=&r</span></tt>”) to specify that the output is an
+“early-clobber” output. Marking an ouput as “early-clobber” ensures that LLVM
+will not use the same register for any inputs (other than an input tied to this
+output).</p>
+</div>
+<div class="section" id="input-constraints">
+<h5><a class="toc-backref" href="#id668">Input constraints</a><a class="headerlink" href="#input-constraints" title="Permalink to this headline">¶</a></h5>
+<p>Input constraints do not have a prefix – just the constraint codes. Each input
+constraint will consume one argument from the call instruction. It is not
+permitted for the asm to write to any input register or memory location (unless
+that input is tied to an output). Note also that multiple inputs may all be
+assigned to the same register, if LLVM can determine that they necessarily all
+contain the same value.</p>
+<p>Instead of providing a Constraint Code, input constraints may also “tie”
+themselves to an output constraint, by providing an integer as the constraint
+string. Tied inputs still consume an argument from the call instruction, and
+take up a position in the asm template numbering as is usual – they will simply
+be constrained to always use the same register as the output they’ve been tied
+to. For example, a constraint string of “<tt class="docutils literal"><span class="pre">=r,0</span></tt>” says to assign a register for
+output, and use that register as an input as well (it being the 0’th
+constraint).</p>
+<p>It is permitted to tie an input to an “early-clobber” output. In that case, no
+<em>other</em> input may share the same register as the input tied to the early-clobber
+(even when the other input has the same value).</p>
+<p>You may only tie an input to an output which has a register constraint, not a
+memory constraint. Only a single input may be tied to an output.</p>
+<p>There is also an “interesting” feature which deserves a bit of explanation: if a
+register class constraint allocates a register which is too small for the value
+type operand provided as input, the input value will be split into multiple
+registers, and all of them passed to the inline asm.</p>
+<p>However, this feature is often not as useful as you might think.</p>
+<p>Firstly, the registers are <em>not</em> guaranteed to be consecutive. So, on those
+architectures that have instructions which operate on multiple consecutive
+instructions, this is not an appropriate way to support them. (e.g. the 32-bit
+SparcV8 has a 64-bit load, which instruction takes a single 32-bit register. The
+hardware then loads into both the named register, and the next register. This
+feature of inline asm would not be useful to support that.)</p>
+<p>A few of the targets provide a template string modifier allowing explicit access
+to the second register of a two-register operand (e.g. MIPS <tt class="docutils literal"><span class="pre">L</span></tt>, <tt class="docutils literal"><span class="pre">M</span></tt>, and
+<tt class="docutils literal"><span class="pre">D</span></tt>). On such an architecture, you can actually access the second allocated
+register (yet, still, not any subsequent ones). But, in that case, you’re still
+probably better off simply splitting the value into two separate operands, for
+clarity. (e.g. see the description of the <tt class="docutils literal"><span class="pre">A</span></tt> constraint on X86, which,
+despite existing only for use with this feature, is not really a good idea to
+use)</p>
+</div>
+<div class="section" id="indirect-inputs-and-outputs">
+<h5><a class="toc-backref" href="#id669">Indirect inputs and outputs</a><a class="headerlink" href="#indirect-inputs-and-outputs" title="Permalink to this headline">¶</a></h5>
+<p>Indirect output or input constraints can be specified by the “<tt class="docutils literal"><span class="pre">*</span></tt>” modifier
+(which goes after the “<tt class="docutils literal"><span class="pre">=</span></tt>” in case of an output). This indicates that the asm
+will write to or read from the contents of an <em>address</em> provided as an input
+argument. (Note that in this way, indirect outputs act more like an <em>input</em> than
+an output: just like an input, they consume an argument of the call expression,
+rather than producing a return value. An indirect output constraint is an
+“output” only in that the asm is expected to write to the contents of the input
+memory location, instead of just read from it).</p>
+<p>This is most typically used for memory constraint, e.g. “<tt class="docutils literal"><span class="pre">=*m</span></tt>”, to pass the
+address of a variable as a value.</p>
+<p>It is also possible to use an indirect <em>register</em> constraint, but only on output
+(e.g. “<tt class="docutils literal"><span class="pre">=*r</span></tt>”). This will cause LLVM to allocate a register for an output
+value normally, and then, separately emit a store to the address provided as
+input, after the provided inline asm. (It’s not clear what value this
+functionality provides, compared to writing the store explicitly after the asm
+statement, and it can only produce worse code, since it bypasses many
+optimization passes. I would recommend not using it.)</p>
+</div>
+<div class="section" id="clobber-constraints">
+<h5><a class="toc-backref" href="#id670">Clobber constraints</a><a class="headerlink" href="#clobber-constraints" title="Permalink to this headline">¶</a></h5>
+<p>A clobber constraint is indicated by a “<tt class="docutils literal"><span class="pre">~</span></tt>” prefix. A clobber does not
+consume an input operand, nor generate an output. Clobbers cannot use any of the
+general constraint code letters – they may use only explicit register
+constraints, e.g. “<tt class="docutils literal"><span class="pre">~{eax}</span></tt>”. The one exception is that a clobber string of
+“<tt class="docutils literal"><span class="pre">~{memory}</span></tt>” indicates that the assembly writes to arbitrary undeclared
+memory locations – not only the memory pointed to by a declared indirect
+output.</p>
+</div>
+<div class="section" id="constraint-codes">
+<h5><a class="toc-backref" href="#id671">Constraint Codes</a><a class="headerlink" href="#constraint-codes" title="Permalink to this headline">¶</a></h5>
+<p>After a potential prefix comes constraint code, or codes.</p>
+<p>A Constraint Code is either a single letter (e.g. “<tt class="docutils literal"><span class="pre">r</span></tt>”), a “<tt class="docutils literal"><span class="pre">^</span></tt>” character
+followed by two letters (e.g. “<tt class="docutils literal"><span class="pre">^wc</span></tt>”), or “<tt class="docutils literal"><span class="pre">{</span></tt>” register-name “<tt class="docutils literal"><span class="pre">}</span></tt>”
+(e.g. “<tt class="docutils literal"><span class="pre">{eax}</span></tt>”).</p>
+<p>The one and two letter constraint codes are typically chosen to be the same as
+GCC’s constraint codes.</p>
+<p>A single constraint may include one or more than constraint code in it, leaving
+it up to LLVM to choose which one to use. This is included mainly for
+compatibility with the translation of GCC inline asm coming from clang.</p>
+<p>There are two ways to specify alternatives, and either or both may be used in an
+inline asm constraint list:</p>
+<ol class="arabic simple">
+<li>Append the codes to each other, making a constraint code set. E.g. “<tt class="docutils literal"><span class="pre">im</span></tt>”
+or “<tt class="docutils literal"><span class="pre">{eax}m</span></tt>”. This means “choose any of the options in the set”. The
+choice of constraint is made independently for each constraint in the
+constraint list.</li>
+<li>Use “<tt class="docutils literal"><span class="pre">|</span></tt>” between constraint code sets, creating alternatives. Every
+constraint in the constraint list must have the same number of alternative
+sets. With this syntax, the same alternative in <em>all</em> of the items in the
+constraint list will be chosen together.</li>
+</ol>
+<p>Putting those together, you might have a two operand constraint string like
+<tt class="docutils literal"><span class="pre">"rm|r,ri|rm"</span></tt>. This indicates that if operand 0 is <tt class="docutils literal"><span class="pre">r</span></tt> or <tt class="docutils literal"><span class="pre">m</span></tt>, then
+operand 1 may be one of <tt class="docutils literal"><span class="pre">r</span></tt> or <tt class="docutils literal"><span class="pre">i</span></tt>. If operand 0 is <tt class="docutils literal"><span class="pre">r</span></tt>, then operand 1
+may be one of <tt class="docutils literal"><span class="pre">r</span></tt> or <tt class="docutils literal"><span class="pre">m</span></tt>. But, operand 0 and 1 cannot both be of type m.</p>
+<p>However, the use of either of the alternatives features is <em>NOT</em> recommended, as
+LLVM is not able to make an intelligent choice about which one to use. (At the
+point it currently needs to choose, not enough information is available to do so
+in a smart way.) Thus, it simply tries to make a choice that’s most likely to
+compile, not one that will be optimal performance. (e.g., given “<tt class="docutils literal"><span class="pre">rm</span></tt>”, it’ll
+always choose to use memory, not registers). And, if given multiple registers,
+or multiple register classes, it will simply choose the first one. (In fact, it
+doesn’t currently even ensure explicitly specified physical registers are
+unique, so specifying multiple physical registers as alternatives, like
+<tt class="docutils literal"><span class="pre">{r11}{r12},{r11}{r12}</span></tt>, will assign r11 to both operands, not at all what was
+intended.)</p>
+</div>
+<div class="section" id="supported-constraint-code-list">
+<h5><a class="toc-backref" href="#id672">Supported Constraint Code List</a><a class="headerlink" href="#supported-constraint-code-list" title="Permalink to this headline">¶</a></h5>
+<p>The constraint codes are, in general, expected to behave the same way they do in
+GCC. LLVM’s support is often implemented on an ‘as-needed’ basis, to support C
+inline asm code which was supported by GCC. A mismatch in behavior between LLVM
+and GCC likely indicates a bug in LLVM.</p>
+<p>Some constraint codes are typically supported by all targets:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A register in the target’s general purpose register class.</li>
+<li><tt class="docutils literal"><span class="pre">m</span></tt>: A memory address operand. It is target-specific what addressing modes
+are supported, typical examples are register, or register + register offset,
+or register + immediate offset (of some target-specific size).</li>
+<li><tt class="docutils literal"><span class="pre">i</span></tt>: An integer constant (of target-specific width). Allows either a simple
+immediate, or a relocatable value.</li>
+<li><tt class="docutils literal"><span class="pre">n</span></tt>: An integer constant – <em>not</em> including relocatable values.</li>
+<li><tt class="docutils literal"><span class="pre">s</span></tt>: An integer constant, but allowing <em>only</em> relocatable values.</li>
+<li><tt class="docutils literal"><span class="pre">X</span></tt>: Allows an operand of any kind, no constraint whatsoever. Typically
+useful to pass a label for an asm branch or call.</li>
+<li><tt class="docutils literal"><span class="pre">{register-name}</span></tt>: Requires exactly the named physical register.</li>
+</ul>
+<p>Other constraints are target-specific:</p>
+<p>AArch64:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">z</span></tt>: An immediate integer 0. Outputs <tt class="docutils literal"><span class="pre">WZR</span></tt> or <tt class="docutils literal"><span class="pre">XZR</span></tt>, as appropriate.</li>
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: An immediate integer valid for an <tt class="docutils literal"><span class="pre">ADD</span></tt> or <tt class="docutils literal"><span class="pre">SUB</span></tt> instruction,
+i.e. 0 to 4095 with optional shift by 12.</li>
+<li><tt class="docutils literal"><span class="pre">J</span></tt>: An immediate integer that, when negated, is valid for an <tt class="docutils literal"><span class="pre">ADD</span></tt> or
+<tt class="docutils literal"><span class="pre">SUB</span></tt> instruction, i.e. -1 to -4095 with optional left shift by 12.</li>
+<li><tt class="docutils literal"><span class="pre">K</span></tt>: An immediate integer that is valid for the ‘bitmask immediate 32’ of a
+logical instruction like <tt class="docutils literal"><span class="pre">AND</span></tt>, <tt class="docutils literal"><span class="pre">EOR</span></tt>, or <tt class="docutils literal"><span class="pre">ORR</span></tt> with a 32-bit register.</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: An immediate integer that is valid for the ‘bitmask immediate 64’ of a
+logical instruction like <tt class="docutils literal"><span class="pre">AND</span></tt>, <tt class="docutils literal"><span class="pre">EOR</span></tt>, or <tt class="docutils literal"><span class="pre">ORR</span></tt> with a 64-bit register.</li>
+<li><tt class="docutils literal"><span class="pre">M</span></tt>: An immediate integer for use with the <tt class="docutils literal"><span class="pre">MOV</span></tt> assembly alias on a
+32-bit register. This is a superset of <tt class="docutils literal"><span class="pre">K</span></tt>: in addition to the bitmask
+immediate, also allows immediate integers which can be loaded with a single
+<tt class="docutils literal"><span class="pre">MOVZ</span></tt> or <tt class="docutils literal"><span class="pre">MOVL</span></tt> instruction.</li>
+<li><tt class="docutils literal"><span class="pre">N</span></tt>: An immediate integer for use with the <tt class="docutils literal"><span class="pre">MOV</span></tt> assembly alias on a
+64-bit register. This is a superset of <tt class="docutils literal"><span class="pre">L</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">Q</span></tt>: Memory address operand must be in a single register (no
+offsets). (However, LLVM currently does this for the <tt class="docutils literal"><span class="pre">m</span></tt> constraint as
+well.)</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A 32 or 64-bit integer register (W* or X*).</li>
+<li><tt class="docutils literal"><span class="pre">w</span></tt>: A 32, 64, or 128-bit floating-point/SIMD register.</li>
+<li><tt class="docutils literal"><span class="pre">x</span></tt>: A lower 128-bit floating-point/SIMD register (<tt class="docutils literal"><span class="pre">V0</span></tt> to <tt class="docutils literal"><span class="pre">V15</span></tt>).</li>
+</ul>
+<p>AMDGPU:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A 32 or 64-bit integer register.</li>
+<li><tt class="docutils literal"><span class="pre">[0-9]v</span></tt>: The 32-bit VGPR register, number 0-9.</li>
+<li><tt class="docutils literal"><span class="pre">[0-9]s</span></tt>: The 32-bit SGPR register, number 0-9.</li>
+</ul>
+<p>All ARM modes:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">Q</span></tt>, <tt class="docutils literal"><span class="pre">Um</span></tt>, <tt class="docutils literal"><span class="pre">Un</span></tt>, <tt class="docutils literal"><span class="pre">Uq</span></tt>, <tt class="docutils literal"><span class="pre">Us</span></tt>, <tt class="docutils literal"><span class="pre">Ut</span></tt>, <tt class="docutils literal"><span class="pre">Uv</span></tt>, <tt class="docutils literal"><span class="pre">Uy</span></tt>: Memory address
+operand. Treated the same as operand <tt class="docutils literal"><span class="pre">m</span></tt>, at the moment.</li>
+</ul>
+<p>ARM and ARM’s Thumb2 mode:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">j</span></tt>: An immediate integer between 0 and 65535 (valid for <tt class="docutils literal"><span class="pre">MOVW</span></tt>)</li>
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: An immediate integer valid for a data-processing instruction.</li>
+<li><tt class="docutils literal"><span class="pre">J</span></tt>: An immediate integer between -4095 and 4095.</li>
+<li><tt class="docutils literal"><span class="pre">K</span></tt>: An immediate integer whose bitwise inverse is valid for a
+data-processing instruction. (Can be used with template modifier “<tt class="docutils literal"><span class="pre">B</span></tt>” to
+print the inverted value).</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: An immediate integer whose negation is valid for a data-processing
+instruction. (Can be used with template modifier “<tt class="docutils literal"><span class="pre">n</span></tt>” to print the negated
+value).</li>
+<li><tt class="docutils literal"><span class="pre">M</span></tt>: A power of two or a integer between 0 and 32.</li>
+<li><tt class="docutils literal"><span class="pre">N</span></tt>: Invalid immediate constraint.</li>
+<li><tt class="docutils literal"><span class="pre">O</span></tt>: Invalid immediate constraint.</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A general-purpose 32-bit integer register (<tt class="docutils literal"><span class="pre">r0-r15</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">l</span></tt>: In Thumb2 mode, low 32-bit GPR registers (<tt class="docutils literal"><span class="pre">r0-r7</span></tt>). In ARM mode, same
+as <tt class="docutils literal"><span class="pre">r</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">h</span></tt>: In Thumb2 mode, a high 32-bit GPR register (<tt class="docutils literal"><span class="pre">r8-r15</span></tt>). In ARM mode,
+invalid.</li>
+<li><tt class="docutils literal"><span class="pre">w</span></tt>: A 32, 64, or 128-bit floating-point/SIMD register: <tt class="docutils literal"><span class="pre">s0-s31</span></tt>,
+<tt class="docutils literal"><span class="pre">d0-d31</span></tt>, or <tt class="docutils literal"><span class="pre">q0-q15</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">x</span></tt>: A 32, 64, or 128-bit floating-point/SIMD register: <tt class="docutils literal"><span class="pre">s0-s15</span></tt>,
+<tt class="docutils literal"><span class="pre">d0-d7</span></tt>, or <tt class="docutils literal"><span class="pre">q0-q3</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">t</span></tt>: A floating-point/SIMD register, only supports 32-bit values:
+<tt class="docutils literal"><span class="pre">s0-s31</span></tt>.</li>
+</ul>
+<p>ARM’s Thumb1 mode:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: An immediate integer between 0 and 255.</li>
+<li><tt class="docutils literal"><span class="pre">J</span></tt>: An immediate integer between -255 and -1.</li>
+<li><tt class="docutils literal"><span class="pre">K</span></tt>: An immediate integer between 0 and 255, with optional left-shift by
+some amount.</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: An immediate integer between -7 and 7.</li>
+<li><tt class="docutils literal"><span class="pre">M</span></tt>: An immediate integer which is a multiple of 4 between 0 and 1020.</li>
+<li><tt class="docutils literal"><span class="pre">N</span></tt>: An immediate integer between 0 and 31.</li>
+<li><tt class="docutils literal"><span class="pre">O</span></tt>: An immediate integer which is a multiple of 4 between -508 and 508.</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A low 32-bit GPR register (<tt class="docutils literal"><span class="pre">r0-r7</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">l</span></tt>: A low 32-bit GPR register (<tt class="docutils literal"><span class="pre">r0-r7</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">h</span></tt>: A high GPR register (<tt class="docutils literal"><span class="pre">r0-r7</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">w</span></tt>: A 32, 64, or 128-bit floating-point/SIMD register: <tt class="docutils literal"><span class="pre">s0-s31</span></tt>,
+<tt class="docutils literal"><span class="pre">d0-d31</span></tt>, or <tt class="docutils literal"><span class="pre">q0-q15</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">x</span></tt>: A 32, 64, or 128-bit floating-point/SIMD register: <tt class="docutils literal"><span class="pre">s0-s15</span></tt>,
+<tt class="docutils literal"><span class="pre">d0-d7</span></tt>, or <tt class="docutils literal"><span class="pre">q0-q3</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">t</span></tt>: A floating-point/SIMD register, only supports 32-bit values:
+<tt class="docutils literal"><span class="pre">s0-s31</span></tt>.</li>
+</ul>
+<p>Hexagon:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">o</span></tt>, <tt class="docutils literal"><span class="pre">v</span></tt>: A memory address operand, treated the same as constraint <tt class="docutils literal"><span class="pre">m</span></tt>,
+at the moment.</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A 32 or 64-bit register.</li>
+</ul>
+<p>MSP430:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: An 8 or 16-bit register.</li>
+</ul>
+<p>MIPS:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: An immediate signed 16-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">J</span></tt>: An immediate integer zero.</li>
+<li><tt class="docutils literal"><span class="pre">K</span></tt>: An immediate unsigned 16-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: An immediate 32-bit integer, where the lower 16 bits are 0.</li>
+<li><tt class="docutils literal"><span class="pre">N</span></tt>: An immediate integer between -65535 and -1.</li>
+<li><tt class="docutils literal"><span class="pre">O</span></tt>: An immediate signed 15-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">P</span></tt>: An immediate integer between 1 and 65535.</li>
+<li><tt class="docutils literal"><span class="pre">m</span></tt>: A memory address operand. In MIPS-SE mode, allows a base address
+register plus 16-bit immediate offset. In MIPS mode, just a base register.</li>
+<li><tt class="docutils literal"><span class="pre">R</span></tt>: A memory address operand. In MIPS-SE mode, allows a base address
+register plus a 9-bit signed offset. In MIPS mode, the same as constraint
+<tt class="docutils literal"><span class="pre">m</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ZC</span></tt>: A memory address operand, suitable for use in a <tt class="docutils literal"><span class="pre">pref</span></tt>, <tt class="docutils literal"><span class="pre">ll</span></tt>, or
+<tt class="docutils literal"><span class="pre">sc</span></tt> instruction on the given subtarget (details vary).</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt>, <tt class="docutils literal"><span class="pre">d</span></tt>, <tt class="docutils literal"><span class="pre">y</span></tt>: A 32 or 64-bit GPR register.</li>
+<li><tt class="docutils literal"><span class="pre">f</span></tt>: A 32 or 64-bit FPU register (<tt class="docutils literal"><span class="pre">F0-F31</span></tt>), or a 128-bit MSA register
+(<tt class="docutils literal"><span class="pre">W0-W31</span></tt>). In the case of MSA registers, it is recommended to use the <tt class="docutils literal"><span class="pre">w</span></tt>
+argument modifier for compatibility with GCC.</li>
+<li><tt class="docutils literal"><span class="pre">c</span></tt>: A 32-bit or 64-bit GPR register suitable for indirect jump (always
+<tt class="docutils literal"><span class="pre">25</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">l</span></tt>: The <tt class="docutils literal"><span class="pre">lo</span></tt> register, 32 or 64-bit.</li>
+<li><tt class="docutils literal"><span class="pre">x</span></tt>: Invalid.</li>
+</ul>
+<p>NVPTX:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">b</span></tt>: A 1-bit integer register.</li>
+<li><tt class="docutils literal"><span class="pre">c</span></tt> or <tt class="docutils literal"><span class="pre">h</span></tt>: A 16-bit integer register.</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A 32-bit integer register.</li>
+<li><tt class="docutils literal"><span class="pre">l</span></tt> or <tt class="docutils literal"><span class="pre">N</span></tt>: A 64-bit integer register.</li>
+<li><tt class="docutils literal"><span class="pre">f</span></tt>: A 32-bit float register.</li>
+<li><tt class="docutils literal"><span class="pre">d</span></tt>: A 64-bit float register.</li>
+</ul>
+<p>PowerPC:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: An immediate signed 16-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">J</span></tt>: An immediate unsigned 16-bit integer, shifted left 16 bits.</li>
+<li><tt class="docutils literal"><span class="pre">K</span></tt>: An immediate unsigned 16-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: An immediate signed 16-bit integer, shifted left 16 bits.</li>
+<li><tt class="docutils literal"><span class="pre">M</span></tt>: An immediate integer greater than 31.</li>
+<li><tt class="docutils literal"><span class="pre">N</span></tt>: An immediate integer that is an exact power of 2.</li>
+<li><tt class="docutils literal"><span class="pre">O</span></tt>: The immediate integer constant 0.</li>
+<li><tt class="docutils literal"><span class="pre">P</span></tt>: An immediate integer constant whose negation is a signed 16-bit
+constant.</li>
+<li><tt class="docutils literal"><span class="pre">es</span></tt>, <tt class="docutils literal"><span class="pre">o</span></tt>, <tt class="docutils literal"><span class="pre">Q</span></tt>, <tt class="docutils literal"><span class="pre">Z</span></tt>, <tt class="docutils literal"><span class="pre">Zy</span></tt>: A memory address operand, currently
+treated the same as <tt class="docutils literal"><span class="pre">m</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A 32 or 64-bit integer register.</li>
+<li><tt class="docutils literal"><span class="pre">b</span></tt>: A 32 or 64-bit integer register, excluding <tt class="docutils literal"><span class="pre">R0</span></tt> (that is:
+<tt class="docutils literal"><span class="pre">R1-R31</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">f</span></tt>: A 32 or 64-bit float register (<tt class="docutils literal"><span class="pre">F0-F31</span></tt>), or when QPX is enabled, a
+128 or 256-bit QPX register (<tt class="docutils literal"><span class="pre">Q0-Q31</span></tt>; aliases the <tt class="docutils literal"><span class="pre">F</span></tt> registers).</li>
+<li><tt class="docutils literal"><span class="pre">v</span></tt>: For <tt class="docutils literal"><span class="pre">4</span> <span class="pre">x</span> <span class="pre">f32</span></tt> or <tt class="docutils literal"><span class="pre">4</span> <span class="pre">x</span> <span class="pre">f64</span></tt> types, when QPX is enabled, a
+128 or 256-bit QPX register (<tt class="docutils literal"><span class="pre">Q0-Q31</span></tt>), otherwise a 128-bit
+altivec vector register (<tt class="docutils literal"><span class="pre">V0-V31</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">y</span></tt>: Condition register (<tt class="docutils literal"><span class="pre">CR0-CR7</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">wc</span></tt>: An individual CR bit in a CR register.</li>
+<li><tt class="docutils literal"><span class="pre">wa</span></tt>, <tt class="docutils literal"><span class="pre">wd</span></tt>, <tt class="docutils literal"><span class="pre">wf</span></tt>: Any 128-bit VSX vector register, from the full VSX
+register set (overlapping both the floating-point and vector register files).</li>
+<li><tt class="docutils literal"><span class="pre">ws</span></tt>: A 32 or 64-bit floating point register, from the full VSX register
+set.</li>
+</ul>
+<p>Sparc:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: An immediate 13-bit signed integer.</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A 32-bit integer register.</li>
+</ul>
+<p>SystemZ:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: An immediate unsigned 8-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">J</span></tt>: An immediate unsigned 12-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">K</span></tt>: An immediate signed 16-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: An immediate signed 20-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">M</span></tt>: An immediate integer 0x7fffffff.</li>
+<li><tt class="docutils literal"><span class="pre">Q</span></tt>, <tt class="docutils literal"><span class="pre">R</span></tt>, <tt class="docutils literal"><span class="pre">S</span></tt>, <tt class="docutils literal"><span class="pre">T</span></tt>: A memory address operand, treated the same as
+<tt class="docutils literal"><span class="pre">m</span></tt>, at the moment.</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt> or <tt class="docutils literal"><span class="pre">d</span></tt>: A 32, 64, or 128-bit integer register.</li>
+<li><tt class="docutils literal"><span class="pre">a</span></tt>: A 32, 64, or 128-bit integer address register (excludes R0, which in an
+address context evaluates as zero).</li>
+<li><tt class="docutils literal"><span class="pre">h</span></tt>: A 32-bit value in the high part of a 64bit data register
+(LLVM-specific)</li>
+<li><tt class="docutils literal"><span class="pre">f</span></tt>: A 32, 64, or 128-bit floating point register.</li>
+</ul>
+<p>X86:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: An immediate integer between 0 and 31.</li>
+<li><tt class="docutils literal"><span class="pre">J</span></tt>: An immediate integer between 0 and 64.</li>
+<li><tt class="docutils literal"><span class="pre">K</span></tt>: An immediate signed 8-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: An immediate integer, 0xff or 0xffff or (in 64-bit mode only)
+0xffffffff.</li>
+<li><tt class="docutils literal"><span class="pre">M</span></tt>: An immediate integer between 0 and 3.</li>
+<li><tt class="docutils literal"><span class="pre">N</span></tt>: An immediate unsigned 8-bit integer.</li>
+<li><tt class="docutils literal"><span class="pre">O</span></tt>: An immediate integer between 0 and 127.</li>
+<li><tt class="docutils literal"><span class="pre">e</span></tt>: An immediate 32-bit signed integer.</li>
+<li><tt class="docutils literal"><span class="pre">Z</span></tt>: An immediate 32-bit unsigned integer.</li>
+<li><tt class="docutils literal"><span class="pre">o</span></tt>, <tt class="docutils literal"><span class="pre">v</span></tt>: Treated the same as <tt class="docutils literal"><span class="pre">m</span></tt>, at the moment.</li>
+<li><tt class="docutils literal"><span class="pre">q</span></tt>: An 8, 16, 32, or 64-bit register which can be accessed as an 8-bit
+<tt class="docutils literal"><span class="pre">l</span></tt> integer register. On X86-32, this is the <tt class="docutils literal"><span class="pre">a</span></tt>, <tt class="docutils literal"><span class="pre">b</span></tt>, <tt class="docutils literal"><span class="pre">c</span></tt>, and <tt class="docutils literal"><span class="pre">d</span></tt>
+registers, and on X86-64, it is all of the integer registers.</li>
+<li><tt class="docutils literal"><span class="pre">Q</span></tt>: An 8, 16, 32, or 64-bit register which can be accessed as an 8-bit
+<tt class="docutils literal"><span class="pre">h</span></tt> integer register. This is the <tt class="docutils literal"><span class="pre">a</span></tt>, <tt class="docutils literal"><span class="pre">b</span></tt>, <tt class="docutils literal"><span class="pre">c</span></tt>, and <tt class="docutils literal"><span class="pre">d</span></tt> registers.</li>
+<li><tt class="docutils literal"><span class="pre">r</span></tt> or <tt class="docutils literal"><span class="pre">l</span></tt>: An 8, 16, 32, or 64-bit integer register.</li>
+<li><tt class="docutils literal"><span class="pre">R</span></tt>: An 8, 16, 32, or 64-bit “legacy” integer register – one which has
+existed since i386, and can be accessed without the REX prefix.</li>
+<li><tt class="docutils literal"><span class="pre">f</span></tt>: A 32, 64, or 80-bit ‘387 FPU stack pseudo-register.</li>
+<li><tt class="docutils literal"><span class="pre">y</span></tt>: A 64-bit MMX register, if MMX is enabled.</li>
+<li><tt class="docutils literal"><span class="pre">x</span></tt>: If SSE is enabled: a 32 or 64-bit scalar operand, or 128-bit vector
+operand in a SSE register. If AVX is also enabled, can also be a 256-bit
+vector operand in an AVX register. If AVX-512 is also enabled, can also be a
+512-bit vector operand in an AVX512 register, Otherwise, an error.</li>
+<li><tt class="docutils literal"><span class="pre">Y</span></tt>: The same as <tt class="docutils literal"><span class="pre">x</span></tt>, if <em>SSE2</em> is enabled, otherwise an error.</li>
+<li><tt class="docutils literal"><span class="pre">A</span></tt>: Special case: allocates EAX first, then EDX, for a single operand (in
+32-bit mode, a 64-bit integer operand will get split into two registers). It
+is not recommended to use this constraint, as in 64-bit mode, the 64-bit
+operand will get allocated only to RAX – if two 32-bit operands are needed,
+you’re better off splitting it yourself, before passing it to the asm
+statement.</li>
+</ul>
+<p>XCore:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: A 32-bit integer register.</li>
+</ul>
+</div>
+</div>
+<div class="section" id="asm-template-argument-modifiers">
+<span id="inline-asm-modifiers"></span><h4><a class="toc-backref" href="#id673">Asm template argument modifiers</a><a class="headerlink" href="#asm-template-argument-modifiers" title="Permalink to this headline">¶</a></h4>
+<p>In the asm template string, modifiers can be used on the operand reference, like
+“<tt class="docutils literal"><span class="pre">${0:n}</span></tt>”.</p>
+<p>The modifiers are, in general, expected to behave the same way they do in
+GCC. LLVM’s support is often implemented on an ‘as-needed’ basis, to support C
+inline asm code which was supported by GCC. A mismatch in behavior between LLVM
+and GCC likely indicates a bug in LLVM.</p>
+<p>Target-independent:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">c</span></tt>: Print an immediate integer constant unadorned, without
+the target-specific immediate punctuation (e.g. no <tt class="docutils literal"><span class="pre">$</span></tt> prefix).</li>
+<li><tt class="docutils literal"><span class="pre">n</span></tt>: Negate and print immediate integer constant unadorned, without the
+target-specific immediate punctuation (e.g. no <tt class="docutils literal"><span class="pre">$</span></tt> prefix).</li>
+<li><tt class="docutils literal"><span class="pre">l</span></tt>: Print as an unadorned label, without the target-specific label
+punctuation (e.g. no <tt class="docutils literal"><span class="pre">$</span></tt> prefix).</li>
+</ul>
+<p>AArch64:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">w</span></tt>: Print a GPR register with a <tt class="docutils literal"><span class="pre">w*</span></tt> name instead of <tt class="docutils literal"><span class="pre">x*</span></tt> name. E.g.,
+instead of <tt class="docutils literal"><span class="pre">x30</span></tt>, print <tt class="docutils literal"><span class="pre">w30</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">x</span></tt>: Print a GPR register with a <tt class="docutils literal"><span class="pre">x*</span></tt> name. (this is the default, anyhow).</li>
+<li><tt class="docutils literal"><span class="pre">b</span></tt>, <tt class="docutils literal"><span class="pre">h</span></tt>, <tt class="docutils literal"><span class="pre">s</span></tt>, <tt class="docutils literal"><span class="pre">d</span></tt>, <tt class="docutils literal"><span class="pre">q</span></tt>: Print a floating-point/SIMD register with a
+<tt class="docutils literal"><span class="pre">b*</span></tt>, <tt class="docutils literal"><span class="pre">h*</span></tt>, <tt class="docutils literal"><span class="pre">s*</span></tt>, <tt class="docutils literal"><span class="pre">d*</span></tt>, or <tt class="docutils literal"><span class="pre">q*</span></tt> name, rather than the default of
+<tt class="docutils literal"><span class="pre">v*</span></tt>.</li>
+</ul>
+<p>AMDGPU:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: No effect.</li>
+</ul>
+<p>ARM:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">a</span></tt>: Print an operand as an address (with <tt class="docutils literal"><span class="pre">[</span></tt> and <tt class="docutils literal"><span class="pre">]</span></tt> surrounding a
+register).</li>
+<li><tt class="docutils literal"><span class="pre">P</span></tt>: No effect.</li>
+<li><tt class="docutils literal"><span class="pre">q</span></tt>: No effect.</li>
+<li><tt class="docutils literal"><span class="pre">y</span></tt>: Print a VFP single-precision register as an indexed double (e.g. print
+as <tt class="docutils literal"><span class="pre">d4[1]</span></tt> instead of <tt class="docutils literal"><span class="pre">s9</span></tt>)</li>
+<li><tt class="docutils literal"><span class="pre">B</span></tt>: Bitwise invert and print an immediate integer constant without <tt class="docutils literal"><span class="pre">#</span></tt>
+prefix.</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: Print the low 16-bits of an immediate integer constant.</li>
+<li><tt class="docutils literal"><span class="pre">M</span></tt>: Print as a register set suitable for ldm/stm. Also prints <em>all</em>
+register operands subsequent to the specified one (!), so use carefully.</li>
+<li><tt class="docutils literal"><span class="pre">Q</span></tt>: Print the low-order register of a register-pair, or the low-order
+register of a two-register operand.</li>
+<li><tt class="docutils literal"><span class="pre">R</span></tt>: Print the high-order register of a register-pair, or the high-order
+register of a two-register operand.</li>
+<li><tt class="docutils literal"><span class="pre">H</span></tt>: Print the second register of a register-pair. (On a big-endian system,
+<tt class="docutils literal"><span class="pre">H</span></tt> is equivalent to <tt class="docutils literal"><span class="pre">Q</span></tt>, and on little-endian system, <tt class="docutils literal"><span class="pre">H</span></tt> is equivalent
+to <tt class="docutils literal"><span class="pre">R</span></tt>.)</li>
+<li><tt class="docutils literal"><span class="pre">e</span></tt>: Print the low doubleword register of a NEON quad register.</li>
+<li><tt class="docutils literal"><span class="pre">f</span></tt>: Print the high doubleword register of a NEON quad register.</li>
+<li><tt class="docutils literal"><span class="pre">m</span></tt>: Print the base register of a memory operand without the <tt class="docutils literal"><span class="pre">[</span></tt> and <tt class="docutils literal"><span class="pre">]</span></tt>
+adornment.</li>
+</ul>
+<p>Hexagon:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: Print the second register of a two-register operand. Requires that it
+has been allocated consecutively to the first.</li>
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: Print the letter ‘i’ if the operand is an integer constant, otherwise
+nothing. Used to print ‘addi’ vs ‘add’ instructions.</li>
+</ul>
+<p>MSP430:</p>
+<p>No additional modifiers.</p>
+<p>MIPS:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">X</span></tt>: Print an immediate integer as hexadecimal</li>
+<li><tt class="docutils literal"><span class="pre">x</span></tt>: Print the low 16 bits of an immediate integer as hexadecimal.</li>
+<li><tt class="docutils literal"><span class="pre">d</span></tt>: Print an immediate integer as decimal.</li>
+<li><tt class="docutils literal"><span class="pre">m</span></tt>: Subtract one and print an immediate integer as decimal.</li>
+<li><tt class="docutils literal"><span class="pre">z</span></tt>: Print $0 if an immediate zero, otherwise print normally.</li>
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: Print the low-order register of a two-register operand, or prints the
+address of the low-order word of a double-word memory operand.</li>
+<li><tt class="docutils literal"><span class="pre">M</span></tt>: Print the high-order register of a two-register operand, or prints the
+address of the high-order word of a double-word memory operand.</li>
+<li><tt class="docutils literal"><span class="pre">D</span></tt>: Print the second register of a two-register operand, or prints the
+second word of a double-word memory operand. (On a big-endian system, <tt class="docutils literal"><span class="pre">D</span></tt> is
+equivalent to <tt class="docutils literal"><span class="pre">L</span></tt>, and on little-endian system, <tt class="docutils literal"><span class="pre">D</span></tt> is equivalent to
+<tt class="docutils literal"><span class="pre">M</span></tt>.)</li>
+<li><tt class="docutils literal"><span class="pre">w</span></tt>: No effect. Provided for compatibility with GCC which requires this
+modifier in order to print MSA registers (<tt class="docutils literal"><span class="pre">W0-W31</span></tt>) with the <tt class="docutils literal"><span class="pre">f</span></tt>
+constraint.</li>
+</ul>
+<p>NVPTX:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: No effect.</li>
+</ul>
+<p>PowerPC:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">L</span></tt>: Print the second register of a two-register operand. Requires that it
+has been allocated consecutively to the first.</li>
+<li><tt class="docutils literal"><span class="pre">I</span></tt>: Print the letter ‘i’ if the operand is an integer constant, otherwise
+nothing. Used to print ‘addi’ vs ‘add’ instructions.</li>
+<li><tt class="docutils literal"><span class="pre">y</span></tt>: For a memory operand, prints formatter for a two-register X-form
+instruction. (Currently always prints <tt class="docutils literal"><span class="pre">r0,OPERAND</span></tt>).</li>
+<li><tt class="docutils literal"><span class="pre">U</span></tt>: Prints ‘u’ if the memory operand is an update form, and nothing
+otherwise. (NOTE: LLVM does not support update form, so this will currently
+always print nothing)</li>
+<li><tt class="docutils literal"><span class="pre">X</span></tt>: Prints ‘x’ if the memory operand is an indexed form. (NOTE: LLVM does
+not support indexed form, so this will currently always print nothing)</li>
+</ul>
+<p>Sparc:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">r</span></tt>: No effect.</li>
+</ul>
+<p>SystemZ:</p>
+<p>SystemZ implements only <tt class="docutils literal"><span class="pre">n</span></tt>, and does <em>not</em> support any of the other
+target-independent modifiers.</p>
+<p>X86:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">c</span></tt>: Print an unadorned integer or symbol name. (The latter is
+target-specific behavior for this typically target-independent modifier).</li>
+<li><tt class="docutils literal"><span class="pre">A</span></tt>: Print a register name with a ‘<tt class="docutils literal"><span class="pre">*</span></tt>‘ before it.</li>
+<li><tt class="docutils literal"><span class="pre">b</span></tt>: Print an 8-bit register name (e.g. <tt class="docutils literal"><span class="pre">al</span></tt>); do nothing on a memory
+operand.</li>
+<li><tt class="docutils literal"><span class="pre">h</span></tt>: Print the upper 8-bit register name (e.g. <tt class="docutils literal"><span class="pre">ah</span></tt>); do nothing on a
+memory operand.</li>
+<li><tt class="docutils literal"><span class="pre">w</span></tt>: Print the 16-bit register name (e.g. <tt class="docutils literal"><span class="pre">ax</span></tt>); do nothing on a memory
+operand.</li>
+<li><tt class="docutils literal"><span class="pre">k</span></tt>: Print the 32-bit register name (e.g. <tt class="docutils literal"><span class="pre">eax</span></tt>); do nothing on a memory
+operand.</li>
+<li><tt class="docutils literal"><span class="pre">q</span></tt>: Print the 64-bit register name (e.g. <tt class="docutils literal"><span class="pre">rax</span></tt>), if 64-bit registers are
+available, otherwise the 32-bit register name; do nothing on a memory operand.</li>
+<li><tt class="docutils literal"><span class="pre">n</span></tt>: Negate and print an unadorned integer, or, for operands other than an
+immediate integer (e.g. a relocatable symbol expression), print a ‘-‘ before
+the operand. (The behavior for relocatable symbol expressions is a
+target-specific behavior for this typically target-independent modifier)</li>
+<li><tt class="docutils literal"><span class="pre">H</span></tt>: Print a memory reference with additional offset +8.</li>
+<li><tt class="docutils literal"><span class="pre">P</span></tt>: Print a memory reference or operand for use as the argument of a call
+instruction. (E.g. omit <tt class="docutils literal"><span class="pre">(rip)</span></tt>, even though it’s PC-relative.)</li>
+</ul>
+<p>XCore:</p>
+<p>No additional modifiers.</p>
+</div>
+<div class="section" id="inline-asm-metadata">
+<h4><a class="toc-backref" href="#id674">Inline Asm Metadata</a><a class="headerlink" href="#inline-asm-metadata" title="Permalink to this headline">¶</a></h4>
+<p>The call instructions that wrap inline asm nodes may have a
+“<tt class="docutils literal"><span class="pre">!srcloc</span></tt>” MDNode attached to it that contains a list of constant
+integers. If present, the code generator will use the integer as the
+location cookie value when report errors through the <tt class="docutils literal"><span class="pre">LLVMContext</span></tt>
+error reporting mechanisms. This allows a front-end to correlate backend
+errors that occur with inline asm back to the source code that produced
+it. For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">call</span> <span class="kt">void</span> <span class="k">asm</span> <span class="k">sideeffect</span> <span class="s">"something bad"</span><span class="p">,</span> <span class="s">""</span><span class="p">(),</span> <span class="nv">!srcloc</span> <span class="nv-Anonymous">!42</span>
+<span class="p">...</span>
+<span class="nv-Anonymous">!42</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i32</span> <span class="m">1234567</span> <span class="p">}</span>
+</pre></div>
+</div>
+<p>It is up to the front-end to make sense of the magic numbers it places
+in the IR. If the MDNode contains multiple constants, the code generator
+will use the one that corresponds to the line of the asm that the error
+occurs on.</p>
+</div>
+</div>
+</div>
+<div class="section" id="metadata">
+<span id="id4"></span><h2><a class="toc-backref" href="#id675">Metadata</a><a class="headerlink" href="#metadata" title="Permalink to this headline">¶</a></h2>
+<p>LLVM IR allows metadata to be attached to instructions in the program
+that can convey extra information about the code to the optimizers and
+code generator. One example application of metadata is source-level
+debug information. There are two metadata primitives: strings and nodes.</p>
+<p>Metadata does not have a type, and is not a value. If referenced from a
+<tt class="docutils literal"><span class="pre">call</span></tt> instruction, it uses the <tt class="docutils literal"><span class="pre">metadata</span></tt> type.</p>
+<p>All metadata are identified in syntax by a exclamation point (‘<tt class="docutils literal"><span class="pre">!</span></tt>‘).</p>
+<div class="section" id="metadata-nodes-and-metadata-strings">
+<span id="metadata-string"></span><h3><a class="toc-backref" href="#id676">Metadata Nodes and Metadata Strings</a><a class="headerlink" href="#metadata-nodes-and-metadata-strings" title="Permalink to this headline">¶</a></h3>
+<p>A metadata string is a string surrounded by double quotes. It can
+contain any character by escaping non-printable characters with
+“<tt class="docutils literal"><span class="pre">\xx</span></tt>” where “<tt class="docutils literal"><span class="pre">xx</span></tt>” is the two digit hex code. For example:
+“<tt class="docutils literal"><span class="pre">!"test\00"</span></tt>”.</p>
+<p>Metadata nodes are represented with notation similar to structure
+constants (a comma separated list of elements, surrounded by braces and
+preceded by an exclamation point). Metadata nodes can have any values as
+their operand. For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="p">!{</span> <span class="nv">!"test\00"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">10</span><span class="p">}</span>
+</pre></div>
+</div>
+<p>Metadata nodes that aren’t uniqued use the <tt class="docutils literal"><span class="pre">distinct</span></tt> keyword. For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = distinct !{!"test\00", i32 10}
+</pre></div>
+</div>
+<p><tt class="docutils literal"><span class="pre">distinct</span></tt> nodes are useful when nodes shouldn’t be merged based on their
+content. They can also occur when transformations cause uniquing collisions
+when metadata operands change.</p>
+<p>A <a class="reference internal" href="#namedmetadatastructure"><em>named metadata</em></a> is a collection of
+metadata nodes, which can be looked up in the module symbol table. For
+example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">!foo</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!4</span><span class="p">,</span> <span class="nv-Anonymous">!3</span><span class="p">}</span>
+</pre></div>
+</div>
+<p>Metadata can be used as function arguments. Here <tt class="docutils literal"><span class="pre">llvm.dbg.value</span></tt>
+function is using two metadata arguments:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">call</span> <span class="kt">void</span> <span class="vg">@llvm.dbg.value</span><span class="p">(</span><span class="kt">metadata</span> <span class="nv-Anonymous">!24</span><span class="p">,</span> <span class="k">i64</span> <span class="m">0</span><span class="p">,</span> <span class="kt">metadata</span> <span class="nv-Anonymous">!25</span><span class="p">)</span>
+</pre></div>
+</div>
+<p>Metadata can be attached with an instruction. Here metadata <tt class="docutils literal"><span class="pre">!21</span></tt> is
+attached to the <tt class="docutils literal"><span class="pre">add</span></tt> instruction using the <tt class="docutils literal"><span class="pre">!dbg</span></tt> identifier:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%indvar.next</span> <span class="p">=</span> <span class="k">add</span> <span class="k">i64</span> <span class="nv">%indvar</span><span class="p">,</span> <span class="m">1</span><span class="p">,</span> <span class="nv">!dbg</span> <span class="nv-Anonymous">!21</span>
+</pre></div>
+</div>
+<p>More information about specific metadata nodes recognized by the
+optimizers and code generator is found below.</p>
+<div class="section" id="specialized-metadata-nodes">
+<span id="specialized-metadata"></span><h4><a class="toc-backref" href="#id677">Specialized Metadata Nodes</a><a class="headerlink" href="#specialized-metadata-nodes" title="Permalink to this headline">¶</a></h4>
+<p>Specialized metadata nodes are custom data structures in metadata (as opposed
+to generic tuples). Their fields are labelled, and can be specified in any
+order.</p>
+<p>These aren’t inherently debug info centric, but currently all the specialized
+metadata nodes are related to debug info.</p>
+<div class="section" id="dicompileunit">
+<span id="id5"></span><h5><a class="toc-backref" href="#id678">DICompileUnit</a><a class="headerlink" href="#dicompileunit" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DICompileUnit</span></tt> nodes represent a compile unit. The <tt class="docutils literal"><span class="pre">enums:</span></tt>,
+<tt class="docutils literal"><span class="pre">retainedTypes:</span></tt>, <tt class="docutils literal"><span class="pre">subprograms:</span></tt>, <tt class="docutils literal"><span class="pre">globals:</span></tt> and <tt class="docutils literal"><span class="pre">imports:</span></tt> fields are
+tuples containing the debug info to be emitted along with the compile unit,
+regardless of code optimizations (some nodes are only emitted if there are
+references to them from instructions).</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang",
+ isOptimized: true, flags: "-O2", runtimeVersion: 2,
+ splitDebugFilename: "abc.debug", emissionKind: 1,
+ enums: !2, retainedTypes: !3, subprograms: !4,
+ globals: !5, imports: !6)
+</pre></div>
+</div>
+<p>Compile unit descriptors provide the root scope for objects declared in a
+specific compilation unit. File descriptors are defined using this scope.
+These descriptors are collected by a named metadata <tt class="docutils literal"><span class="pre">!llvm.dbg.cu</span></tt>. They
+keep track of subprograms, global variables, type information, and imported
+entities (declarations and namespaces).</p>
+</div>
+<div class="section" id="difile">
+<span id="id6"></span><h5><a class="toc-backref" href="#id679">DIFile</a><a class="headerlink" href="#difile" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DIFile</span></tt> nodes represent files. The <tt class="docutils literal"><span class="pre">filename:</span></tt> can include slashes.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DIFile</span><span class="p">(</span><span class="nl">filename:</span> <span class="s">"path/to/file"</span><span class="p">,</span> <span class="nl">directory:</span> <span class="s">"/path/to/dir"</span><span class="p">)</span>
+</pre></div>
+</div>
+<p>Files are sometimes used in <tt class="docutils literal"><span class="pre">scope:</span></tt> fields, and are the only valid target
+for <tt class="docutils literal"><span class="pre">file:</span></tt> fields.</p>
+</div>
+<div class="section" id="dibasictype">
+<span id="id7"></span><h5><a class="toc-backref" href="#id680">DIBasicType</a><a class="headerlink" href="#dibasictype" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DIBasicType</span></tt> nodes represent primitive types, such as <tt class="docutils literal"><span class="pre">int</span></tt>, <tt class="docutils literal"><span class="pre">bool</span></tt> and
+<tt class="docutils literal"><span class="pre">float</span></tt>. <tt class="docutils literal"><span class="pre">tag:</span></tt> defaults to <tt class="docutils literal"><span class="pre">DW_TAG_base_type</span></tt>.</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = !DIBasicType(name: "unsigned char", size: 8, align: 8,
+ encoding: DW_ATE_unsigned_char)
+!1 = !DIBasicType(tag: DW_TAG_unspecified_type, name: "decltype(nullptr)")
+</pre></div>
+</div>
+<p>The <tt class="docutils literal"><span class="pre">encoding:</span></tt> describes the details of the type. Usually it’s one of the
+following:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>DW_ATE_address = 1
+DW_ATE_boolean = 2
+DW_ATE_float = 4
+DW_ATE_signed = 5
+DW_ATE_signed_char = 6
+DW_ATE_unsigned = 7
+DW_ATE_unsigned_char = 8
+</pre></div>
+</div>
+</div>
+<div class="section" id="disubroutinetype">
+<span id="id8"></span><h5><a class="toc-backref" href="#id681">DISubroutineType</a><a class="headerlink" href="#disubroutinetype" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DISubroutineType</span></tt> nodes represent subroutine types. Their <tt class="docutils literal"><span class="pre">types:</span></tt> field
+refers to a tuple; the first operand is the return type, while the rest are the
+types of the formal arguments in order. If the first operand is <tt class="docutils literal"><span class="pre">null</span></tt>, that
+represents a function with no return value (such as <tt class="docutils literal"><span class="pre">void</span> <span class="pre">foo()</span> <span class="pre">{}</span></tt> in C++).</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = !BasicType(name: "int", size: 32, align: 32, DW_ATE_signed)
+!1 = !BasicType(name: "char", size: 8, align: 8, DW_ATE_signed_char)
+!2 = !DISubroutineType(types: !{null, !0, !1}) ; void (int, char)
+</pre></div>
+</div>
+</div>
+<div class="section" id="diderivedtype">
+<span id="id9"></span><h5><a class="toc-backref" href="#id682">DIDerivedType</a><a class="headerlink" href="#diderivedtype" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DIDerivedType</span></tt> nodes represent types derived from other types, such as
+qualified types.</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = !DIBasicType(name: "unsigned char", size: 8, align: 8,
+ encoding: DW_ATE_unsigned_char)
+!1 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !0, size: 32,
+ align: 32)
+</pre></div>
+</div>
+<p>The following <tt class="docutils literal"><span class="pre">tag:</span></tt> values are valid:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>DW_TAG_formal_parameter = 5
+DW_TAG_member = 13
+DW_TAG_pointer_type = 15
+DW_TAG_reference_type = 16
+DW_TAG_typedef = 22
+DW_TAG_ptr_to_member_type = 31
+DW_TAG_const_type = 38
+DW_TAG_volatile_type = 53
+DW_TAG_restrict_type = 55
+</pre></div>
+</div>
+<p><tt class="docutils literal"><span class="pre">DW_TAG_member</span></tt> is used to define a member of a <a class="reference internal" href="#dicompositetype"><em>composite type</em></a> or <a class="reference internal" href="#disubprogram"><em>subprogram</em></a>. The type of the member
+is the <tt class="docutils literal"><span class="pre">baseType:</span></tt>. The <tt class="docutils literal"><span class="pre">offset:</span></tt> is the member’s bit offset.
+<tt class="docutils literal"><span class="pre">DW_TAG_formal_parameter</span></tt> is used to define a member which is a formal
+argument of a subprogram.</p>
+<p><tt class="docutils literal"><span class="pre">DW_TAG_typedef</span></tt> is used to provide a name for the <tt class="docutils literal"><span class="pre">baseType:</span></tt>.</p>
+<p><tt class="docutils literal"><span class="pre">DW_TAG_pointer_type</span></tt>, <tt class="docutils literal"><span class="pre">DW_TAG_reference_type</span></tt>, <tt class="docutils literal"><span class="pre">DW_TAG_const_type</span></tt>,
+<tt class="docutils literal"><span class="pre">DW_TAG_volatile_type</span></tt> and <tt class="docutils literal"><span class="pre">DW_TAG_restrict_type</span></tt> are used to qualify the
+<tt class="docutils literal"><span class="pre">baseType:</span></tt>.</p>
+<p>Note that the <tt class="docutils literal"><span class="pre">void</span> <span class="pre">*</span></tt> type is expressed as a type derived from NULL.</p>
+</div>
+<div class="section" id="dicompositetype">
+<span id="id10"></span><h5><a class="toc-backref" href="#id683">DICompositeType</a><a class="headerlink" href="#dicompositetype" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DICompositeType</span></tt> nodes represent types composed of other types, like
+structures and unions. <tt class="docutils literal"><span class="pre">elements:</span></tt> points to a tuple of the composed types.</p>
+<p>If the source language supports ODR, the <tt class="docutils literal"><span class="pre">identifier:</span></tt> field gives the unique
+identifier used for type merging between modules. When specified, other types
+can refer to composite types indirectly via a <a class="reference internal" href="#metadata-string"><em>metadata string</em></a> that matches their identifier.</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = !DIEnumerator(name: "SixKind", value: 7)
+!1 = !DIEnumerator(name: "SevenKind", value: 7)
+!2 = !DIEnumerator(name: "NegEightKind", value: -8)
+!3 = !DICompositeType(tag: DW_TAG_enumeration_type, name: "Enum", file: !12,
+ line: 2, size: 32, align: 32, identifier: "_M4Enum",
+ elements: !{!0, !1, !2})
+</pre></div>
+</div>
+<p>The following <tt class="docutils literal"><span class="pre">tag:</span></tt> values are valid:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>DW_TAG_array_type = 1
+DW_TAG_class_type = 2
+DW_TAG_enumeration_type = 4
+DW_TAG_structure_type = 19
+DW_TAG_union_type = 23
+DW_TAG_subroutine_type = 21
+DW_TAG_inheritance = 28
+</pre></div>
+</div>
+<p>For <tt class="docutils literal"><span class="pre">DW_TAG_array_type</span></tt>, the <tt class="docutils literal"><span class="pre">elements:</span></tt> should be <a class="reference internal" href="#disubrange"><em>subrange
+descriptors</em></a>, each representing the range of subscripts at that
+level of indexing. The <tt class="docutils literal"><span class="pre">DIFlagVector</span></tt> flag to <tt class="docutils literal"><span class="pre">flags:</span></tt> indicates that an
+array type is a native packed vector.</p>
+<p>For <tt class="docutils literal"><span class="pre">DW_TAG_enumeration_type</span></tt>, the <tt class="docutils literal"><span class="pre">elements:</span></tt> should be <a class="reference internal" href="#dienumerator"><em>enumerator
+descriptors</em></a>, each representing the definition of an enumeration
+value for the set. All enumeration type descriptors are collected in the
+<tt class="docutils literal"><span class="pre">enums:</span></tt> field of the <a class="reference internal" href="#dicompileunit"><em>compile unit</em></a>.</p>
+<p>For <tt class="docutils literal"><span class="pre">DW_TAG_structure_type</span></tt>, <tt class="docutils literal"><span class="pre">DW_TAG_class_type</span></tt>, and
+<tt class="docutils literal"><span class="pre">DW_TAG_union_type</span></tt>, the <tt class="docutils literal"><span class="pre">elements:</span></tt> should be <a class="reference internal" href="#diderivedtype"><em>derived types</em></a> with <tt class="docutils literal"><span class="pre">tag:</span> <span class="pre">DW_TAG_member</span></tt> or <tt class="docutils literal"><span class="pre">tag:</span> <span class="pre">DW_TAG_inheritance</span></tt>.</p>
+</div>
+<div class="section" id="disubrange">
+<span id="id11"></span><h5><a class="toc-backref" href="#id684">DISubrange</a><a class="headerlink" href="#disubrange" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DISubrange</span></tt> nodes are the elements for <tt class="docutils literal"><span class="pre">DW_TAG_array_type</span></tt> variants of
+<a class="reference internal" href="#dicompositetype"><em>DICompositeType</em></a>. <tt class="docutils literal"><span class="pre">count:</span> <span class="pre">-1</span></tt> indicates an empty array.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DISubrange</span><span class="p">(</span><span class="nl">count:</span> <span class="m">5</span><span class="p">,</span> <span class="nl">lowerBound:</span> <span class="m">0</span><span class="p">)</span> <span class="c">; array counting from 0</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="nv">!DISubrange</span><span class="p">(</span><span class="nl">count:</span> <span class="m">5</span><span class="p">,</span> <span class="nl">lowerBound:</span> <span class="m">1</span><span class="p">)</span> <span class="c">; array counting from 1</span>
+<span class="nv-Anonymous">!2</span> <span class="p">=</span> <span class="nv">!DISubrange</span><span class="p">(</span><span class="nl">count:</span> <span class="m">-1</span><span class="p">)</span> <span class="c">; empty array.</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="dienumerator">
+<span id="id12"></span><h5><a class="toc-backref" href="#id685">DIEnumerator</a><a class="headerlink" href="#dienumerator" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DIEnumerator</span></tt> nodes are the elements for <tt class="docutils literal"><span class="pre">DW_TAG_enumeration_type</span></tt>
+variants of <a class="reference internal" href="#dicompositetype"><em>DICompositeType</em></a>.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DIEnumerator</span><span class="p">(</span><span class="nl">name:</span> <span class="s">"SixKind"</span><span class="p">,</span> <span class="nl">value:</span> <span class="m">7</span><span class="p">)</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="nv">!DIEnumerator</span><span class="p">(</span><span class="nl">name:</span> <span class="s">"SevenKind"</span><span class="p">,</span> <span class="nl">value:</span> <span class="m">7</span><span class="p">)</span>
+<span class="nv-Anonymous">!2</span> <span class="p">=</span> <span class="nv">!DIEnumerator</span><span class="p">(</span><span class="nl">name:</span> <span class="s">"NegEightKind"</span><span class="p">,</span> <span class="nl">value:</span> <span class="m">-8</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="ditemplatetypeparameter">
+<h5><a class="toc-backref" href="#id686">DITemplateTypeParameter</a><a class="headerlink" href="#ditemplatetypeparameter" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DITemplateTypeParameter</span></tt> nodes represent type parameters to generic source
+language constructs. They are used (optionally) in <a class="reference internal" href="#dicompositetype"><em>DICompositeType</em></a> and
+<a class="reference internal" href="#disubprogram"><em>DISubprogram</em></a> <tt class="docutils literal"><span class="pre">templateParams:</span></tt> fields.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DITemplateTypeParameter</span><span class="p">(</span><span class="nl">name:</span> <span class="s">"Ty"</span><span class="p">,</span> <span class="nl">type:</span> <span class="nv-Anonymous">!1</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="ditemplatevalueparameter">
+<h5><a class="toc-backref" href="#id687">DITemplateValueParameter</a><a class="headerlink" href="#ditemplatevalueparameter" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DITemplateValueParameter</span></tt> nodes represent value parameters to generic source
+language constructs. <tt class="docutils literal"><span class="pre">tag:</span></tt> defaults to <tt class="docutils literal"><span class="pre">DW_TAG_template_value_parameter</span></tt>,
+but if specified can also be set to <tt class="docutils literal"><span class="pre">DW_TAG_GNU_template_template_param</span></tt> or
+<tt class="docutils literal"><span class="pre">DW_TAG_GNU_template_param_pack</span></tt>. They are used (optionally) in
+<a class="reference internal" href="#dicompositetype"><em>DICompositeType</em></a> and <a class="reference internal" href="#disubprogram"><em>DISubprogram</em></a> <tt class="docutils literal"><span class="pre">templateParams:</span></tt> fields.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DITemplateValueParameter</span><span class="p">(</span><span class="nl">name:</span> <span class="s">"Ty"</span><span class="p">,</span> <span class="nl">type:</span> <span class="nv-Anonymous">!1</span><span class="p">,</span> <span class="nl">value:</span> <span class="k">i32</span> <span class="m">7</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="dinamespace">
+<h5><a class="toc-backref" href="#id688">DINamespace</a><a class="headerlink" href="#dinamespace" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DINamespace</span></tt> nodes represent namespaces in the source language.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DINamespace</span><span class="p">(</span><span class="nl">name:</span> <span class="s">"myawesomeproject"</span><span class="p">,</span> <span class="nl">scope:</span> <span class="nv-Anonymous">!1</span><span class="p">,</span> <span class="nl">file:</span> <span class="nv-Anonymous">!2</span><span class="p">,</span> <span class="nl">line:</span> <span class="m">7</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="diglobalvariable">
+<h5><a class="toc-backref" href="#id689">DIGlobalVariable</a><a class="headerlink" href="#diglobalvariable" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DIGlobalVariable</span></tt> nodes represent global variables in the source language.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DIGlobalVariable</span><span class="p">(</span><span class="nl">name:</span> <span class="s">"foo"</span><span class="p">,</span> <span class="nl">linkageName:</span> <span class="s">"foo"</span><span class="p">,</span> <span class="nl">scope:</span> <span class="nv-Anonymous">!1</span><span class="p">,</span>
+ <span class="nl">file:</span> <span class="nv-Anonymous">!2</span><span class="p">,</span> <span class="nl">line:</span> <span class="m">7</span><span class="p">,</span> <span class="nl">type:</span> <span class="nv-Anonymous">!3</span><span class="p">,</span> <span class="nl">isLocal:</span> <span class="k">true</span><span class="p">,</span>
+ <span class="nl">isDefinition:</span> <span class="k">false</span><span class="p">,</span> <span class="nl">variable:</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@foo</span><span class="p">,</span>
+ <span class="nl">declaration:</span> <span class="nv-Anonymous">!4</span><span class="p">)</span>
+</pre></div>
+</div>
+<p>All global variables should be referenced by the <cite>globals:</cite> field of a
+<a class="reference internal" href="#dicompileunit"><em>compile unit</em></a>.</p>
+</div>
+<div class="section" id="disubprogram">
+<span id="id13"></span><h5><a class="toc-backref" href="#id690">DISubprogram</a><a class="headerlink" href="#disubprogram" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DISubprogram</span></tt> nodes represent functions from the source language. The
+<tt class="docutils literal"><span class="pre">variables:</span></tt> field points at <a class="reference internal" href="#dilocalvariable"><em>variables</em></a> that must be
+retained, even if their IR counterparts are optimized out of the IR. The
+<tt class="docutils literal"><span class="pre">type:</span></tt> field must point at an <a class="reference internal" href="#disubroutinetype"><em>DISubroutineType</em></a>.</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = !DISubprogram(name: "foo", linkageName: "_Zfoov", scope: !1,
+ file: !2, line: 7, type: !3, isLocal: true,
+ isDefinition: false, scopeLine: 8, containingType: !4,
+ virtuality: DW_VIRTUALITY_pure_virtual, virtualIndex: 10,
+ flags: DIFlagPrototyped, isOptimized: true,
+ function: void ()* @_Z3foov,
+ templateParams: !5, declaration: !6, variables: !7)
+</pre></div>
+</div>
+</div>
+<div class="section" id="dilexicalblock">
+<span id="id14"></span><h5><a class="toc-backref" href="#id691">DILexicalBlock</a><a class="headerlink" href="#dilexicalblock" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DILexicalBlock</span></tt> nodes describe nested blocks within a <a class="reference internal" href="#disubprogram"><em>subprogram</em></a>. The line number and column numbers are used to dinstinguish
+two lexical blocks at same depth. They are valid targets for <tt class="docutils literal"><span class="pre">scope:</span></tt>
+fields.</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = distinct !DILexicalBlock(scope: !1, file: !2, line: 7, column: 35)
+</pre></div>
+</div>
+<p>Usually lexical blocks are <tt class="docutils literal"><span class="pre">distinct</span></tt> to prevent node merging based on
+operands.</p>
+</div>
+<div class="section" id="dilexicalblockfile">
+<span id="id15"></span><h5><a class="toc-backref" href="#id692">DILexicalBlockFile</a><a class="headerlink" href="#dilexicalblockfile" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DILexicalBlockFile</span></tt> nodes are used to discriminate between sections of a
+<a class="reference internal" href="#dilexicalblock"><em>lexical block</em></a>. The <tt class="docutils literal"><span class="pre">file:</span></tt> field can be changed to
+indicate textual inclusion, or the <tt class="docutils literal"><span class="pre">discriminator:</span></tt> field can be used to
+discriminate between control flow within a single block in the source language.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DILexicalBlock</span><span class="p">(</span><span class="nl">scope:</span> <span class="nv-Anonymous">!3</span><span class="p">,</span> <span class="nl">file:</span> <span class="nv-Anonymous">!4</span><span class="p">,</span> <span class="nl">line:</span> <span class="m">7</span><span class="p">,</span> <span class="nl">column:</span> <span class="m">35</span><span class="p">)</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="nv">!DILexicalBlockFile</span><span class="p">(</span><span class="nl">scope:</span> <span class="nv-Anonymous">!0</span><span class="p">,</span> <span class="nl">file:</span> <span class="nv-Anonymous">!4</span><span class="p">,</span> <span class="nl">discriminator:</span> <span class="m">0</span><span class="p">)</span>
+<span class="nv-Anonymous">!2</span> <span class="p">=</span> <span class="nv">!DILexicalBlockFile</span><span class="p">(</span><span class="nl">scope:</span> <span class="nv-Anonymous">!0</span><span class="p">,</span> <span class="nl">file:</span> <span class="nv-Anonymous">!4</span><span class="p">,</span> <span class="nl">discriminator:</span> <span class="m">1</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="dilocation">
+<span id="id16"></span><h5><a class="toc-backref" href="#id693">DILocation</a><a class="headerlink" href="#dilocation" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DILocation</span></tt> nodes represent source debug locations. The <tt class="docutils literal"><span class="pre">scope:</span></tt> field is
+mandatory, and points at an <a class="reference internal" href="#dilexicalblockfile"><em>DILexicalBlockFile</em></a>, an
+<a class="reference internal" href="#dilexicalblock"><em>DILexicalBlock</em></a>, or an <a class="reference internal" href="#disubprogram"><em>DISubprogram</em></a>.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="nv">!DILocation</span><span class="p">(</span><span class="nl">line:</span> <span class="m">2900</span><span class="p">,</span> <span class="nl">column:</span> <span class="m">42</span><span class="p">,</span> <span class="nl">scope:</span> <span class="nv-Anonymous">!1</span><span class="p">,</span> <span class="nl">inlinedAt:</span> <span class="nv-Anonymous">!2</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="dilocalvariable">
+<span id="id17"></span><h5><a class="toc-backref" href="#id694">DILocalVariable</a><a class="headerlink" href="#dilocalvariable" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DILocalVariable</span></tt> nodes represent local variables in the source language.
+Instead of <tt class="docutils literal"><span class="pre">DW_TAG_variable</span></tt>, they use LLVM-specific fake tags to
+discriminate between local variables (<tt class="docutils literal"><span class="pre">DW_TAG_auto_variable</span></tt>) and subprogram
+arguments (<tt class="docutils literal"><span class="pre">DW_TAG_arg_variable</span></tt>). In the latter case, the <tt class="docutils literal"><span class="pre">arg:</span></tt> field
+specifies the argument position, and this variable will be included in the
+<tt class="docutils literal"><span class="pre">variables:</span></tt> field of its <a class="reference internal" href="#disubprogram"><em>DISubprogram</em></a>.</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = !DILocalVariable(tag: DW_TAG_arg_variable, name: "this", arg: 0,
+ scope: !3, file: !2, line: 7, type: !3,
+ flags: DIFlagArtificial)
+!1 = !DILocalVariable(tag: DW_TAG_arg_variable, name: "x", arg: 1,
+ scope: !4, file: !2, line: 7, type: !3)
+!1 = !DILocalVariable(tag: DW_TAG_auto_variable, name: "y",
+ scope: !5, file: !2, line: 7, type: !3)
+</pre></div>
+</div>
+</div>
+<div class="section" id="diexpression">
+<h5><a class="toc-backref" href="#id695">DIExpression</a><a class="headerlink" href="#diexpression" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DIExpression</span></tt> nodes represent DWARF expression sequences. They are used in
+<a class="reference internal" href="#dbg-intrinsics"><em>debug intrinsics</em></a> (such as <tt class="docutils literal"><span class="pre">llvm.dbg.declare</span></tt>) to
+describe how the referenced LLVM variable relates to the source language
+variable.</p>
+<p>The current supported vocabulary is limited:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">DW_OP_deref</span></tt> dereferences the working expression.</li>
+<li><tt class="docutils literal"><span class="pre">DW_OP_plus,</span> <span class="pre">93</span></tt> adds <tt class="docutils literal"><span class="pre">93</span></tt> to the working expression.</li>
+<li><tt class="docutils literal"><span class="pre">DW_OP_bit_piece,</span> <span class="pre">16,</span> <span class="pre">8</span></tt> specifies the offset and size (<tt class="docutils literal"><span class="pre">16</span></tt> and <tt class="docutils literal"><span class="pre">8</span></tt>
+here, respectively) of the variable piece from the working expression.</li>
+</ul>
+<div class="highlight-llvm"><div class="highlight"><pre>!0 = !DIExpression(DW_OP_deref)
+!1 = !DIExpression(DW_OP_plus, 3)
+!2 = !DIExpression(DW_OP_bit_piece, 3, 7)
+!3 = !DIExpression(DW_OP_deref, DW_OP_plus, 3, DW_OP_bit_piece, 3, 7)
+</pre></div>
+</div>
+</div>
+<div class="section" id="diobjcproperty">
+<h5><a class="toc-backref" href="#id696">DIObjCProperty</a><a class="headerlink" href="#diobjcproperty" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DIObjCProperty</span></tt> nodes represent Objective-C property nodes.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!3</span> <span class="p">=</span> <span class="nv">!DIObjCProperty</span><span class="p">(</span><span class="nl">name:</span> <span class="s">"foo"</span><span class="p">,</span> <span class="nl">file:</span> <span class="nv-Anonymous">!1</span><span class="p">,</span> <span class="nl">line:</span> <span class="m">7</span><span class="p">,</span> <span class="nl">setter:</span> <span class="s">"setFoo"</span><span class="p">,</span>
+ <span class="nl">getter:</span> <span class="s">"getFoo"</span><span class="p">,</span> <span class="nl">attributes:</span> <span class="m">7</span><span class="p">,</span> <span class="nl">type:</span> <span class="nv-Anonymous">!2</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="diimportedentity">
+<h5><a class="toc-backref" href="#id697">DIImportedEntity</a><a class="headerlink" href="#diimportedentity" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">DIImportedEntity</span></tt> nodes represent entities (such as modules) imported into a
+compile unit.</p>
+<div class="highlight-llvm"><div class="highlight"><pre>!2 = !DIImportedEntity(tag: DW_TAG_imported_module, name: "foo", scope: !0,
+ entity: !1, line: 7)
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="tbaa-metadata">
+<h4><a class="toc-backref" href="#id698">‘<tt class="docutils literal"><span class="pre">tbaa</span></tt>‘ Metadata</a><a class="headerlink" href="#tbaa-metadata" title="Permalink to this headline">¶</a></h4>
+<p>In LLVM IR, memory does not have types, so LLVM’s own type system is not
+suitable for doing TBAA. Instead, metadata is added to the IR to
+describe a type system of a higher level language. This can be used to
+implement typical C/C++ TBAA, but it can also be used to implement
+custom alias analysis behavior for other languages.</p>
+<p>The current metadata format is very simple. TBAA metadata nodes have up
+to three fields, e.g.:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span> <span class="nv">!"an example type tree"</span> <span class="p">}</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="p">!{</span> <span class="nv">!"int"</span><span class="p">,</span> <span class="nv-Anonymous">!0</span> <span class="p">}</span>
+<span class="nv-Anonymous">!2</span> <span class="p">=</span> <span class="p">!{</span> <span class="nv">!"float"</span><span class="p">,</span> <span class="nv-Anonymous">!0</span> <span class="p">}</span>
+<span class="nv-Anonymous">!3</span> <span class="p">=</span> <span class="p">!{</span> <span class="nv">!"const float"</span><span class="p">,</span> <span class="nv-Anonymous">!2</span><span class="p">,</span> <span class="k">i64</span> <span class="m">1</span> <span class="p">}</span>
+</pre></div>
+</div>
+<p>The first field is an identity field. It can be any value, usually a
+metadata string, which uniquely identifies the type. The most important
+name in the tree is the name of the root node. Two trees with different
+root node names are entirely disjoint, even if they have leaves with
+common names.</p>
+<p>The second field identifies the type’s parent node in the tree, or is
+null or omitted for a root node. A type is considered to alias all of
+its descendants and all of its ancestors in the tree. Also, a type is
+considered to alias all types in other trees, so that bitcode produced
+from multiple front-ends is handled conservatively.</p>
+<p>If the third field is present, it’s an integer which if equal to 1
+indicates that the type is “constant” (meaning
+<tt class="docutils literal"><span class="pre">pointsToConstantMemory</span></tt> should return true; see <a class="reference external" href="AliasAnalysis.html#OtherItfs">other useful
+AliasAnalysis methods</a>).</p>
+</div>
+<div class="section" id="tbaa-struct-metadata">
+<h4><a class="toc-backref" href="#id699">‘<tt class="docutils literal"><span class="pre">tbaa.struct</span></tt>‘ Metadata</a><a class="headerlink" href="#tbaa-struct-metadata" title="Permalink to this headline">¶</a></h4>
+<p>The <a class="reference internal" href="#int-memcpy"><em>llvm.memcpy</em></a> is often used to implement
+aggregate assignment operations in C and similar languages, however it
+is defined to copy a contiguous region of memory, which is more than
+strictly necessary for aggregate types which contain holes due to
+padding. Also, it doesn’t contain any TBAA information about the fields
+of the aggregate.</p>
+<p><tt class="docutils literal"><span class="pre">!tbaa.struct</span></tt> metadata can describe which memory subregions in a
+memcpy are padding and what the TBAA tags of the struct are.</p>
+<p>The current metadata format is very simple. <tt class="docutils literal"><span class="pre">!tbaa.struct</span></tt> metadata
+nodes are a list of operands which are in conceptual groups of three.
+For each group of three, the first operand gives the byte offset of a
+field in bytes, the second gives its size in bytes, and the third gives
+its tbaa tag. e.g.:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!4</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i64</span> <span class="m">0</span><span class="p">,</span> <span class="k">i64</span> <span class="m">4</span><span class="p">,</span> <span class="nv-Anonymous">!1</span><span class="p">,</span> <span class="k">i64</span> <span class="m">8</span><span class="p">,</span> <span class="k">i64</span> <span class="m">4</span><span class="p">,</span> <span class="nv-Anonymous">!2</span> <span class="p">}</span>
+</pre></div>
+</div>
+<p>This describes a struct with two fields. The first is at offset 0 bytes
+with size 4 bytes, and has tbaa tag !1. The second is at offset 8 bytes
+and has size 4 bytes and has tbaa tag !2.</p>
+<p>Note that the fields need not be contiguous. In this example, there is a
+4 byte gap between the two fields. This gap represents padding which
+does not carry useful data and need not be preserved.</p>
+</div>
+<div class="section" id="noalias-and-alias-scope-metadata">
+<h4><a class="toc-backref" href="#id700">‘<tt class="docutils literal"><span class="pre">noalias</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">alias.scope</span></tt>‘ Metadata</a><a class="headerlink" href="#noalias-and-alias-scope-metadata" title="Permalink to this headline">¶</a></h4>
+<p><tt class="docutils literal"><span class="pre">noalias</span></tt> and <tt class="docutils literal"><span class="pre">alias.scope</span></tt> metadata provide the ability to specify generic
+noalias memory-access sets. This means that some collection of memory access
+instructions (loads, stores, memory-accessing calls, etc.) that carry
+<tt class="docutils literal"><span class="pre">noalias</span></tt> metadata can specifically be specified not to alias with some other
+collection of memory access instructions that carry <tt class="docutils literal"><span class="pre">alias.scope</span></tt> metadata.
+Each type of metadata specifies a list of scopes where each scope has an id and
+a domain. When evaluating an aliasing query, if for some domain, the set
+of scopes with that domain in one instruction’s <tt class="docutils literal"><span class="pre">alias.scope</span></tt> list is a
+subset of (or equal to) the set of scopes for that domain in another
+instruction’s <tt class="docutils literal"><span class="pre">noalias</span></tt> list, then the two memory accesses are assumed not to
+alias.</p>
+<p>The metadata identifying each domain is itself a list containing one or two
+entries. The first entry is the name of the domain. Note that if the name is a
+string then it can be combined accross functions and translation units. A
+self-reference can be used to create globally unique domain names. A
+descriptive string may optionally be provided as a second list entry.</p>
+<p>The metadata identifying each scope is also itself a list containing two or
+three entries. The first entry is the name of the scope. Note that if the name
+is a string then it can be combined accross functions and translation units. A
+self-reference can be used to create globally unique scope names. A metadata
+reference to the scope’s domain is the second entry. A descriptive string may
+optionally be provided as a third list entry.</p>
+<p>For example,</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="c">; Two scope domains:</span>
+<span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!0</span><span class="p">}</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!1</span><span class="p">}</span>
+
+<span class="c">; Some scopes in these domains:</span>
+<span class="nv-Anonymous">!2</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!2</span><span class="p">,</span> <span class="nv-Anonymous">!0</span><span class="p">}</span>
+<span class="nv-Anonymous">!3</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!3</span><span class="p">,</span> <span class="nv-Anonymous">!0</span><span class="p">}</span>
+<span class="nv-Anonymous">!4</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!4</span><span class="p">,</span> <span class="nv-Anonymous">!1</span><span class="p">}</span>
+
+<span class="c">; Some scope lists:</span>
+<span class="nv-Anonymous">!5</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!4</span><span class="p">}</span> <span class="c">; A list containing only scope !4</span>
+<span class="nv-Anonymous">!6</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!4</span><span class="p">,</span> <span class="nv-Anonymous">!3</span><span class="p">,</span> <span class="nv-Anonymous">!2</span><span class="p">}</span>
+<span class="nv-Anonymous">!7</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!3</span><span class="p">}</span>
+
+<span class="c">; These two instructions don't alias:</span>
+<span class="nv-Anonymous">%0</span> <span class="p">=</span> <span class="k">load</span> <span class="kt">float</span><span class="p">,</span> <span class="kt">float</span><span class="p">*</span> <span class="nv">%c</span><span class="p">,</span> <span class="k">align</span> <span class="m">4</span><span class="p">,</span> <span class="nv">!alias.scope</span> <span class="nv-Anonymous">!5</span>
+<span class="k">store</span> <span class="kt">float</span> <span class="nv-Anonymous">%0</span><span class="p">,</span> <span class="kt">float</span><span class="p">*</span> <span class="nv">%arrayidx.i</span><span class="p">,</span> <span class="k">align</span> <span class="m">4</span><span class="p">,</span> <span class="nv">!noalias</span> <span class="nv-Anonymous">!5</span>
+
+<span class="c">; These two instructions also don't alias (for domain !1, the set of scopes</span>
+<span class="c">; in the !alias.scope equals that in the !noalias list):</span>
+<span class="nv-Anonymous">%2</span> <span class="p">=</span> <span class="k">load</span> <span class="kt">float</span><span class="p">,</span> <span class="kt">float</span><span class="p">*</span> <span class="nv">%c</span><span class="p">,</span> <span class="k">align</span> <span class="m">4</span><span class="p">,</span> <span class="nv">!alias.scope</span> <span class="nv-Anonymous">!5</span>
+<span class="k">store</span> <span class="kt">float</span> <span class="nv-Anonymous">%2</span><span class="p">,</span> <span class="kt">float</span><span class="p">*</span> <span class="nv">%arrayidx.i2</span><span class="p">,</span> <span class="k">align</span> <span class="m">4</span><span class="p">,</span> <span class="nv">!noalias</span> <span class="nv-Anonymous">!6</span>
+
+<span class="c">; These two instructions may alias (for domain !0, the set of scopes in</span>
+<span class="c">; the !noalias list is not a superset of, or equal to, the scopes in the</span>
+<span class="c">; !alias.scope list):</span>
+<span class="nv-Anonymous">%2</span> <span class="p">=</span> <span class="k">load</span> <span class="kt">float</span><span class="p">,</span> <span class="kt">float</span><span class="p">*</span> <span class="nv">%c</span><span class="p">,</span> <span class="k">align</span> <span class="m">4</span><span class="p">,</span> <span class="nv">!alias.scope</span> <span class="nv-Anonymous">!6</span>
+<span class="k">store</span> <span class="kt">float</span> <span class="nv-Anonymous">%0</span><span class="p">,</span> <span class="kt">float</span><span class="p">*</span> <span class="nv">%arrayidx.i</span><span class="p">,</span> <span class="k">align</span> <span class="m">4</span><span class="p">,</span> <span class="nv">!noalias</span> <span class="nv-Anonymous">!7</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="fpmath-metadata">
+<h4><a class="toc-backref" href="#id701">‘<tt class="docutils literal"><span class="pre">fpmath</span></tt>‘ Metadata</a><a class="headerlink" href="#fpmath-metadata" title="Permalink to this headline">¶</a></h4>
+<p><tt class="docutils literal"><span class="pre">fpmath</span></tt> metadata may be attached to any instruction of floating point
+type. It can be used to express the maximum acceptable error in the
+result of that instruction, in ULPs, thus potentially allowing the
+compiler to use a more efficient but less accurate method of computing
+it. ULP is defined as follows:</p>
+<blockquote>
+<div>If <tt class="docutils literal"><span class="pre">x</span></tt> is a real number that lies between two finite consecutive
+floating-point numbers <tt class="docutils literal"><span class="pre">a</span></tt> and <tt class="docutils literal"><span class="pre">b</span></tt>, without being equal to one
+of them, then <tt class="docutils literal"><span class="pre">ulp(x)</span> <span class="pre">=</span> <span class="pre">|b</span> <span class="pre">-</span> <span class="pre">a|</span></tt>, otherwise <tt class="docutils literal"><span class="pre">ulp(x)</span></tt> is the
+distance between the two non-equal finite floating-point numbers
+nearest <tt class="docutils literal"><span class="pre">x</span></tt>. Moreover, <tt class="docutils literal"><span class="pre">ulp(NaN)</span></tt> is <tt class="docutils literal"><span class="pre">NaN</span></tt>.</div></blockquote>
+<p>The metadata node shall consist of a single positive floating point
+number representing the maximum relative error, for example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span> <span class="kt">float</span> <span class="m">2.5</span> <span class="p">}</span> <span class="c">; maximum acceptable inaccuracy is 2.5 ULPs</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="range-metadata">
+<span id="id18"></span><h4><a class="toc-backref" href="#id702">‘<tt class="docutils literal"><span class="pre">range</span></tt>‘ Metadata</a><a class="headerlink" href="#range-metadata" title="Permalink to this headline">¶</a></h4>
+<p><tt class="docutils literal"><span class="pre">range</span></tt> metadata may be attached only to <tt class="docutils literal"><span class="pre">load</span></tt>, <tt class="docutils literal"><span class="pre">call</span></tt> and <tt class="docutils literal"><span class="pre">invoke</span></tt> of
+integer types. It expresses the possible ranges the loaded value or the value
+returned by the called function at this call site is in. The ranges are
+represented with a flattened list of integers. The loaded value or the value
+returned is known to be in the union of the ranges defined by each consecutive
+pair. Each pair has the following properties:</p>
+<ul class="simple">
+<li>The type must match the type loaded by the instruction.</li>
+<li>The pair <tt class="docutils literal"><span class="pre">a,b</span></tt> represents the range <tt class="docutils literal"><span class="pre">[a,b)</span></tt>.</li>
+<li>Both <tt class="docutils literal"><span class="pre">a</span></tt> and <tt class="docutils literal"><span class="pre">b</span></tt> are constants.</li>
+<li>The range is allowed to wrap.</li>
+<li>The range should not represent the full or empty set. That is,
+<tt class="docutils literal"><span class="pre">a!=b</span></tt>.</li>
+</ul>
+<p>In addition, the pairs must be in signed order of the lower bound and
+they must be non-contiguous.</p>
+<p>Examples:</p>
+<div class="highlight-llvm"><div class="highlight"><pre> <span class="nv">%a</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i8</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%x</span><span class="p">,</span> <span class="k">align</span> <span class="m">1</span><span class="p">,</span> <span class="nv">!range</span> <span class="nv-Anonymous">!0</span> <span class="c">; Can only be 0 or 1</span>
+ <span class="nv">%b</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i8</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%y</span><span class="p">,</span> <span class="k">align</span> <span class="m">1</span><span class="p">,</span> <span class="nv">!range</span> <span class="nv-Anonymous">!1</span> <span class="c">; Can only be 255 (-1), 0 or 1</span>
+ <span class="nv">%c</span> <span class="p">=</span> <span class="k">call</span> <span class="k">i8</span> <span class="vg">@foo</span><span class="p">(),</span> <span class="nv">!range</span> <span class="nv-Anonymous">!2</span> <span class="c">; Can only be 0, 1, 3, 4 or 5</span>
+ <span class="nv">%d</span> <span class="p">=</span> <span class="k">invoke</span> <span class="k">i8</span> <span class="vg">@bar</span><span class="p">()</span> <span class="k">to</span> <span class="kt">label</span> <span class="nv">%cont</span>
+ <span class="k">unwind</span> <span class="kt">label</span> <span class="nv">%lpad</span><span class="p">,</span> <span class="nv">!range</span> <span class="nv-Anonymous">!3</span> <span class="c">; Can only be -2, -1, 3, 4 or 5</span>
+<span class="p">...</span>
+<span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i8</span> <span class="m">0</span><span class="p">,</span> <span class="k">i8</span> <span class="m">2</span> <span class="p">}</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i8</span> <span class="m">255</span><span class="p">,</span> <span class="k">i8</span> <span class="m">2</span> <span class="p">}</span>
+<span class="nv-Anonymous">!2</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i8</span> <span class="m">0</span><span class="p">,</span> <span class="k">i8</span> <span class="m">2</span><span class="p">,</span> <span class="k">i8</span> <span class="m">3</span><span class="p">,</span> <span class="k">i8</span> <span class="m">6</span> <span class="p">}</span>
+<span class="nv-Anonymous">!3</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i8</span> <span class="m">-2</span><span class="p">,</span> <span class="k">i8</span> <span class="m">0</span><span class="p">,</span> <span class="k">i8</span> <span class="m">3</span><span class="p">,</span> <span class="k">i8</span> <span class="m">6</span> <span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="llvm-loop">
+<h4><a class="toc-backref" href="#id703">‘<tt class="docutils literal"><span class="pre">llvm.loop</span></tt>‘</a><a class="headerlink" href="#llvm-loop" title="Permalink to this headline">¶</a></h4>
+<p>It is sometimes useful to attach information to loop constructs. Currently,
+loop metadata is implemented as metadata attached to the branch instruction
+in the loop latch block. This type of metadata refer to a metadata node that is
+guaranteed to be separate for each loop. The loop identifier metadata is
+specified with the name <tt class="docutils literal"><span class="pre">llvm.loop</span></tt>.</p>
+<p>The loop identifier metadata is implemented using a metadata that refers to
+itself to avoid merging it with any other identifier metadata, e.g.,
+during module linkage or function inlining. That is, each loop should refer
+to their own identification metadata even if they reside in separate functions.
+The following example contains loop identifier metadata for two separate loop
+constructs:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!0</span><span class="p">}</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!1</span><span class="p">}</span>
+</pre></div>
+</div>
+<p>The loop identifier metadata can be used to specify additional
+per-loop metadata. Any operands after the first operand can be treated
+as user-defined metadata. For example the <tt class="docutils literal"><span class="pre">llvm.loop.unroll.count</span></tt>
+suggests an unroll factor to the loop unroller:</p>
+<div class="highlight-llvm"><div class="highlight"><pre> <span class="k">br</span> <span class="k">i1</span> <span class="nv">%exitcond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%._crit_edge</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%.lr.ph</span><span class="p">,</span> <span class="nv">!llvm.loop</span> <span class="nv-Anonymous">!0</span>
+<span class="p">...</span>
+<span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!0</span><span class="p">,</span> <span class="nv-Anonymous">!1</span><span class="p">}</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.unroll.count"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">4</span><span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="llvm-loop-vectorize-and-llvm-loop-interleave">
+<h4><a class="toc-backref" href="#id704">‘<tt class="docutils literal"><span class="pre">llvm.loop.vectorize</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">llvm.loop.interleave</span></tt>‘</a><a class="headerlink" href="#llvm-loop-vectorize-and-llvm-loop-interleave" title="Permalink to this headline">¶</a></h4>
+<p>Metadata prefixed with <tt class="docutils literal"><span class="pre">llvm.loop.vectorize</span></tt> or <tt class="docutils literal"><span class="pre">llvm.loop.interleave</span></tt> are
+used to control per-loop vectorization and interleaving parameters such as
+vectorization width and interleave count. These metadata should be used in
+conjunction with <tt class="docutils literal"><span class="pre">llvm.loop</span></tt> loop identification metadata. The
+<tt class="docutils literal"><span class="pre">llvm.loop.vectorize</span></tt> and <tt class="docutils literal"><span class="pre">llvm.loop.interleave</span></tt> metadata are only
+optimization hints and the optimizer will only interleave and vectorize loops if
+it believes it is safe to do so. The <tt class="docutils literal"><span class="pre">llvm.mem.parallel_loop_access</span></tt> metadata
+which contains information about loop-carried memory dependencies can be helpful
+in determining the safety of these transformations.</p>
+</div>
+<div class="section" id="llvm-loop-interleave-count-metadata">
+<h4><a class="toc-backref" href="#id705">‘<tt class="docutils literal"><span class="pre">llvm.loop.interleave.count</span></tt>‘ Metadata</a><a class="headerlink" href="#llvm-loop-interleave-count-metadata" title="Permalink to this headline">¶</a></h4>
+<p>This metadata suggests an interleave count to the loop interleaver.
+The first operand is the string <tt class="docutils literal"><span class="pre">llvm.loop.interleave.count</span></tt> and the
+second operand is an integer specifying the interleave count. For
+example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.interleave.count"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">4</span><span class="p">}</span>
+</pre></div>
+</div>
+<p>Note that setting <tt class="docutils literal"><span class="pre">llvm.loop.interleave.count</span></tt> to 1 disables interleaving
+multiple iterations of the loop. If <tt class="docutils literal"><span class="pre">llvm.loop.interleave.count</span></tt> is set to 0
+then the interleave count will be determined automatically.</p>
+</div>
+<div class="section" id="llvm-loop-vectorize-enable-metadata">
+<h4><a class="toc-backref" href="#id706">‘<tt class="docutils literal"><span class="pre">llvm.loop.vectorize.enable</span></tt>‘ Metadata</a><a class="headerlink" href="#llvm-loop-vectorize-enable-metadata" title="Permalink to this headline">¶</a></h4>
+<p>This metadata selectively enables or disables vectorization for the loop. The
+first operand is the string <tt class="docutils literal"><span class="pre">llvm.loop.vectorize.enable</span></tt> and the second operand
+is a bit. If the bit operand value is 1 vectorization is enabled. A value of
+0 disables vectorization:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.vectorize.enable"</span><span class="p">,</span> <span class="k">i1</span> <span class="m">0</span><span class="p">}</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.vectorize.enable"</span><span class="p">,</span> <span class="k">i1</span> <span class="m">1</span><span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="llvm-loop-vectorize-width-metadata">
+<h4><a class="toc-backref" href="#id707">‘<tt class="docutils literal"><span class="pre">llvm.loop.vectorize.width</span></tt>‘ Metadata</a><a class="headerlink" href="#llvm-loop-vectorize-width-metadata" title="Permalink to this headline">¶</a></h4>
+<p>This metadata sets the target width of the vectorizer. The first
+operand is the string <tt class="docutils literal"><span class="pre">llvm.loop.vectorize.width</span></tt> and the second
+operand is an integer specifying the width. For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.vectorize.width"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">4</span><span class="p">}</span>
+</pre></div>
+</div>
+<p>Note that setting <tt class="docutils literal"><span class="pre">llvm.loop.vectorize.width</span></tt> to 1 disables
+vectorization of the loop. If <tt class="docutils literal"><span class="pre">llvm.loop.vectorize.width</span></tt> is set to
+0 or if the loop does not have this metadata the width will be
+determined automatically.</p>
+</div>
+<div class="section" id="llvm-loop-unroll">
+<h4><a class="toc-backref" href="#id708">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll</span></tt>‘</a><a class="headerlink" href="#llvm-loop-unroll" title="Permalink to this headline">¶</a></h4>
+<p>Metadata prefixed with <tt class="docutils literal"><span class="pre">llvm.loop.unroll</span></tt> are loop unrolling
+optimization hints such as the unroll factor. <tt class="docutils literal"><span class="pre">llvm.loop.unroll</span></tt>
+metadata should be used in conjunction with <tt class="docutils literal"><span class="pre">llvm.loop</span></tt> loop
+identification metadata. The <tt class="docutils literal"><span class="pre">llvm.loop.unroll</span></tt> metadata are only
+optimization hints and the unrolling will only be performed if the
+optimizer believes it is safe to do so.</p>
+</div>
+<div class="section" id="llvm-loop-unroll-count-metadata">
+<h4><a class="toc-backref" href="#id709">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll.count</span></tt>‘ Metadata</a><a class="headerlink" href="#llvm-loop-unroll-count-metadata" title="Permalink to this headline">¶</a></h4>
+<p>This metadata suggests an unroll factor to the loop unroller. The
+first operand is the string <tt class="docutils literal"><span class="pre">llvm.loop.unroll.count</span></tt> and the second
+operand is a positive integer specifying the unroll factor. For
+example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.unroll.count"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">4</span><span class="p">}</span>
+</pre></div>
+</div>
+<p>If the trip count of the loop is less than the unroll count the loop
+will be partially unrolled.</p>
+</div>
+<div class="section" id="llvm-loop-unroll-disable-metadata">
+<h4><a class="toc-backref" href="#id710">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll.disable</span></tt>‘ Metadata</a><a class="headerlink" href="#llvm-loop-unroll-disable-metadata" title="Permalink to this headline">¶</a></h4>
+<p>This metadata disables loop unrolling. The metadata has a single operand
+which is the string <tt class="docutils literal"><span class="pre">llvm.loop.unroll.disable</span></tt>. For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.unroll.disable"</span><span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="llvm-loop-unroll-runtime-disable-metadata">
+<h4><a class="toc-backref" href="#id711">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll.runtime.disable</span></tt>‘ Metadata</a><a class="headerlink" href="#llvm-loop-unroll-runtime-disable-metadata" title="Permalink to this headline">¶</a></h4>
+<p>This metadata disables runtime loop unrolling. The metadata has a single
+operand which is the string <tt class="docutils literal"><span class="pre">llvm.loop.unroll.runtime.disable</span></tt>. For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.unroll.runtime.disable"</span><span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="llvm-loop-unroll-full-metadata">
+<h4><a class="toc-backref" href="#id712">‘<tt class="docutils literal"><span class="pre">llvm.loop.unroll.full</span></tt>‘ Metadata</a><a class="headerlink" href="#llvm-loop-unroll-full-metadata" title="Permalink to this headline">¶</a></h4>
+<p>This metadata suggests that the loop should be unrolled fully. The
+metadata has a single operand which is the string <tt class="docutils literal"><span class="pre">llvm.loop.unroll.full</span></tt>.
+For example:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv">!"llvm.loop.unroll.full"</span><span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="llvm-mem">
+<h4><a class="toc-backref" href="#id713">‘<tt class="docutils literal"><span class="pre">llvm.mem</span></tt>‘</a><a class="headerlink" href="#llvm-mem" title="Permalink to this headline">¶</a></h4>
+<p>Metadata types used to annotate memory accesses with information helpful
+for optimizations are prefixed with <tt class="docutils literal"><span class="pre">llvm.mem</span></tt>.</p>
+</div>
+<div class="section" id="llvm-mem-parallel-loop-access-metadata">
+<h4><a class="toc-backref" href="#id714">‘<tt class="docutils literal"><span class="pre">llvm.mem.parallel_loop_access</span></tt>‘ Metadata</a><a class="headerlink" href="#llvm-mem-parallel-loop-access-metadata" title="Permalink to this headline">¶</a></h4>
+<p>The <tt class="docutils literal"><span class="pre">llvm.mem.parallel_loop_access</span></tt> metadata refers to a loop identifier,
+or metadata containing a list of loop identifiers for nested loops.
+The metadata is attached to memory accessing instructions and denotes that
+no loop carried memory dependence exist between it and other instructions denoted
+with the same loop identifier.</p>
+<p>Precisely, given two instructions <tt class="docutils literal"><span class="pre">m1</span></tt> and <tt class="docutils literal"><span class="pre">m2</span></tt> that both have the
+<tt class="docutils literal"><span class="pre">llvm.mem.parallel_loop_access</span></tt> metadata, with <tt class="docutils literal"><span class="pre">L1</span></tt> and <tt class="docutils literal"><span class="pre">L2</span></tt> being the
+set of loops associated with that metadata, respectively, then there is no loop
+carried dependence between <tt class="docutils literal"><span class="pre">m1</span></tt> and <tt class="docutils literal"><span class="pre">m2</span></tt> for loops in both <tt class="docutils literal"><span class="pre">L1</span></tt> and
+<tt class="docutils literal"><span class="pre">L2</span></tt>.</p>
+<p>As a special case, if all memory accessing instructions in a loop have
+<tt class="docutils literal"><span class="pre">llvm.mem.parallel_loop_access</span></tt> metadata that refers to that loop, then the
+loop has no loop carried memory dependences and is considered to be a parallel
+loop.</p>
+<p>Note that if not all memory access instructions have such metadata referring to
+the loop, then the loop is considered not being trivially parallel. Additional
+memory dependence analysis is required to make that determination. As a fail
+safe mechanism, this causes loops that were originally parallel to be considered
+sequential (if optimization passes that are unaware of the parallel semantics
+insert new memory instructions into the loop body).</p>
+<p>Example of a loop that is considered parallel due to its correct use of
+both <tt class="docutils literal"><span class="pre">llvm.loop</span></tt> and <tt class="docutils literal"><span class="pre">llvm.mem.parallel_loop_access</span></tt>
+metadata types that refer to the same loop identifier metadata.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nl">for.body:</span>
+ <span class="p">...</span>
+ <span class="nv">%val0</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%arrayidx</span><span class="p">,</span> <span class="nv">!llvm.mem.parallel_loop_access</span> <span class="nv-Anonymous">!0</span>
+ <span class="p">...</span>
+ <span class="k">store</span> <span class="k">i32</span> <span class="nv">%val0</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%arrayidx1</span><span class="p">,</span> <span class="nv">!llvm.mem.parallel_loop_access</span> <span class="nv-Anonymous">!0</span>
+ <span class="p">...</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%exitcond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%for.end</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%for.body</span><span class="p">,</span> <span class="nv">!llvm.loop</span> <span class="nv-Anonymous">!0</span>
+
+<span class="nl">for.end:</span>
+<span class="p">...</span>
+<span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!0</span><span class="p">}</span>
+</pre></div>
+</div>
+<p>It is also possible to have nested parallel loops. In that case the
+memory accesses refer to a list of loop identifier metadata nodes instead of
+the loop identifier metadata node directly:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nl">outer.for.body:</span>
+ <span class="p">...</span>
+ <span class="nv">%val1</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%arrayidx3</span><span class="p">,</span> <span class="nv">!llvm.mem.parallel_loop_access</span> <span class="nv-Anonymous">!2</span>
+ <span class="p">...</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%inner.for.body</span>
+
+<span class="nl">inner.for.body:</span>
+ <span class="p">...</span>
+ <span class="nv">%val0</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%arrayidx1</span><span class="p">,</span> <span class="nv">!llvm.mem.parallel_loop_access</span> <span class="nv-Anonymous">!0</span>
+ <span class="p">...</span>
+ <span class="k">store</span> <span class="k">i32</span> <span class="nv">%val0</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%arrayidx2</span><span class="p">,</span> <span class="nv">!llvm.mem.parallel_loop_access</span> <span class="nv-Anonymous">!0</span>
+ <span class="p">...</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%exitcond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%inner.for.end</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%inner.for.body</span><span class="p">,</span> <span class="nv">!llvm.loop</span> <span class="nv-Anonymous">!1</span>
+
+<span class="nl">inner.for.end:</span>
+ <span class="p">...</span>
+ <span class="k">store</span> <span class="k">i32</span> <span class="nv">%val1</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%arrayidx4</span><span class="p">,</span> <span class="nv">!llvm.mem.parallel_loop_access</span> <span class="nv-Anonymous">!2</span>
+ <span class="p">...</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%exitcond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%outer.for.end</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%outer.for.body</span><span class="p">,</span> <span class="nv">!llvm.loop</span> <span class="nv-Anonymous">!2</span>
+
+<span class="nl">outer.for.end:</span> <span class="c">; preds = %for.body</span>
+<span class="p">...</span>
+<span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!1</span><span class="p">,</span> <span class="nv-Anonymous">!2</span><span class="p">}</span> <span class="c">; a list of loop identifiers</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!1</span><span class="p">}</span> <span class="c">; an identifier for the inner loop</span>
+<span class="nv-Anonymous">!2</span> <span class="p">=</span> <span class="p">!{</span><span class="nv-Anonymous">!2</span><span class="p">}</span> <span class="c">; an identifier for the outer loop</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="llvm-bitsets">
+<h4><a class="toc-backref" href="#id715">‘<tt class="docutils literal"><span class="pre">llvm.bitsets</span></tt>‘</a><a class="headerlink" href="#llvm-bitsets" title="Permalink to this headline">¶</a></h4>
+<p>The <tt class="docutils literal"><span class="pre">llvm.bitsets</span></tt> global metadata is used to implement
+<a class="reference internal" href="BitSets.html"><em>bitsets</em></a>.</p>
+</div>
+</div>
+</div>
+<div class="section" id="module-flags-metadata">
+<h2><a class="toc-backref" href="#id716">Module Flags Metadata</a><a class="headerlink" href="#module-flags-metadata" title="Permalink to this headline">¶</a></h2>
+<p>Information about the module as a whole is difficult to convey to LLVM’s
+subsystems. The LLVM IR isn’t sufficient to transmit this information.
+The <tt class="docutils literal"><span class="pre">llvm.module.flags</span></tt> named metadata exists in order to facilitate
+this. These flags are in the form of key / value pairs — much like a
+dictionary — making it easy for any subsystem who cares about a flag to
+look it up.</p>
+<p>The <tt class="docutils literal"><span class="pre">llvm.module.flags</span></tt> metadata contains a list of metadata triplets.
+Each triplet has the following form:</p>
+<ul class="simple">
+<li>The first element is a <em>behavior</em> flag, which specifies the behavior
+when two (or more) modules are merged together, and it encounters two
+(or more) metadata with the same ID. The supported behaviors are
+described below.</li>
+<li>The second element is a metadata string that is a unique ID for the
+metadata. Each module may only have one flag entry for each unique ID (not
+including entries with the <strong>Require</strong> behavior).</li>
+<li>The third element is the value of the flag.</li>
+</ul>
+<p>When two (or more) modules are merged together, the resulting
+<tt class="docutils literal"><span class="pre">llvm.module.flags</span></tt> metadata is the union of the modules’ flags. That is, for
+each unique metadata ID string, there will be exactly one entry in the merged
+modules <tt class="docutils literal"><span class="pre">llvm.module.flags</span></tt> metadata table, and the value for that entry will
+be determined by the merge behavior flag, as described below. The only exception
+is that entries with the <em>Require</em> behavior are always preserved.</p>
+<p>The following behaviors are supported:</p>
+<table border="1" class="docutils">
+<colgroup>
+<col width="10%" />
+<col width="90%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">Value</th>
+<th class="head">Behavior</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>1</td>
+<td><dl class="first last docutils">
+<dt><strong>Error</strong></dt>
+<dd>Emits an error if two values disagree, otherwise the resulting value
+is that of the operands.</dd>
+</dl>
+</td>
+</tr>
+<tr class="row-odd"><td>2</td>
+<td><dl class="first last docutils">
+<dt><strong>Warning</strong></dt>
+<dd>Emits a warning if two values disagree. The result value will be the
+operand for the flag from the first module being linked.</dd>
+</dl>
+</td>
+</tr>
+<tr class="row-even"><td>3</td>
+<td><dl class="first last docutils">
+<dt><strong>Require</strong></dt>
+<dd>Adds a requirement that another module flag be present and have a
+specified value after linking is performed. The value must be a
+metadata pair, where the first element of the pair is the ID of the
+module flag to be restricted, and the second element of the pair is
+the value the module flag should be restricted to. This behavior can
+be used to restrict the allowable results (via triggering of an
+error) of linking IDs with the <strong>Override</strong> behavior.</dd>
+</dl>
+</td>
+</tr>
+<tr class="row-odd"><td>4</td>
+<td><dl class="first last docutils">
+<dt><strong>Override</strong></dt>
+<dd>Uses the specified value, regardless of the behavior or value of the
+other module. If both modules specify <strong>Override</strong>, but the values
+differ, an error will be emitted.</dd>
+</dl>
+</td>
+</tr>
+<tr class="row-even"><td>5</td>
+<td><dl class="first last docutils">
+<dt><strong>Append</strong></dt>
+<dd>Appends the two values, which are required to be metadata nodes.</dd>
+</dl>
+</td>
+</tr>
+<tr class="row-odd"><td>6</td>
+<td><dl class="first last docutils">
+<dt><strong>AppendUnique</strong></dt>
+<dd>Appends the two values, which are required to be metadata
+nodes. However, duplicate entries in the second list are dropped
+during the append operation.</dd>
+</dl>
+</td>
+</tr>
+</tbody>
+</table>
+<p>It is an error for a particular unique flag ID to have multiple behaviors,
+except in the case of <strong>Require</strong> (which adds restrictions on another metadata
+value) or <strong>Override</strong>.</p>
+<p>An example of module flags:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">!0</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="nv">!"foo"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span> <span class="p">}</span>
+<span class="nv-Anonymous">!1</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i32</span> <span class="m">4</span><span class="p">,</span> <span class="nv">!"bar"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">37</span> <span class="p">}</span>
+<span class="nv-Anonymous">!2</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i32</span> <span class="m">2</span><span class="p">,</span> <span class="nv">!"qux"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">42</span> <span class="p">}</span>
+<span class="nv-Anonymous">!3</span> <span class="p">=</span> <span class="p">!{</span> <span class="k">i32</span> <span class="m">3</span><span class="p">,</span> <span class="nv">!"qux"</span><span class="p">,</span>
+ <span class="p">!{</span>
+ <span class="nv">!"foo"</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span>
+ <span class="p">}</span>
+<span class="p">}</span>
+<span class="nv">!llvm.module.flags</span> <span class="p">=</span> <span class="p">!{</span> <span class="nv-Anonymous">!0</span><span class="p">,</span> <span class="nv-Anonymous">!1</span><span class="p">,</span> <span class="nv-Anonymous">!2</span><span class="p">,</span> <span class="nv-Anonymous">!3</span> <span class="p">}</span>
+</pre></div>
+</div>
+<ul>
+<li><p class="first">Metadata <tt class="docutils literal"><span class="pre">!0</span></tt> has the ID <tt class="docutils literal"><span class="pre">!"foo"</span></tt> and the value ‘1’. The behavior
+if two or more <tt class="docutils literal"><span class="pre">!"foo"</span></tt> flags are seen is to emit an error if their
+values are not equal.</p>
+</li>
+<li><p class="first">Metadata <tt class="docutils literal"><span class="pre">!1</span></tt> has the ID <tt class="docutils literal"><span class="pre">!"bar"</span></tt> and the value ‘37’. The
+behavior if two or more <tt class="docutils literal"><span class="pre">!"bar"</span></tt> flags are seen is to use the value
+‘37’.</p>
+</li>
+<li><p class="first">Metadata <tt class="docutils literal"><span class="pre">!2</span></tt> has the ID <tt class="docutils literal"><span class="pre">!"qux"</span></tt> and the value ‘42’. The
+behavior if two or more <tt class="docutils literal"><span class="pre">!"qux"</span></tt> flags are seen is to emit a
+warning if their values are not equal.</p>
+</li>
+<li><p class="first">Metadata <tt class="docutils literal"><span class="pre">!3</span></tt> has the ID <tt class="docutils literal"><span class="pre">!"qux"</span></tt> and the value:</p>
+<div class="highlight-python"><div class="highlight"><pre>!{ !"foo", i32 1 }
+</pre></div>
+</div>
+<p>The behavior is to emit an error if the <tt class="docutils literal"><span class="pre">llvm.module.flags</span></tt> does not
+contain a flag with the ID <tt class="docutils literal"><span class="pre">!"foo"</span></tt> that has the value ‘1’ after linking is
+performed.</p>
+</li>
+</ul>
+<div class="section" id="objective-c-garbage-collection-module-flags-metadata">
+<h3><a class="toc-backref" href="#id717">Objective-C Garbage Collection Module Flags Metadata</a><a class="headerlink" href="#objective-c-garbage-collection-module-flags-metadata" title="Permalink to this headline">¶</a></h3>
+<p>On the Mach-O platform, Objective-C stores metadata about garbage
+collection in a special section called “image info”. The metadata
+consists of a version number and a bitmask specifying what types of
+garbage collection are supported (if any) by the file. If two or more
+modules are linked together their garbage collection metadata needs to
+be merged rather than appended together.</p>
+<p>The Objective-C garbage collection module flags metadata consists of the
+following key-value pairs:</p>
+<table border="1" class="docutils">
+<colgroup>
+<col width="30%" />
+<col width="70%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">Key</th>
+<th class="head">Value</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Version</span></tt></td>
+<td><strong>[Required]</strong> — The Objective-C ABI version. Valid values are 1 and 2.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Image</span> <span class="pre">Info</span> <span class="pre">Version</span></tt></td>
+<td><strong>[Required]</strong> — The version of the image info section. Currently
+always 0.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Image</span> <span class="pre">Info</span> <span class="pre">Section</span></tt></td>
+<td><strong>[Required]</strong> — The section to place the metadata. Valid values are
+<tt class="docutils literal"><span class="pre">"__OBJC,</span> <span class="pre">__image_info,</span> <span class="pre">regular"</span></tt> for Objective-C ABI version 1, and
+<tt class="docutils literal"><span class="pre">"__DATA,__objc_imageinfo,</span> <span class="pre">regular,</span> <span class="pre">no_dead_strip"</span></tt> for
+Objective-C ABI version 2.</td>
+</tr>
+<tr class="row-odd"><td><tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Garbage</span> <span class="pre">Collection</span></tt></td>
+<td><strong>[Required]</strong> — Specifies whether garbage collection is supported or
+not. Valid values are 0, for no garbage collection, and 2, for garbage
+collection supported.</td>
+</tr>
+<tr class="row-even"><td><tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">GC</span> <span class="pre">Only</span></tt></td>
+<td><strong>[Optional]</strong> — Specifies that only garbage collection is supported.
+If present, its value must be 6. This flag requires that the
+<tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Garbage</span> <span class="pre">Collection</span></tt> flag have the value 2.</td>
+</tr>
+</tbody>
+</table>
+<p>Some important flag interactions:</p>
+<ul class="simple">
+<li>If a module with <tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Garbage</span> <span class="pre">Collection</span></tt> set to 0 is
+merged with a module with <tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Garbage</span> <span class="pre">Collection</span></tt> set to
+2, then the resulting module has the
+<tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Garbage</span> <span class="pre">Collection</span></tt> flag set to 0.</li>
+<li>A module with <tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">Garbage</span> <span class="pre">Collection</span></tt> set to 0 cannot be
+merged with a module with <tt class="docutils literal"><span class="pre">Objective-C</span> <span class="pre">GC</span> <span class="pre">Only</span></tt> set to 6.</li>
+</ul>
+</div>
+<div class="section" id="automatic-linker-flags-module-flags-metadata">
+<h3><a class="toc-backref" href="#id718">Automatic Linker Flags Module Flags Metadata</a><a class="headerlink" href="#automatic-linker-flags-module-flags-metadata" title="Permalink to this headline">¶</a></h3>
+<p>Some targets support embedding flags to the linker inside individual object
+files. Typically this is used in conjunction with language extensions which
+allow source files to explicitly declare the libraries they depend on, and have
+these automatically be transmitted to the linker via object files.</p>
+<p>These flags are encoded in the IR using metadata in the module flags section,
+using the <tt class="docutils literal"><span class="pre">Linker</span> <span class="pre">Options</span></tt> key. The merge behavior for this flag is required
+to be <tt class="docutils literal"><span class="pre">AppendUnique</span></tt>, and the value for the key is expected to be a metadata
+node which should be a list of other metadata nodes, each of which should be a
+list of metadata strings defining linker options.</p>
+<p>For example, the following metadata section specifies two separate sets of
+linker options, presumably to link against <tt class="docutils literal"><span class="pre">libz</span></tt> and the <tt class="docutils literal"><span class="pre">Cocoa</span></tt>
+framework:</p>
+<div class="highlight-python"><div class="highlight"><pre>!0 = !{ i32 6, !"Linker Options",
+ !{
+ !{ !"-lz" },
+ !{ !"-framework", !"Cocoa" } } }
+!llvm.module.flags = !{ !0 }
+</pre></div>
+</div>
+<p>The metadata encoding as lists of lists of options, as opposed to a collapsed
+list of options, is chosen so that the IR encoding can use multiple option
+strings to specify e.g., a single library, while still having that specifier be
+preserved as an atomic element that can be recognized by a target specific
+assembly writer or object file emitter.</p>
+<p>Each individual option is required to be either a valid option for the target’s
+linker, or an option that is reserved by the target specific assembly writer or
+object file emitter. No other aspect of these options is defined by the IR.</p>
+</div>
+<div class="section" id="c-type-width-module-flags-metadata">
+<h3><a class="toc-backref" href="#id719">C type width Module Flags Metadata</a><a class="headerlink" href="#c-type-width-module-flags-metadata" title="Permalink to this headline">¶</a></h3>
+<p>The ARM backend emits a section into each generated object file describing the
+options that it was compiled with (in a compiler-independent way) to prevent
+linking incompatible objects, and to allow automatic library selection. Some
+of these options are not visible at the IR level, namely wchar_t width and enum
+width.</p>
+<p>To pass this information to the backend, these options are encoded in module
+flags metadata, using the following key-value pairs:</p>
+<table border="1" class="docutils">
+<colgroup>
+<col width="30%" />
+<col width="70%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">Key</th>
+<th class="head">Value</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>short_wchar</td>
+<td><ul class="first last simple">
+<li>0 — sizeof(wchar_t) == 4</li>
+<li>1 — sizeof(wchar_t) == 2</li>
+</ul>
+</td>
+</tr>
+<tr class="row-odd"><td>short_enum</td>
+<td><ul class="first last simple">
+<li>0 — Enums are at least as large as an <tt class="docutils literal"><span class="pre">int</span></tt>.</li>
+<li>1 — Enums are stored in the smallest integer type which can
+represent all of its values.</li>
+</ul>
+</td>
+</tr>
+</tbody>
+</table>
+<p>For example, the following metadata section specifies that the module was
+compiled with a <tt class="docutils literal"><span class="pre">wchar_t</span></tt> width of 4 bytes, and the underlying type of an
+enum is the smallest type which can represent all of its values:</p>
+<div class="highlight-python"><div class="highlight"><pre>!llvm.module.flags = !{!0, !1}
+!0 = !{i32 1, !"short_wchar", i32 1}
+!1 = !{i32 1, !"short_enum", i32 0}
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="intrinsic-global-variables">
+<span id="intrinsicglobalvariables"></span><h2><a class="toc-backref" href="#id720">Intrinsic Global Variables</a><a class="headerlink" href="#intrinsic-global-variables" title="Permalink to this headline">¶</a></h2>
+<p>LLVM has a number of “magic” global variables that contain data that
+affect code generation or other IR semantics. These are documented here.
+All globals of this sort should have a section specified as
+“<tt class="docutils literal"><span class="pre">llvm.metadata</span></tt>”. This section and all globals that start with
+“<tt class="docutils literal"><span class="pre">llvm.</span></tt>” are reserved for use by LLVM.</p>
+<div class="section" id="the-llvm-used-global-variable">
+<span id="gv-llvmused"></span><h3><a class="toc-backref" href="#id721">The ‘<tt class="docutils literal"><span class="pre">llvm.used</span></tt>‘ Global Variable</a><a class="headerlink" href="#the-llvm-used-global-variable" title="Permalink to this headline">¶</a></h3>
+<p>The <tt class="docutils literal"><span class="pre">@llvm.used</span></tt> global is an array which has
+<a class="reference internal" href="#linkage-appending"><em>appending linkage</em></a>. This array contains a list of
+pointers to named global variables, functions and aliases which may optionally
+have a pointer cast formed of bitcast or getelementptr. For example, a legal
+use of it is:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="vg">@X</span> <span class="p">=</span> <span class="k">global</span> <span class="k">i8</span> <span class="m">4</span>
+<span class="vg">@Y</span> <span class="p">=</span> <span class="k">global</span> <span class="k">i32</span> <span class="m">123</span>
+
+<span class="vg">@llvm.used</span> <span class="p">=</span> <span class="k">appending</span> <span class="k">global</span> <span class="p">[</span><span class="m">2</span> <span class="k">x</span> <span class="k">i8</span><span class="p">*]</span> <span class="p">[</span>
+ <span class="k">i8</span><span class="p">*</span> <span class="vg">@X</span><span class="p">,</span>
+ <span class="k">i8</span><span class="p">*</span> <span class="k">bitcast</span> <span class="p">(</span><span class="k">i32</span><span class="p">*</span> <span class="vg">@Y</span> <span class="k">to</span> <span class="k">i8</span><span class="p">*)</span>
+<span class="p">],</span> <span class="k">section</span> <span class="s">"llvm.metadata"</span>
+</pre></div>
+</div>
+<p>If a symbol appears in the <tt class="docutils literal"><span class="pre">@llvm.used</span></tt> list, then the compiler, assembler,
+and linker are required to treat the symbol as if there is a reference to the
+symbol that it cannot see (which is why they have to be named). For example, if
+a variable has internal linkage and no references other than that from the
+<tt class="docutils literal"><span class="pre">@llvm.used</span></tt> list, it cannot be deleted. This is commonly used to represent
+references from inline asms and other things the compiler cannot “see”, and
+corresponds to “<tt class="docutils literal"><span class="pre">attribute((used))</span></tt>” in GNU C.</p>
+<p>On some targets, the code generator must emit a directive to the
+assembler or object file to prevent the assembler and linker from
+molesting the symbol.</p>
+</div>
+<div class="section" id="the-llvm-compiler-used-global-variable">
+<span id="gv-llvmcompilerused"></span><h3><a class="toc-backref" href="#id722">The ‘<tt class="docutils literal"><span class="pre">llvm.compiler.used</span></tt>‘ Global Variable</a><a class="headerlink" href="#the-llvm-compiler-used-global-variable" title="Permalink to this headline">¶</a></h3>
+<p>The <tt class="docutils literal"><span class="pre">@llvm.compiler.used</span></tt> directive is the same as the <tt class="docutils literal"><span class="pre">@llvm.used</span></tt>
+directive, except that it only prevents the compiler from touching the
+symbol. On targets that support it, this allows an intelligent linker to
+optimize references to the symbol without being impeded as it would be
+by <tt class="docutils literal"><span class="pre">@llvm.used</span></tt>.</p>
+<p>This is a rare construct that should only be used in rare circumstances,
+and should not be exposed to source languages.</p>
+</div>
+<div class="section" id="the-llvm-global-ctors-global-variable">
+<span id="gv-llvmglobalctors"></span><h3><a class="toc-backref" href="#id723">The ‘<tt class="docutils literal"><span class="pre">llvm.global_ctors</span></tt>‘ Global Variable</a><a class="headerlink" href="#the-llvm-global-ctors-global-variable" title="Permalink to this headline">¶</a></h3>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">%0</span> <span class="p">=</span> <span class="k">type</span> <span class="p">{</span> <span class="k">i32</span><span class="p">,</span> <span class="kt">void</span> <span class="p">()*,</span> <span class="k">i8</span><span class="p">*</span> <span class="p">}</span>
+<span class="vg">@llvm.global_ctors</span> <span class="p">=</span> <span class="k">appending</span> <span class="k">global</span> <span class="p">[</span><span class="m">1</span> <span class="k">x</span> <span class="nv-Anonymous">%0</span><span class="p">]</span> <span class="p">[</span><span class="nv-Anonymous">%0</span> <span class="p">{</span> <span class="k">i32</span> <span class="m">65535</span><span class="p">,</span> <span class="kt">void</span> <span class="p">()*</span> <span class="vg">@ctor</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="vg">@data</span> <span class="p">}]</span>
+</pre></div>
+</div>
+<p>The <tt class="docutils literal"><span class="pre">@llvm.global_ctors</span></tt> array contains a list of constructor
+functions, priorities, and an optional associated global or function.
+The functions referenced by this array will be called in ascending order
+of priority (i.e. lowest first) when the module is loaded. The order of
+functions with the same priority is not defined.</p>
+<p>If the third field is present, non-null, and points to a global variable
+or function, the initializer function will only run if the associated
+data from the current module is not discarded.</p>
+</div>
+<div class="section" id="the-llvm-global-dtors-global-variable">
+<span id="llvmglobaldtors"></span><h3><a class="toc-backref" href="#id724">The ‘<tt class="docutils literal"><span class="pre">llvm.global_dtors</span></tt>‘ Global Variable</a><a class="headerlink" href="#the-llvm-global-dtors-global-variable" title="Permalink to this headline">¶</a></h3>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv-Anonymous">%0</span> <span class="p">=</span> <span class="k">type</span> <span class="p">{</span> <span class="k">i32</span><span class="p">,</span> <span class="kt">void</span> <span class="p">()*,</span> <span class="k">i8</span><span class="p">*</span> <span class="p">}</span>
+<span class="vg">@llvm.global_dtors</span> <span class="p">=</span> <span class="k">appending</span> <span class="k">global</span> <span class="p">[</span><span class="m">1</span> <span class="k">x</span> <span class="nv-Anonymous">%0</span><span class="p">]</span> <span class="p">[</span><span class="nv-Anonymous">%0</span> <span class="p">{</span> <span class="k">i32</span> <span class="m">65535</span><span class="p">,</span> <span class="kt">void</span> <span class="p">()*</span> <span class="vg">@dtor</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="vg">@data</span> <span class="p">}]</span>
+</pre></div>
+</div>
+<p>The <tt class="docutils literal"><span class="pre">@llvm.global_dtors</span></tt> array contains a list of destructor
+functions, priorities, and an optional associated global or function.
+The functions referenced by this array will be called in descending
+order of priority (i.e. highest first) when the module is unloaded. The
+order of functions with the same priority is not defined.</p>
+<p>If the third field is present, non-null, and points to a global variable
+or function, the destructor function will only run if the associated
+data from the current module is not discarded.</p>
+</div>
+</div>
+<div class="section" id="instruction-reference">
+<h2><a class="toc-backref" href="#id725">Instruction Reference</a><a class="headerlink" href="#instruction-reference" title="Permalink to this headline">¶</a></h2>
+<p>The LLVM instruction set consists of several different classifications
+of instructions: <a class="reference internal" href="#terminators"><em>terminator instructions</em></a>, <a class="reference internal" href="#binaryops"><em>binary
+instructions</em></a>, <a class="reference internal" href="#bitwiseops"><em>bitwise binary
+instructions</em></a>, <a class="reference internal" href="#memoryops"><em>memory instructions</em></a>, and
+<a class="reference internal" href="#otherops"><em>other instructions</em></a>.</p>
+<div class="section" id="terminator-instructions">
+<span id="terminators"></span><h3><a class="toc-backref" href="#id726">Terminator Instructions</a><a class="headerlink" href="#terminator-instructions" title="Permalink to this headline">¶</a></h3>
+<p>As mentioned <a class="reference internal" href="#functionstructure"><em>previously</em></a>, every basic block in a
+program ends with a “Terminator” instruction, which indicates which
+block should be executed after the current block is finished. These
+terminator instructions typically yield a ‘<tt class="docutils literal"><span class="pre">void</span></tt>‘ value: they produce
+control flow, not values (the one exception being the
+‘<a class="reference internal" href="#i-invoke"><em>invoke</em></a>‘ instruction).</p>
+<p>The terminator instructions are: ‘<a class="reference internal" href="#i-ret"><em>ret</em></a>‘,
+‘<a class="reference internal" href="#i-br"><em>br</em></a>‘, ‘<a class="reference internal" href="#i-switch"><em>switch</em></a>‘,
+‘<a class="reference internal" href="#i-indirectbr"><em>indirectbr</em></a>‘, ‘<a class="reference internal" href="#i-invoke"><em>invoke</em></a>‘,
+‘<a class="reference internal" href="#i-resume"><em>resume</em></a>‘, and ‘<a class="reference internal" href="#i-unreachable"><em>unreachable</em></a>‘.</p>
+<div class="section" id="ret-instruction">
+<span id="i-ret"></span><h4><a class="toc-backref" href="#id727">‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ Instruction</a><a class="headerlink" href="#ret-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="syntax">
+<h5><a class="toc-backref" href="#id728">Syntax:</a><a class="headerlink" href="#syntax" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>ret <type> <value> ; Return a value from a non-void function
+ret void ; Return from void function
+</pre></div>
+</div>
+</div>
+<div class="section" id="overview">
+<h5><a class="toc-backref" href="#id729">Overview:</a><a class="headerlink" href="#overview" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ instruction is used to return control flow (and optionally
+a value) from a function back to the caller.</p>
+<p>There are two forms of the ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ instruction: one that returns a
+value and then causes control flow, and one that just causes control
+flow to occur.</p>
+</div>
+<div class="section" id="arguments">
+<h5><a class="toc-backref" href="#id730">Arguments:</a><a class="headerlink" href="#arguments" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ instruction optionally accepts a single argument, the
+return value. The type of the return value must be a ‘<a class="reference internal" href="#t-firstclass"><em>first
+class</em></a>‘ type.</p>
+<p>A function is not <a class="reference internal" href="#wellformed"><em>well formed</em></a> if it it has a non-void
+return type and contains a ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ instruction with no return value or
+a return value with a type that does not match its type, or if it has a
+void return type and contains a ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ instruction with a return
+value.</p>
+</div>
+<div class="section" id="semantics">
+<h5><a class="toc-backref" href="#id731">Semantics:</a><a class="headerlink" href="#semantics" title="Permalink to this headline">¶</a></h5>
+<p>When the ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ instruction is executed, control flow returns back to
+the calling function’s context. If the caller is a
+“<a class="reference internal" href="#i-call"><em>call</em></a>” instruction, execution continues at the
+instruction after the call. If the caller was an
+“<a class="reference internal" href="#i-invoke"><em>invoke</em></a>” instruction, execution continues at the
+beginning of the “normal” destination block. If the instruction returns
+a value, that value shall set the call or invoke instruction’s return
+value.</p>
+</div>
+<div class="section" id="example">
+<h5><a class="toc-backref" href="#id732">Example:</a><a class="headerlink" href="#example" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">ret</span> <span class="k">i32</span> <span class="m">5</span> <span class="c">; Return an integer value of 5</span>
+<span class="k">ret</span> <span class="kt">void</span> <span class="c">; Return from a void function</span>
+<span class="k">ret</span> <span class="p">{</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i8</span> <span class="p">}</span> <span class="p">{</span> <span class="k">i32</span> <span class="m">4</span><span class="p">,</span> <span class="k">i8</span> <span class="m">2</span> <span class="p">}</span> <span class="c">; Return a struct of values 4 and 2</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="br-instruction">
+<span id="i-br"></span><h4><a class="toc-backref" href="#id733">‘<tt class="docutils literal"><span class="pre">br</span></tt>‘ Instruction</a><a class="headerlink" href="#br-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id19">
+<h5><a class="toc-backref" href="#id734">Syntax:</a><a class="headerlink" href="#id19" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>br i1 <cond>, label <iftrue>, label <iffalse>
+br label <dest> ; Unconditional branch
+</pre></div>
+</div>
+</div>
+<div class="section" id="id20">
+<h5><a class="toc-backref" href="#id735">Overview:</a><a class="headerlink" href="#id20" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">br</span></tt>‘ instruction is used to cause control flow to transfer to a
+different basic block in the current function. There are two forms of
+this instruction, corresponding to a conditional branch and an
+unconditional branch.</p>
+</div>
+<div class="section" id="id21">
+<h5><a class="toc-backref" href="#id736">Arguments:</a><a class="headerlink" href="#id21" title="Permalink to this headline">¶</a></h5>
+<p>The conditional branch form of the ‘<tt class="docutils literal"><span class="pre">br</span></tt>‘ instruction takes a single
+‘<tt class="docutils literal"><span class="pre">i1</span></tt>‘ value and two ‘<tt class="docutils literal"><span class="pre">label</span></tt>‘ values. The unconditional form of the
+‘<tt class="docutils literal"><span class="pre">br</span></tt>‘ instruction takes a single ‘<tt class="docutils literal"><span class="pre">label</span></tt>‘ value as a target.</p>
+</div>
+<div class="section" id="id22">
+<h5><a class="toc-backref" href="#id737">Semantics:</a><a class="headerlink" href="#id22" title="Permalink to this headline">¶</a></h5>
+<p>Upon execution of a conditional ‘<tt class="docutils literal"><span class="pre">br</span></tt>‘ instruction, the ‘<tt class="docutils literal"><span class="pre">i1</span></tt>‘
+argument is evaluated. If the value is <tt class="docutils literal"><span class="pre">true</span></tt>, control flows to the
+‘<tt class="docutils literal"><span class="pre">iftrue</span></tt>‘ <tt class="docutils literal"><span class="pre">label</span></tt> argument. If “cond” is <tt class="docutils literal"><span class="pre">false</span></tt>, control flows
+to the ‘<tt class="docutils literal"><span class="pre">iffalse</span></tt>‘ <tt class="docutils literal"><span class="pre">label</span></tt> argument.</p>
+</div>
+<div class="section" id="id23">
+<h5><a class="toc-backref" href="#id738">Example:</a><a class="headerlink" href="#id23" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nl">Test:</span>
+ <span class="nv">%cond</span> <span class="p">=</span> <span class="k">icmp</span> <span class="k">eq</span> <span class="k">i32</span> <span class="nv">%a</span><span class="p">,</span> <span class="nv">%b</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%cond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%IfEqual</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%IfUnequal</span>
+<span class="nl">IfEqual:</span>
+ <span class="k">ret</span> <span class="k">i32</span> <span class="m">1</span>
+<span class="nl">IfUnequal:</span>
+ <span class="k">ret</span> <span class="k">i32</span> <span class="m">0</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="switch-instruction">
+<span id="i-switch"></span><h4><a class="toc-backref" href="#id739">‘<tt class="docutils literal"><span class="pre">switch</span></tt>‘ Instruction</a><a class="headerlink" href="#switch-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id24">
+<h5><a class="toc-backref" href="#id740">Syntax:</a><a class="headerlink" href="#id24" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
+</pre></div>
+</div>
+</div>
+<div class="section" id="id25">
+<h5><a class="toc-backref" href="#id741">Overview:</a><a class="headerlink" href="#id25" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">switch</span></tt>‘ instruction is used to transfer control flow to one of
+several different places. It is a generalization of the ‘<tt class="docutils literal"><span class="pre">br</span></tt>‘
+instruction, allowing a branch to occur to one of many possible
+destinations.</p>
+</div>
+<div class="section" id="id26">
+<h5><a class="toc-backref" href="#id742">Arguments:</a><a class="headerlink" href="#id26" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">switch</span></tt>‘ instruction uses three parameters: an integer
+comparison value ‘<tt class="docutils literal"><span class="pre">value</span></tt>‘, a default ‘<tt class="docutils literal"><span class="pre">label</span></tt>‘ destination, and an
+array of pairs of comparison value constants and ‘<tt class="docutils literal"><span class="pre">label</span></tt>‘s. The table
+is not allowed to contain duplicate constant entries.</p>
+</div>
+<div class="section" id="id27">
+<h5><a class="toc-backref" href="#id743">Semantics:</a><a class="headerlink" href="#id27" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">switch</span></tt> instruction specifies a table of values and destinations.
+When the ‘<tt class="docutils literal"><span class="pre">switch</span></tt>‘ instruction is executed, this table is searched
+for the given value. If the value is found, control flow is transferred
+to the corresponding destination; otherwise, control flow is transferred
+to the default destination.</p>
+</div>
+<div class="section" id="implementation">
+<h5><a class="toc-backref" href="#id744">Implementation:</a><a class="headerlink" href="#implementation" title="Permalink to this headline">¶</a></h5>
+<p>Depending on properties of the target machine and the particular
+<tt class="docutils literal"><span class="pre">switch</span></tt> instruction, this instruction may be code generated in
+different ways. For example, it could be generated as a series of
+chained conditional branches or with a lookup table.</p>
+</div>
+<div class="section" id="id28">
+<h5><a class="toc-backref" href="#id745">Example:</a><a class="headerlink" href="#id28" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="c">; Emulate a conditional br instruction</span>
+<span class="nv">%Val</span> <span class="p">=</span> <span class="k">zext</span> <span class="k">i1</span> <span class="nv">%value</span> <span class="k">to</span> <span class="k">i32</span>
+<span class="k">switch</span> <span class="k">i32</span> <span class="nv">%Val</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%truedest</span> <span class="p">[</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%falsedest</span> <span class="p">]</span>
+
+<span class="c">; Emulate an unconditional br instruction</span>
+<span class="k">switch</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%dest</span> <span class="p">[</span> <span class="p">]</span>
+
+<span class="c">; Implement a jump table:</span>
+<span class="k">switch</span> <span class="k">i32</span> <span class="nv">%val</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%otherwise</span> <span class="p">[</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%onzero</span>
+ <span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%onone</span>
+ <span class="k">i32</span> <span class="m">2</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%ontwo</span> <span class="p">]</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="indirectbr-instruction">
+<span id="i-indirectbr"></span><h4><a class="toc-backref" href="#id746">‘<tt class="docutils literal"><span class="pre">indirectbr</span></tt>‘ Instruction</a><a class="headerlink" href="#indirectbr-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id29">
+<h5><a class="toc-backref" href="#id747">Syntax:</a><a class="headerlink" href="#id29" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
+</pre></div>
+</div>
+</div>
+<div class="section" id="id30">
+<h5><a class="toc-backref" href="#id748">Overview:</a><a class="headerlink" href="#id30" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">indirectbr</span></tt>‘ instruction implements an indirect branch to a
+label within the current function, whose address is specified by
+“<tt class="docutils literal"><span class="pre">address</span></tt>”. Address must be derived from a
+<a class="reference internal" href="#blockaddress"><em>blockaddress</em></a> constant.</p>
+</div>
+<div class="section" id="id31">
+<h5><a class="toc-backref" href="#id749">Arguments:</a><a class="headerlink" href="#id31" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">address</span></tt>‘ argument is the address of the label to jump to. The
+rest of the arguments indicate the full set of possible destinations
+that the address may point to. Blocks are allowed to occur multiple
+times in the destination list, though this isn’t particularly useful.</p>
+<p>This destination list is required so that dataflow analysis has an
+accurate understanding of the CFG.</p>
+</div>
+<div class="section" id="id32">
+<h5><a class="toc-backref" href="#id750">Semantics:</a><a class="headerlink" href="#id32" title="Permalink to this headline">¶</a></h5>
+<p>Control transfers to the block specified in the address argument. All
+possible destination blocks must be listed in the label list, otherwise
+this instruction has undefined behavior. This implies that jumps to
+labels defined in other functions have undefined behavior as well.</p>
+</div>
+<div class="section" id="id33">
+<h5><a class="toc-backref" href="#id751">Implementation:</a><a class="headerlink" href="#id33" title="Permalink to this headline">¶</a></h5>
+<p>This is typically implemented with a jump through a register.</p>
+</div>
+<div class="section" id="id34">
+<h5><a class="toc-backref" href="#id752">Example:</a><a class="headerlink" href="#id34" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="invoke-instruction">
+<span id="i-invoke"></span><h4><a class="toc-backref" href="#id753">‘<tt class="docutils literal"><span class="pre">invoke</span></tt>‘ Instruction</a><a class="headerlink" href="#invoke-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id35">
+<h5><a class="toc-backref" href="#id754">Syntax:</a><a class="headerlink" href="#id35" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = invoke [cconv] [ret attrs] <ptr to function ty> <function ptr val>(<function args>) [fn attrs]
+ to label <normal label> unwind label <exception label>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id36">
+<h5><a class="toc-backref" href="#id755">Overview:</a><a class="headerlink" href="#id36" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">invoke</span></tt>‘ instruction causes control to transfer to a specified
+function, with the possibility of control flow transfer to either the
+‘<tt class="docutils literal"><span class="pre">normal</span></tt>‘ label or the ‘<tt class="docutils literal"><span class="pre">exception</span></tt>‘ label. If the callee function
+returns with the “<tt class="docutils literal"><span class="pre">ret</span></tt>” instruction, control flow will return to the
+“normal” label. If the callee (or any indirect callees) returns via the
+“<a class="reference internal" href="#i-resume"><em>resume</em></a>” instruction or other exception handling
+mechanism, control is interrupted and continued at the dynamically
+nearest “exception” label.</p>
+<p>The ‘<tt class="docutils literal"><span class="pre">exception</span></tt>‘ label is a <a class="reference external" href="ExceptionHandling.html#overview">landing
+pad</a> for the exception. As such,
+‘<tt class="docutils literal"><span class="pre">exception</span></tt>‘ label is required to have the
+“<a class="reference internal" href="#i-landingpad"><em>landingpad</em></a>” instruction, which contains the
+information about the behavior of the program after unwinding happens,
+as its first non-PHI instruction. The restrictions on the
+“<tt class="docutils literal"><span class="pre">landingpad</span></tt>” instruction’s tightly couples it to the “<tt class="docutils literal"><span class="pre">invoke</span></tt>”
+instruction, so that the important information contained within the
+“<tt class="docutils literal"><span class="pre">landingpad</span></tt>” instruction can’t be lost through normal code motion.</p>
+</div>
+<div class="section" id="id37">
+<h5><a class="toc-backref" href="#id756">Arguments:</a><a class="headerlink" href="#id37" title="Permalink to this headline">¶</a></h5>
+<p>This instruction requires several arguments:</p>
+<ol class="arabic simple">
+<li>The optional “cconv” marker indicates which <a class="reference internal" href="#callingconv"><em>calling
+convention</em></a> the call should use. If none is
+specified, the call defaults to using C calling conventions.</li>
+<li>The optional <a class="reference internal" href="#paramattrs"><em>Parameter Attributes</em></a> list for return
+values. Only ‘<tt class="docutils literal"><span class="pre">zeroext</span></tt>‘, ‘<tt class="docutils literal"><span class="pre">signext</span></tt>‘, and ‘<tt class="docutils literal"><span class="pre">inreg</span></tt>‘ attributes
+are valid here.</li>
+<li>‘<tt class="docutils literal"><span class="pre">ptr</span> <span class="pre">to</span> <span class="pre">function</span> <span class="pre">ty</span></tt>‘: shall be the signature of the pointer to
+function value being invoked. In most cases, this is a direct
+function invocation, but indirect <tt class="docutils literal"><span class="pre">invoke</span></tt>‘s are just as possible,
+branching off an arbitrary pointer to function value.</li>
+<li>‘<tt class="docutils literal"><span class="pre">function</span> <span class="pre">ptr</span> <span class="pre">val</span></tt>‘: An LLVM value containing a pointer to a
+function to be invoked.</li>
+<li>‘<tt class="docutils literal"><span class="pre">function</span> <span class="pre">args</span></tt>‘: argument list whose types match the function
+signature argument types and parameter attributes. All arguments must
+be of <a class="reference internal" href="#t-firstclass"><em>first class</em></a> type. If the function signature
+indicates the function accepts a variable number of arguments, the
+extra arguments can be specified.</li>
+<li>‘<tt class="docutils literal"><span class="pre">normal</span> <span class="pre">label</span></tt>‘: the label reached when the called function
+executes a ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ instruction.</li>
+<li>‘<tt class="docutils literal"><span class="pre">exception</span> <span class="pre">label</span></tt>‘: the label reached when a callee returns via
+the <a class="reference internal" href="#i-resume"><em>resume</em></a> instruction or other exception handling
+mechanism.</li>
+<li>The optional <a class="reference internal" href="#fnattrs"><em>function attributes</em></a> list. Only
+‘<tt class="docutils literal"><span class="pre">noreturn</span></tt>‘, ‘<tt class="docutils literal"><span class="pre">nounwind</span></tt>‘, ‘<tt class="docutils literal"><span class="pre">readonly</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">readnone</span></tt>‘
+attributes are valid here.</li>
+</ol>
+</div>
+<div class="section" id="id38">
+<h5><a class="toc-backref" href="#id757">Semantics:</a><a class="headerlink" href="#id38" title="Permalink to this headline">¶</a></h5>
+<p>This instruction is designed to operate as a standard ‘<tt class="docutils literal"><span class="pre">call</span></tt>‘
+instruction in most regards. The primary difference is that it
+establishes an association with a label, which is used by the runtime
+library to unwind the stack.</p>
+<p>This instruction is used in languages with destructors to ensure that
+proper cleanup is performed in the case of either a <tt class="docutils literal"><span class="pre">longjmp</span></tt> or a
+thrown exception. Additionally, this is important for implementation of
+‘<tt class="docutils literal"><span class="pre">catch</span></tt>‘ clauses in high-level languages that support them.</p>
+<p>For the purposes of the SSA form, the definition of the value returned
+by the ‘<tt class="docutils literal"><span class="pre">invoke</span></tt>‘ instruction is deemed to occur on the edge from the
+current block to the “normal” label. If the callee unwinds then no
+return value is available.</p>
+</div>
+<div class="section" id="id39">
+<h5><a class="toc-backref" href="#id758">Example:</a><a class="headerlink" href="#id39" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%retval</span> <span class="p">=</span> <span class="k">invoke</span> <span class="k">i32</span> <span class="vg">@Test</span><span class="p">(</span><span class="k">i32</span> <span class="m">15</span><span class="p">)</span> <span class="k">to</span> <span class="kt">label</span> <span class="nv">%Continue</span>
+ <span class="k">unwind</span> <span class="kt">label</span> <span class="nv">%TestCleanup</span> <span class="c">; i32:retval set</span>
+<span class="nv">%retval</span> <span class="p">=</span> <span class="k">invoke</span> <span class="k">coldcc</span> <span class="k">i32</span> <span class="nv">%Testfnptr</span><span class="p">(</span><span class="k">i32</span> <span class="m">15</span><span class="p">)</span> <span class="k">to</span> <span class="kt">label</span> <span class="nv">%Continue</span>
+ <span class="k">unwind</span> <span class="kt">label</span> <span class="nv">%TestCleanup</span> <span class="c">; i32:retval set</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="resume-instruction">
+<span id="i-resume"></span><h4><a class="toc-backref" href="#id759">‘<tt class="docutils literal"><span class="pre">resume</span></tt>‘ Instruction</a><a class="headerlink" href="#resume-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id40">
+<h5><a class="toc-backref" href="#id760">Syntax:</a><a class="headerlink" href="#id40" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>resume <type> <value>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id41">
+<h5><a class="toc-backref" href="#id761">Overview:</a><a class="headerlink" href="#id41" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">resume</span></tt>‘ instruction is a terminator instruction that has no
+successors.</p>
+</div>
+<div class="section" id="id42">
+<h5><a class="toc-backref" href="#id762">Arguments:</a><a class="headerlink" href="#id42" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">resume</span></tt>‘ instruction requires one argument, which must have the
+same type as the result of any ‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ instruction in the same
+function.</p>
+</div>
+<div class="section" id="id43">
+<h5><a class="toc-backref" href="#id763">Semantics:</a><a class="headerlink" href="#id43" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">resume</span></tt>‘ instruction resumes propagation of an existing
+(in-flight) exception whose unwinding was interrupted with a
+<a class="reference internal" href="#i-landingpad"><em>landingpad</em></a> instruction.</p>
+</div>
+<div class="section" id="id44">
+<h5><a class="toc-backref" href="#id764">Example:</a><a class="headerlink" href="#id44" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>resume { i8*, i32 } %exn
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="unreachable-instruction">
+<span id="i-unreachable"></span><h4><a class="toc-backref" href="#id765">‘<tt class="docutils literal"><span class="pre">unreachable</span></tt>‘ Instruction</a><a class="headerlink" href="#unreachable-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id45">
+<h5><a class="toc-backref" href="#id766">Syntax:</a><a class="headerlink" href="#id45" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><span class="n">unreachable</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id46">
+<h5><a class="toc-backref" href="#id767">Overview:</a><a class="headerlink" href="#id46" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">unreachable</span></tt>‘ instruction has no defined semantics. This
+instruction is used to inform the optimizer that a particular portion of
+the code is not reachable. This can be used to indicate that the code
+after a no-return function cannot be reached, and other facts.</p>
+</div>
+<div class="section" id="id47">
+<h5><a class="toc-backref" href="#id768">Semantics:</a><a class="headerlink" href="#id47" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">unreachable</span></tt>‘ instruction has no defined semantics.</p>
+</div>
+</div>
+</div>
+<div class="section" id="binary-operations">
+<span id="binaryops"></span><h3><a class="toc-backref" href="#id769">Binary Operations</a><a class="headerlink" href="#binary-operations" title="Permalink to this headline">¶</a></h3>
+<p>Binary operators are used to do most of the computation in a program.
+They require two operands of the same type, execute an operation on
+them, and produce a single value. The operands might represent multiple
+data, as is the case with the <a class="reference internal" href="#t-vector"><em>vector</em></a> data type. The
+result value has the same type as its operands.</p>
+<p>There are several different binary operators:</p>
+<div class="section" id="add-instruction">
+<span id="i-add"></span><h4><a class="toc-backref" href="#id770">‘<tt class="docutils literal"><span class="pre">add</span></tt>‘ Instruction</a><a class="headerlink" href="#add-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id48">
+<h5><a class="toc-backref" href="#id771">Syntax:</a><a class="headerlink" href="#id48" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = add <ty> <op1>, <op2> ; yields ty:result
+<result> = add nuw <ty> <op1>, <op2> ; yields ty:result
+<result> = add nsw <ty> <op1>, <op2> ; yields ty:result
+<result> = add nuw nsw <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id49">
+<h5><a class="toc-backref" href="#id772">Overview:</a><a class="headerlink" href="#id49" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">add</span></tt>‘ instruction returns the sum of its two operands.</p>
+</div>
+<div class="section" id="id50">
+<h5><a class="toc-backref" href="#id773">Arguments:</a><a class="headerlink" href="#id50" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">add</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id51">
+<h5><a class="toc-backref" href="#id774">Semantics:</a><a class="headerlink" href="#id51" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the integer sum of the two operands.</p>
+<p>If the sum has unsigned overflow, the result returned is the
+mathematical result modulo 2<sup>n</sup>, where n is the bit width of
+the result.</p>
+<p>Because LLVM integers use a two’s complement representation, this
+instruction is appropriate for both signed and unsigned integers.</p>
+<p><tt class="docutils literal"><span class="pre">nuw</span></tt> and <tt class="docutils literal"><span class="pre">nsw</span></tt> stand for “No Unsigned Wrap” and “No Signed Wrap”,
+respectively. If the <tt class="docutils literal"><span class="pre">nuw</span></tt> and/or <tt class="docutils literal"><span class="pre">nsw</span></tt> keywords are present, the
+result value of the <tt class="docutils literal"><span class="pre">add</span></tt> is a <a class="reference internal" href="#poisonvalues"><em>poison value</em></a> if
+unsigned and/or signed overflow, respectively, occurs.</p>
+</div>
+<div class="section" id="id52">
+<h5><a class="toc-backref" href="#id775">Example:</a><a class="headerlink" href="#id52" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = add i32 4, %var ; yields i32:result = 4 + %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fadd-instruction">
+<span id="i-fadd"></span><h4><a class="toc-backref" href="#id776">‘<tt class="docutils literal"><span class="pre">fadd</span></tt>‘ Instruction</a><a class="headerlink" href="#fadd-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id53">
+<h5><a class="toc-backref" href="#id777">Syntax:</a><a class="headerlink" href="#id53" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fadd [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id54">
+<h5><a class="toc-backref" href="#id778">Overview:</a><a class="headerlink" href="#id54" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fadd</span></tt>‘ instruction returns the sum of its two operands.</p>
+</div>
+<div class="section" id="id55">
+<h5><a class="toc-backref" href="#id779">Arguments:</a><a class="headerlink" href="#id55" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">fadd</span></tt>‘ instruction must be <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of floating point values.
+Both arguments must have identical types.</p>
+</div>
+<div class="section" id="id56">
+<h5><a class="toc-backref" href="#id780">Semantics:</a><a class="headerlink" href="#id56" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the floating point sum of the two operands. This
+instruction can also take any number of <a class="reference internal" href="#fastmath"><em>fast-math flags</em></a>,
+which are optimization hints to enable otherwise unsafe floating point
+optimizations:</p>
+</div>
+<div class="section" id="id57">
+<h5><a class="toc-backref" href="#id781">Example:</a><a class="headerlink" href="#id57" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = fadd float 4.0, %var ; yields float:result = 4.0 + %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="sub-instruction">
+<h4><a class="toc-backref" href="#id782">‘<tt class="docutils literal"><span class="pre">sub</span></tt>‘ Instruction</a><a class="headerlink" href="#sub-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id58">
+<h5><a class="toc-backref" href="#id783">Syntax:</a><a class="headerlink" href="#id58" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = sub <ty> <op1>, <op2> ; yields ty:result
+<result> = sub nuw <ty> <op1>, <op2> ; yields ty:result
+<result> = sub nsw <ty> <op1>, <op2> ; yields ty:result
+<result> = sub nuw nsw <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id59">
+<h5><a class="toc-backref" href="#id784">Overview:</a><a class="headerlink" href="#id59" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">sub</span></tt>‘ instruction returns the difference of its two operands.</p>
+<p>Note that the ‘<tt class="docutils literal"><span class="pre">sub</span></tt>‘ instruction is used to represent the ‘<tt class="docutils literal"><span class="pre">neg</span></tt>‘
+instruction present in most other intermediate representations.</p>
+</div>
+<div class="section" id="id60">
+<h5><a class="toc-backref" href="#id785">Arguments:</a><a class="headerlink" href="#id60" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">sub</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id61">
+<h5><a class="toc-backref" href="#id786">Semantics:</a><a class="headerlink" href="#id61" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the integer difference of the two operands.</p>
+<p>If the difference has unsigned overflow, the result returned is the
+mathematical result modulo 2<sup>n</sup>, where n is the bit width of
+the result.</p>
+<p>Because LLVM integers use a two’s complement representation, this
+instruction is appropriate for both signed and unsigned integers.</p>
+<p><tt class="docutils literal"><span class="pre">nuw</span></tt> and <tt class="docutils literal"><span class="pre">nsw</span></tt> stand for “No Unsigned Wrap” and “No Signed Wrap”,
+respectively. If the <tt class="docutils literal"><span class="pre">nuw</span></tt> and/or <tt class="docutils literal"><span class="pre">nsw</span></tt> keywords are present, the
+result value of the <tt class="docutils literal"><span class="pre">sub</span></tt> is a <a class="reference internal" href="#poisonvalues"><em>poison value</em></a> if
+unsigned and/or signed overflow, respectively, occurs.</p>
+</div>
+<div class="section" id="id62">
+<h5><a class="toc-backref" href="#id787">Example:</a><a class="headerlink" href="#id62" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = sub i32 4, %var ; yields i32:result = 4 - %var
+<result> = sub i32 0, %val ; yields i32:result = -%var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fsub-instruction">
+<span id="i-fsub"></span><h4><a class="toc-backref" href="#id788">‘<tt class="docutils literal"><span class="pre">fsub</span></tt>‘ Instruction</a><a class="headerlink" href="#fsub-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id63">
+<h5><a class="toc-backref" href="#id789">Syntax:</a><a class="headerlink" href="#id63" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fsub [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id64">
+<h5><a class="toc-backref" href="#id790">Overview:</a><a class="headerlink" href="#id64" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fsub</span></tt>‘ instruction returns the difference of its two operands.</p>
+<p>Note that the ‘<tt class="docutils literal"><span class="pre">fsub</span></tt>‘ instruction is used to represent the ‘<tt class="docutils literal"><span class="pre">fneg</span></tt>‘
+instruction present in most other intermediate representations.</p>
+</div>
+<div class="section" id="id65">
+<h5><a class="toc-backref" href="#id791">Arguments:</a><a class="headerlink" href="#id65" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">fsub</span></tt>‘ instruction must be <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of floating point values.
+Both arguments must have identical types.</p>
+</div>
+<div class="section" id="id66">
+<h5><a class="toc-backref" href="#id792">Semantics:</a><a class="headerlink" href="#id66" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the floating point difference of the two operands.
+This instruction can also take any number of <a class="reference internal" href="#fastmath"><em>fast-math
+flags</em></a>, which are optimization hints to enable otherwise
+unsafe floating point optimizations:</p>
+</div>
+<div class="section" id="id67">
+<h5><a class="toc-backref" href="#id793">Example:</a><a class="headerlink" href="#id67" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = fsub float 4.0, %var ; yields float:result = 4.0 - %var
+<result> = fsub float -0.0, %val ; yields float:result = -%var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="mul-instruction">
+<h4><a class="toc-backref" href="#id794">‘<tt class="docutils literal"><span class="pre">mul</span></tt>‘ Instruction</a><a class="headerlink" href="#mul-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id68">
+<h5><a class="toc-backref" href="#id795">Syntax:</a><a class="headerlink" href="#id68" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = mul <ty> <op1>, <op2> ; yields ty:result
+<result> = mul nuw <ty> <op1>, <op2> ; yields ty:result
+<result> = mul nsw <ty> <op1>, <op2> ; yields ty:result
+<result> = mul nuw nsw <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id69">
+<h5><a class="toc-backref" href="#id796">Overview:</a><a class="headerlink" href="#id69" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">mul</span></tt>‘ instruction returns the product of its two operands.</p>
+</div>
+<div class="section" id="id70">
+<h5><a class="toc-backref" href="#id797">Arguments:</a><a class="headerlink" href="#id70" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">mul</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id71">
+<h5><a class="toc-backref" href="#id798">Semantics:</a><a class="headerlink" href="#id71" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the integer product of the two operands.</p>
+<p>If the result of the multiplication has unsigned overflow, the result
+returned is the mathematical result modulo 2<sup>n</sup>, where n is the
+bit width of the result.</p>
+<p>Because LLVM integers use a two’s complement representation, and the
+result is the same width as the operands, this instruction returns the
+correct result for both signed and unsigned integers. If a full product
+(e.g. <tt class="docutils literal"><span class="pre">i32</span></tt> * <tt class="docutils literal"><span class="pre">i32</span></tt> -> <tt class="docutils literal"><span class="pre">i64</span></tt>) is needed, the operands should be
+sign-extended or zero-extended as appropriate to the width of the full
+product.</p>
+<p><tt class="docutils literal"><span class="pre">nuw</span></tt> and <tt class="docutils literal"><span class="pre">nsw</span></tt> stand for “No Unsigned Wrap” and “No Signed Wrap”,
+respectively. If the <tt class="docutils literal"><span class="pre">nuw</span></tt> and/or <tt class="docutils literal"><span class="pre">nsw</span></tt> keywords are present, the
+result value of the <tt class="docutils literal"><span class="pre">mul</span></tt> is a <a class="reference internal" href="#poisonvalues"><em>poison value</em></a> if
+unsigned and/or signed overflow, respectively, occurs.</p>
+</div>
+<div class="section" id="id72">
+<h5><a class="toc-backref" href="#id799">Example:</a><a class="headerlink" href="#id72" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = mul i32 4, %var ; yields i32:result = 4 * %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fmul-instruction">
+<span id="i-fmul"></span><h4><a class="toc-backref" href="#id800">‘<tt class="docutils literal"><span class="pre">fmul</span></tt>‘ Instruction</a><a class="headerlink" href="#fmul-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id73">
+<h5><a class="toc-backref" href="#id801">Syntax:</a><a class="headerlink" href="#id73" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fmul [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id74">
+<h5><a class="toc-backref" href="#id802">Overview:</a><a class="headerlink" href="#id74" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fmul</span></tt>‘ instruction returns the product of its two operands.</p>
+</div>
+<div class="section" id="id75">
+<h5><a class="toc-backref" href="#id803">Arguments:</a><a class="headerlink" href="#id75" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">fmul</span></tt>‘ instruction must be <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of floating point values.
+Both arguments must have identical types.</p>
+</div>
+<div class="section" id="id76">
+<h5><a class="toc-backref" href="#id804">Semantics:</a><a class="headerlink" href="#id76" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the floating point product of the two operands.
+This instruction can also take any number of <a class="reference internal" href="#fastmath"><em>fast-math
+flags</em></a>, which are optimization hints to enable otherwise
+unsafe floating point optimizations:</p>
+</div>
+<div class="section" id="id77">
+<h5><a class="toc-backref" href="#id805">Example:</a><a class="headerlink" href="#id77" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = fmul float 4.0, %var ; yields float:result = 4.0 * %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="udiv-instruction">
+<h4><a class="toc-backref" href="#id806">‘<tt class="docutils literal"><span class="pre">udiv</span></tt>‘ Instruction</a><a class="headerlink" href="#udiv-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id78">
+<h5><a class="toc-backref" href="#id807">Syntax:</a><a class="headerlink" href="#id78" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = udiv <ty> <op1>, <op2> ; yields ty:result
+<result> = udiv exact <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id79">
+<h5><a class="toc-backref" href="#id808">Overview:</a><a class="headerlink" href="#id79" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">udiv</span></tt>‘ instruction returns the quotient of its two operands.</p>
+</div>
+<div class="section" id="id80">
+<h5><a class="toc-backref" href="#id809">Arguments:</a><a class="headerlink" href="#id80" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">udiv</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id81">
+<h5><a class="toc-backref" href="#id810">Semantics:</a><a class="headerlink" href="#id81" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the unsigned integer quotient of the two operands.</p>
+<p>Note that unsigned integer division and signed integer division are
+distinct operations; for signed integer division, use ‘<tt class="docutils literal"><span class="pre">sdiv</span></tt>‘.</p>
+<p>Division by zero leads to undefined behavior.</p>
+<p>If the <tt class="docutils literal"><span class="pre">exact</span></tt> keyword is present, the result value of the <tt class="docutils literal"><span class="pre">udiv</span></tt> is
+a <a class="reference internal" href="#poisonvalues"><em>poison value</em></a> if %op1 is not a multiple of %op2 (as
+such, “((a udiv exact b) mul b) == a”).</p>
+</div>
+<div class="section" id="id82">
+<h5><a class="toc-backref" href="#id811">Example:</a><a class="headerlink" href="#id82" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = udiv i32 4, %var ; yields i32:result = 4 / %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="sdiv-instruction">
+<h4><a class="toc-backref" href="#id812">‘<tt class="docutils literal"><span class="pre">sdiv</span></tt>‘ Instruction</a><a class="headerlink" href="#sdiv-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id83">
+<h5><a class="toc-backref" href="#id813">Syntax:</a><a class="headerlink" href="#id83" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = sdiv <ty> <op1>, <op2> ; yields ty:result
+<result> = sdiv exact <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id84">
+<h5><a class="toc-backref" href="#id814">Overview:</a><a class="headerlink" href="#id84" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">sdiv</span></tt>‘ instruction returns the quotient of its two operands.</p>
+</div>
+<div class="section" id="id85">
+<h5><a class="toc-backref" href="#id815">Arguments:</a><a class="headerlink" href="#id85" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">sdiv</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id86">
+<h5><a class="toc-backref" href="#id816">Semantics:</a><a class="headerlink" href="#id86" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the signed integer quotient of the two operands
+rounded towards zero.</p>
+<p>Note that signed integer division and unsigned integer division are
+distinct operations; for unsigned integer division, use ‘<tt class="docutils literal"><span class="pre">udiv</span></tt>‘.</p>
+<p>Division by zero leads to undefined behavior. Overflow also leads to
+undefined behavior; this is a rare case, but can occur, for example, by
+doing a 32-bit division of -2147483648 by -1.</p>
+<p>If the <tt class="docutils literal"><span class="pre">exact</span></tt> keyword is present, the result value of the <tt class="docutils literal"><span class="pre">sdiv</span></tt> is
+a <a class="reference internal" href="#poisonvalues"><em>poison value</em></a> if the result would be rounded.</p>
+</div>
+<div class="section" id="id87">
+<h5><a class="toc-backref" href="#id817">Example:</a><a class="headerlink" href="#id87" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = sdiv i32 4, %var ; yields i32:result = 4 / %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fdiv-instruction">
+<span id="i-fdiv"></span><h4><a class="toc-backref" href="#id818">‘<tt class="docutils literal"><span class="pre">fdiv</span></tt>‘ Instruction</a><a class="headerlink" href="#fdiv-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id88">
+<h5><a class="toc-backref" href="#id819">Syntax:</a><a class="headerlink" href="#id88" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fdiv [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id89">
+<h5><a class="toc-backref" href="#id820">Overview:</a><a class="headerlink" href="#id89" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fdiv</span></tt>‘ instruction returns the quotient of its two operands.</p>
+</div>
+<div class="section" id="id90">
+<h5><a class="toc-backref" href="#id821">Arguments:</a><a class="headerlink" href="#id90" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">fdiv</span></tt>‘ instruction must be <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of floating point values.
+Both arguments must have identical types.</p>
+</div>
+<div class="section" id="id91">
+<h5><a class="toc-backref" href="#id822">Semantics:</a><a class="headerlink" href="#id91" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is the floating point quotient of the two operands.
+This instruction can also take any number of <a class="reference internal" href="#fastmath"><em>fast-math
+flags</em></a>, which are optimization hints to enable otherwise
+unsafe floating point optimizations:</p>
+</div>
+<div class="section" id="id92">
+<h5><a class="toc-backref" href="#id823">Example:</a><a class="headerlink" href="#id92" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = fdiv float 4.0, %var ; yields float:result = 4.0 / %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="urem-instruction">
+<h4><a class="toc-backref" href="#id824">‘<tt class="docutils literal"><span class="pre">urem</span></tt>‘ Instruction</a><a class="headerlink" href="#urem-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id93">
+<h5><a class="toc-backref" href="#id825">Syntax:</a><a class="headerlink" href="#id93" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = urem <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id94">
+<h5><a class="toc-backref" href="#id826">Overview:</a><a class="headerlink" href="#id94" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">urem</span></tt>‘ instruction returns the remainder from the unsigned
+division of its two arguments.</p>
+</div>
+<div class="section" id="id95">
+<h5><a class="toc-backref" href="#id827">Arguments:</a><a class="headerlink" href="#id95" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">urem</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id96">
+<h5><a class="toc-backref" href="#id828">Semantics:</a><a class="headerlink" href="#id96" title="Permalink to this headline">¶</a></h5>
+<p>This instruction returns the unsigned integer <em>remainder</em> of a division.
+This instruction always performs an unsigned division to get the
+remainder.</p>
+<p>Note that unsigned integer remainder and signed integer remainder are
+distinct operations; for signed integer remainder, use ‘<tt class="docutils literal"><span class="pre">srem</span></tt>‘.</p>
+<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
+</div>
+<div class="section" id="id97">
+<h5><a class="toc-backref" href="#id829">Example:</a><a class="headerlink" href="#id97" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = urem i32 4, %var ; yields i32:result = 4 % %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="srem-instruction">
+<h4><a class="toc-backref" href="#id830">‘<tt class="docutils literal"><span class="pre">srem</span></tt>‘ Instruction</a><a class="headerlink" href="#srem-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id98">
+<h5><a class="toc-backref" href="#id831">Syntax:</a><a class="headerlink" href="#id98" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = srem <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id99">
+<h5><a class="toc-backref" href="#id832">Overview:</a><a class="headerlink" href="#id99" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">srem</span></tt>‘ instruction returns the remainder from the signed
+division of its two operands. This instruction can also take
+<a class="reference internal" href="#t-vector"><em>vector</em></a> versions of the values in which case the elements
+must be integers.</p>
+</div>
+<div class="section" id="id100">
+<h5><a class="toc-backref" href="#id833">Arguments:</a><a class="headerlink" href="#id100" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">srem</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id101">
+<h5><a class="toc-backref" href="#id834">Semantics:</a><a class="headerlink" href="#id101" title="Permalink to this headline">¶</a></h5>
+<p>This instruction returns the <em>remainder</em> of a division (where the result
+is either zero or has the same sign as the dividend, <tt class="docutils literal"><span class="pre">op1</span></tt>), not the
+<em>modulo</em> operator (where the result is either zero or has the same sign
+as the divisor, <tt class="docutils literal"><span class="pre">op2</span></tt>) of a value. For more information about the
+difference, see <a class="reference external" href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The Math
+Forum</a>. For a
+table of how this is implemented in various languages, please see
+<a class="reference external" href="http://en.wikipedia.org/wiki/Modulo_operation">Wikipedia: modulo
+operation</a>.</p>
+<p>Note that signed integer remainder and unsigned integer remainder are
+distinct operations; for unsigned integer remainder, use ‘<tt class="docutils literal"><span class="pre">urem</span></tt>‘.</p>
+<p>Taking the remainder of a division by zero leads to undefined behavior.
+Overflow also leads to undefined behavior; this is a rare case, but can
+occur, for example, by taking the remainder of a 32-bit division of
+-2147483648 by -1. (The remainder doesn’t actually overflow, but this
+rule lets srem be implemented using instructions that return both the
+result of the division and the remainder.)</p>
+</div>
+<div class="section" id="id102">
+<h5><a class="toc-backref" href="#id835">Example:</a><a class="headerlink" href="#id102" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = srem i32 4, %var ; yields i32:result = 4 % %var
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="frem-instruction">
+<span id="i-frem"></span><h4><a class="toc-backref" href="#id836">‘<tt class="docutils literal"><span class="pre">frem</span></tt>‘ Instruction</a><a class="headerlink" href="#frem-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id103">
+<h5><a class="toc-backref" href="#id837">Syntax:</a><a class="headerlink" href="#id103" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = frem [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id104">
+<h5><a class="toc-backref" href="#id838">Overview:</a><a class="headerlink" href="#id104" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">frem</span></tt>‘ instruction returns the remainder from the division of
+its two operands.</p>
+</div>
+<div class="section" id="id105">
+<h5><a class="toc-backref" href="#id839">Arguments:</a><a class="headerlink" href="#id105" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">frem</span></tt>‘ instruction must be <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of floating point values.
+Both arguments must have identical types.</p>
+</div>
+<div class="section" id="id106">
+<h5><a class="toc-backref" href="#id840">Semantics:</a><a class="headerlink" href="#id106" title="Permalink to this headline">¶</a></h5>
+<p>This instruction returns the <em>remainder</em> of a division. The remainder
+has the same sign as the dividend. This instruction can also take any
+number of <a class="reference internal" href="#fastmath"><em>fast-math flags</em></a>, which are optimization hints
+to enable otherwise unsafe floating point optimizations:</p>
+</div>
+<div class="section" id="id107">
+<h5><a class="toc-backref" href="#id841">Example:</a><a class="headerlink" href="#id107" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = frem float 4.0, %var ; yields float:result = 4.0 % %var
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="bitwise-binary-operations">
+<span id="bitwiseops"></span><h3><a class="toc-backref" href="#id842">Bitwise Binary Operations</a><a class="headerlink" href="#bitwise-binary-operations" title="Permalink to this headline">¶</a></h3>
+<p>Bitwise binary operators are used to do various forms of bit-twiddling
+in a program. They are generally very efficient instructions and can
+commonly be strength reduced from other instructions. They require two
+operands of the same type, execute an operation on them, and produce a
+single value. The resulting value is the same type as its operands.</p>
+<div class="section" id="shl-instruction">
+<h4><a class="toc-backref" href="#id843">‘<tt class="docutils literal"><span class="pre">shl</span></tt>‘ Instruction</a><a class="headerlink" href="#shl-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id108">
+<h5><a class="toc-backref" href="#id844">Syntax:</a><a class="headerlink" href="#id108" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = shl <ty> <op1>, <op2> ; yields ty:result
+<result> = shl nuw <ty> <op1>, <op2> ; yields ty:result
+<result> = shl nsw <ty> <op1>, <op2> ; yields ty:result
+<result> = shl nuw nsw <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id109">
+<h5><a class="toc-backref" href="#id845">Overview:</a><a class="headerlink" href="#id109" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">shl</span></tt>‘ instruction returns the first operand shifted to the left
+a specified number of bits.</p>
+</div>
+<div class="section" id="id110">
+<h5><a class="toc-backref" href="#id846">Arguments:</a><a class="headerlink" href="#id110" title="Permalink to this headline">¶</a></h5>
+<p>Both arguments to the ‘<tt class="docutils literal"><span class="pre">shl</span></tt>‘ instruction must be the same
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer type.
+‘<tt class="docutils literal"><span class="pre">op2</span></tt>‘ is treated as an unsigned value.</p>
+</div>
+<div class="section" id="id111">
+<h5><a class="toc-backref" href="#id847">Semantics:</a><a class="headerlink" href="#id111" title="Permalink to this headline">¶</a></h5>
+<p>The value produced is <tt class="docutils literal"><span class="pre">op1</span></tt> * 2<sup>op2</sup> mod 2<sup>n</sup>,
+where <tt class="docutils literal"><span class="pre">n</span></tt> is the width of the result. If <tt class="docutils literal"><span class="pre">op2</span></tt> is (statically or
+dynamically) equal to or larger than the number of bits in
+<tt class="docutils literal"><span class="pre">op1</span></tt>, the result is undefined. If the arguments are vectors, each
+vector element of <tt class="docutils literal"><span class="pre">op1</span></tt> is shifted by the corresponding shift amount
+in <tt class="docutils literal"><span class="pre">op2</span></tt>.</p>
+<p>If the <tt class="docutils literal"><span class="pre">nuw</span></tt> keyword is present, then the shift produces a <a class="reference internal" href="#poisonvalues"><em>poison
+value</em></a> if it shifts out any non-zero bits. If the
+<tt class="docutils literal"><span class="pre">nsw</span></tt> keyword is present, then the shift produces a <a class="reference internal" href="#poisonvalues"><em>poison
+value</em></a> if it shifts out any bits that disagree with the
+resultant sign bit. As such, NUW/NSW have the same semantics as they
+would if the shift were expressed as a mul instruction with the same
+nsw/nuw bits in (mul %op1, (shl 1, %op2)).</p>
+</div>
+<div class="section" id="id112">
+<h5><a class="toc-backref" href="#id848">Example:</a><a class="headerlink" href="#id112" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = shl i32 4, %var ; yields i32: 4 << %var
+<result> = shl i32 4, 2 ; yields i32: 16
+<result> = shl i32 1, 10 ; yields i32: 1024
+<result> = shl i32 1, 32 ; undefined
+<result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 2, i32 4>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="lshr-instruction">
+<h4><a class="toc-backref" href="#id849">‘<tt class="docutils literal"><span class="pre">lshr</span></tt>‘ Instruction</a><a class="headerlink" href="#lshr-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id113">
+<h5><a class="toc-backref" href="#id850">Syntax:</a><a class="headerlink" href="#id113" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = lshr <ty> <op1>, <op2> ; yields ty:result
+<result> = lshr exact <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id114">
+<h5><a class="toc-backref" href="#id851">Overview:</a><a class="headerlink" href="#id114" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">lshr</span></tt>‘ instruction (logical shift right) returns the first
+operand shifted to the right a specified number of bits with zero fill.</p>
+</div>
+<div class="section" id="id115">
+<h5><a class="toc-backref" href="#id852">Arguments:</a><a class="headerlink" href="#id115" title="Permalink to this headline">¶</a></h5>
+<p>Both arguments to the ‘<tt class="docutils literal"><span class="pre">lshr</span></tt>‘ instruction must be the same
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer type.
+‘<tt class="docutils literal"><span class="pre">op2</span></tt>‘ is treated as an unsigned value.</p>
+</div>
+<div class="section" id="id116">
+<h5><a class="toc-backref" href="#id853">Semantics:</a><a class="headerlink" href="#id116" title="Permalink to this headline">¶</a></h5>
+<p>This instruction always performs a logical shift right operation. The
+most significant bits of the result will be filled with zero bits after
+the shift. If <tt class="docutils literal"><span class="pre">op2</span></tt> is (statically or dynamically) equal to or larger
+than the number of bits in <tt class="docutils literal"><span class="pre">op1</span></tt>, the result is undefined. If the
+arguments are vectors, each vector element of <tt class="docutils literal"><span class="pre">op1</span></tt> is shifted by the
+corresponding shift amount in <tt class="docutils literal"><span class="pre">op2</span></tt>.</p>
+<p>If the <tt class="docutils literal"><span class="pre">exact</span></tt> keyword is present, the result value of the <tt class="docutils literal"><span class="pre">lshr</span></tt> is
+a <a class="reference internal" href="#poisonvalues"><em>poison value</em></a> if any of the bits shifted out are
+non-zero.</p>
+</div>
+<div class="section" id="id117">
+<h5><a class="toc-backref" href="#id854">Example:</a><a class="headerlink" href="#id117" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = lshr i32 4, 1 ; yields i32:result = 2
+<result> = lshr i32 4, 2 ; yields i32:result = 1
+<result> = lshr i8 4, 3 ; yields i8:result = 0
+<result> = lshr i8 -2, 1 ; yields i8:result = 0x7F
+<result> = lshr i32 1, 32 ; undefined
+<result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> ; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="ashr-instruction">
+<h4><a class="toc-backref" href="#id855">‘<tt class="docutils literal"><span class="pre">ashr</span></tt>‘ Instruction</a><a class="headerlink" href="#ashr-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id118">
+<h5><a class="toc-backref" href="#id856">Syntax:</a><a class="headerlink" href="#id118" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = ashr <ty> <op1>, <op2> ; yields ty:result
+<result> = ashr exact <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id119">
+<h5><a class="toc-backref" href="#id857">Overview:</a><a class="headerlink" href="#id119" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">ashr</span></tt>‘ instruction (arithmetic shift right) returns the first
+operand shifted to the right a specified number of bits with sign
+extension.</p>
+</div>
+<div class="section" id="id120">
+<h5><a class="toc-backref" href="#id858">Arguments:</a><a class="headerlink" href="#id120" title="Permalink to this headline">¶</a></h5>
+<p>Both arguments to the ‘<tt class="docutils literal"><span class="pre">ashr</span></tt>‘ instruction must be the same
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer type.
+‘<tt class="docutils literal"><span class="pre">op2</span></tt>‘ is treated as an unsigned value.</p>
+</div>
+<div class="section" id="id121">
+<h5><a class="toc-backref" href="#id859">Semantics:</a><a class="headerlink" href="#id121" title="Permalink to this headline">¶</a></h5>
+<p>This instruction always performs an arithmetic shift right operation,
+The most significant bits of the result will be filled with the sign bit
+of <tt class="docutils literal"><span class="pre">op1</span></tt>. If <tt class="docutils literal"><span class="pre">op2</span></tt> is (statically or dynamically) equal to or larger
+than the number of bits in <tt class="docutils literal"><span class="pre">op1</span></tt>, the result is undefined. If the
+arguments are vectors, each vector element of <tt class="docutils literal"><span class="pre">op1</span></tt> is shifted by the
+corresponding shift amount in <tt class="docutils literal"><span class="pre">op2</span></tt>.</p>
+<p>If the <tt class="docutils literal"><span class="pre">exact</span></tt> keyword is present, the result value of the <tt class="docutils literal"><span class="pre">ashr</span></tt> is
+a <a class="reference internal" href="#poisonvalues"><em>poison value</em></a> if any of the bits shifted out are
+non-zero.</p>
+</div>
+<div class="section" id="id122">
+<h5><a class="toc-backref" href="#id860">Example:</a><a class="headerlink" href="#id122" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = ashr i32 4, 1 ; yields i32:result = 2
+<result> = ashr i32 4, 2 ; yields i32:result = 1
+<result> = ashr i8 4, 3 ; yields i8:result = 0
+<result> = ashr i8 -2, 1 ; yields i8:result = -1
+<result> = ashr i32 1, 32 ; undefined
+<result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> ; yields: result=<2 x i32> < i32 -1, i32 0>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="and-instruction">
+<h4><a class="toc-backref" href="#id861">‘<tt class="docutils literal"><span class="pre">and</span></tt>‘ Instruction</a><a class="headerlink" href="#and-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id123">
+<h5><a class="toc-backref" href="#id862">Syntax:</a><a class="headerlink" href="#id123" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = and <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id124">
+<h5><a class="toc-backref" href="#id863">Overview:</a><a class="headerlink" href="#id124" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">and</span></tt>‘ instruction returns the bitwise logical and of its two
+operands.</p>
+</div>
+<div class="section" id="id125">
+<h5><a class="toc-backref" href="#id864">Arguments:</a><a class="headerlink" href="#id125" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">and</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id126">
+<h5><a class="toc-backref" href="#id865">Semantics:</a><a class="headerlink" href="#id126" title="Permalink to this headline">¶</a></h5>
+<p>The truth table used for the ‘<tt class="docutils literal"><span class="pre">and</span></tt>‘ instruction is:</p>
+<table border="1" class="docutils">
+<colgroup>
+<col width="33%" />
+<col width="33%" />
+<col width="33%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td>In0</td>
+<td>In1</td>
+<td>Out</td>
+</tr>
+<tr class="row-even"><td>0</td>
+<td>0</td>
+<td>0</td>
+</tr>
+<tr class="row-odd"><td>0</td>
+<td>1</td>
+<td>0</td>
+</tr>
+<tr class="row-even"><td>1</td>
+<td>0</td>
+<td>0</td>
+</tr>
+<tr class="row-odd"><td>1</td>
+<td>1</td>
+<td>1</td>
+</tr>
+</tbody>
+</table>
+</div>
+<div class="section" id="id127">
+<h5><a class="toc-backref" href="#id866">Example:</a><a class="headerlink" href="#id127" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = and i32 4, %var ; yields i32:result = 4 & %var
+<result> = and i32 15, 40 ; yields i32:result = 8
+<result> = and i32 4, 8 ; yields i32:result = 0
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="or-instruction">
+<h4><a class="toc-backref" href="#id867">‘<tt class="docutils literal"><span class="pre">or</span></tt>‘ Instruction</a><a class="headerlink" href="#or-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id128">
+<h5><a class="toc-backref" href="#id868">Syntax:</a><a class="headerlink" href="#id128" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = or <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id129">
+<h5><a class="toc-backref" href="#id869">Overview:</a><a class="headerlink" href="#id129" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">or</span></tt>‘ instruction returns the bitwise logical inclusive or of its
+two operands.</p>
+</div>
+<div class="section" id="id130">
+<h5><a class="toc-backref" href="#id870">Arguments:</a><a class="headerlink" href="#id130" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">or</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id131">
+<h5><a class="toc-backref" href="#id871">Semantics:</a><a class="headerlink" href="#id131" title="Permalink to this headline">¶</a></h5>
+<p>The truth table used for the ‘<tt class="docutils literal"><span class="pre">or</span></tt>‘ instruction is:</p>
+<table border="1" class="docutils">
+<colgroup>
+<col width="33%" />
+<col width="33%" />
+<col width="33%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td>In0</td>
+<td>In1</td>
+<td>Out</td>
+</tr>
+<tr class="row-even"><td>0</td>
+<td>0</td>
+<td>0</td>
+</tr>
+<tr class="row-odd"><td>0</td>
+<td>1</td>
+<td>1</td>
+</tr>
+<tr class="row-even"><td>1</td>
+<td>0</td>
+<td>1</td>
+</tr>
+<tr class="row-odd"><td>1</td>
+<td>1</td>
+<td>1</td>
+</tr>
+</tbody>
+</table>
+</div>
+<div class="section" id="id132">
+<h5><a class="toc-backref" href="#id872">Example:</a><a class="headerlink" href="#id132" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = or i32 4, %var ; yields i32:result = 4 | %var
+<result> = or i32 15, 40 ; yields i32:result = 47
+<result> = or i32 4, 8 ; yields i32:result = 12
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="xor-instruction">
+<h4><a class="toc-backref" href="#id873">‘<tt class="docutils literal"><span class="pre">xor</span></tt>‘ Instruction</a><a class="headerlink" href="#xor-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id133">
+<h5><a class="toc-backref" href="#id874">Syntax:</a><a class="headerlink" href="#id133" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = xor <ty> <op1>, <op2> ; yields ty:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id134">
+<h5><a class="toc-backref" href="#id875">Overview:</a><a class="headerlink" href="#id134" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">xor</span></tt>‘ instruction returns the bitwise logical exclusive or of
+its two operands. The <tt class="docutils literal"><span class="pre">xor</span></tt> is used to implement the “one’s
+complement” operation, which is the “~” operator in C.</p>
+</div>
+<div class="section" id="id135">
+<h5><a class="toc-backref" href="#id876">Arguments:</a><a class="headerlink" href="#id135" title="Permalink to this headline">¶</a></h5>
+<p>The two arguments to the ‘<tt class="docutils literal"><span class="pre">xor</span></tt>‘ instruction must be
+<a class="reference internal" href="#t-integer"><em>integer</em></a> or <a class="reference internal" href="#t-vector"><em>vector</em></a> of integer values. Both
+arguments must have identical types.</p>
+</div>
+<div class="section" id="id136">
+<h5><a class="toc-backref" href="#id877">Semantics:</a><a class="headerlink" href="#id136" title="Permalink to this headline">¶</a></h5>
+<p>The truth table used for the ‘<tt class="docutils literal"><span class="pre">xor</span></tt>‘ instruction is:</p>
+<table border="1" class="docutils">
+<colgroup>
+<col width="33%" />
+<col width="33%" />
+<col width="33%" />
+</colgroup>
+<tbody valign="top">
+<tr class="row-odd"><td>In0</td>
+<td>In1</td>
+<td>Out</td>
+</tr>
+<tr class="row-even"><td>0</td>
+<td>0</td>
+<td>0</td>
+</tr>
+<tr class="row-odd"><td>0</td>
+<td>1</td>
+<td>1</td>
+</tr>
+<tr class="row-even"><td>1</td>
+<td>0</td>
+<td>1</td>
+</tr>
+<tr class="row-odd"><td>1</td>
+<td>1</td>
+<td>0</td>
+</tr>
+</tbody>
+</table>
+</div>
+<div class="section" id="id137">
+<h5><a class="toc-backref" href="#id878">Example:</a><a class="headerlink" href="#id137" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = xor i32 4, %var ; yields i32:result = 4 ^ %var
+<result> = xor i32 15, 40 ; yields i32:result = 39
+<result> = xor i32 4, 8 ; yields i32:result = 12
+<result> = xor i32 %V, -1 ; yields i32:result = ~%V
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="vector-operations">
+<h3><a class="toc-backref" href="#id879">Vector Operations</a><a class="headerlink" href="#vector-operations" title="Permalink to this headline">¶</a></h3>
+<p>LLVM supports several instructions to represent vector operations in a
+target-independent manner. These instructions cover the element-access
+and vector-specific operations needed to process vectors effectively.
+While LLVM does directly support these vector operations, many
+sophisticated algorithms will want to use target-specific intrinsics to
+take full advantage of a specific target.</p>
+<div class="section" id="extractelement-instruction">
+<span id="i-extractelement"></span><h4><a class="toc-backref" href="#id880">‘<tt class="docutils literal"><span class="pre">extractelement</span></tt>‘ Instruction</a><a class="headerlink" href="#extractelement-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id138">
+<h5><a class="toc-backref" href="#id881">Syntax:</a><a class="headerlink" href="#id138" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = extractelement <n x <ty>> <val>, <ty2> <idx> ; yields <ty>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id139">
+<h5><a class="toc-backref" href="#id882">Overview:</a><a class="headerlink" href="#id139" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">extractelement</span></tt>‘ instruction extracts a single scalar element
+from a vector at a specified index.</p>
+</div>
+<div class="section" id="id140">
+<h5><a class="toc-backref" href="#id883">Arguments:</a><a class="headerlink" href="#id140" title="Permalink to this headline">¶</a></h5>
+<p>The first operand of an ‘<tt class="docutils literal"><span class="pre">extractelement</span></tt>‘ instruction is a value of
+<a class="reference internal" href="#t-vector"><em>vector</em></a> type. The second operand is an index indicating
+the position from which to extract the element. The index may be a
+variable of any integer type.</p>
+</div>
+<div class="section" id="id141">
+<h5><a class="toc-backref" href="#id884">Semantics:</a><a class="headerlink" href="#id141" title="Permalink to this headline">¶</a></h5>
+<p>The result is a scalar of the same type as the element type of <tt class="docutils literal"><span class="pre">val</span></tt>.
+Its value is the value at position <tt class="docutils literal"><span class="pre">idx</span></tt> of <tt class="docutils literal"><span class="pre">val</span></tt>. If <tt class="docutils literal"><span class="pre">idx</span></tt>
+exceeds the length of <tt class="docutils literal"><span class="pre">val</span></tt>, the results are undefined.</p>
+</div>
+<div class="section" id="id142">
+<h5><a class="toc-backref" href="#id885">Example:</a><a class="headerlink" href="#id142" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = extractelement <4 x i32> %vec, i32 0 ; yields i32
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="insertelement-instruction">
+<span id="i-insertelement"></span><h4><a class="toc-backref" href="#id886">‘<tt class="docutils literal"><span class="pre">insertelement</span></tt>‘ Instruction</a><a class="headerlink" href="#insertelement-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id143">
+<h5><a class="toc-backref" href="#id887">Syntax:</a><a class="headerlink" href="#id143" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = insertelement <n x <ty>> <val>, <ty> <elt>, <ty2> <idx> ; yields <n x <ty>>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id144">
+<h5><a class="toc-backref" href="#id888">Overview:</a><a class="headerlink" href="#id144" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">insertelement</span></tt>‘ instruction inserts a scalar element into a
+vector at a specified index.</p>
+</div>
+<div class="section" id="id145">
+<h5><a class="toc-backref" href="#id889">Arguments:</a><a class="headerlink" href="#id145" title="Permalink to this headline">¶</a></h5>
+<p>The first operand of an ‘<tt class="docutils literal"><span class="pre">insertelement</span></tt>‘ instruction is a value of
+<a class="reference internal" href="#t-vector"><em>vector</em></a> type. The second operand is a scalar value whose
+type must equal the element type of the first operand. The third operand
+is an index indicating the position at which to insert the value. The
+index may be a variable of any integer type.</p>
+</div>
+<div class="section" id="id146">
+<h5><a class="toc-backref" href="#id890">Semantics:</a><a class="headerlink" href="#id146" title="Permalink to this headline">¶</a></h5>
+<p>The result is a vector of the same type as <tt class="docutils literal"><span class="pre">val</span></tt>. Its element values
+are those of <tt class="docutils literal"><span class="pre">val</span></tt> except at position <tt class="docutils literal"><span class="pre">idx</span></tt>, where it gets the value
+<tt class="docutils literal"><span class="pre">elt</span></tt>. If <tt class="docutils literal"><span class="pre">idx</span></tt> exceeds the length of <tt class="docutils literal"><span class="pre">val</span></tt>, the results are
+undefined.</p>
+</div>
+<div class="section" id="id147">
+<h5><a class="toc-backref" href="#id891">Example:</a><a class="headerlink" href="#id147" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = insertelement <4 x i32> %vec, i32 1, i32 0 ; yields <4 x i32>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="shufflevector-instruction">
+<span id="i-shufflevector"></span><h4><a class="toc-backref" href="#id892">‘<tt class="docutils literal"><span class="pre">shufflevector</span></tt>‘ Instruction</a><a class="headerlink" href="#shufflevector-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id148">
+<h5><a class="toc-backref" href="#id893">Syntax:</a><a class="headerlink" href="#id148" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> ; yields <m x <ty>>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id149">
+<h5><a class="toc-backref" href="#id894">Overview:</a><a class="headerlink" href="#id149" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">shufflevector</span></tt>‘ instruction constructs a permutation of elements
+from two input vectors, returning a vector with the same element type as
+the input and length that is the same as the shuffle mask.</p>
+</div>
+<div class="section" id="id150">
+<h5><a class="toc-backref" href="#id895">Arguments:</a><a class="headerlink" href="#id150" title="Permalink to this headline">¶</a></h5>
+<p>The first two operands of a ‘<tt class="docutils literal"><span class="pre">shufflevector</span></tt>‘ instruction are vectors
+with the same type. The third argument is a shuffle mask whose element
+type is always ‘i32’. The result of the instruction is a vector whose
+length is the same as the shuffle mask and whose element type is the
+same as the element type of the first two operands.</p>
+<p>The shuffle mask operand is required to be a constant vector with either
+constant integer or undef values.</p>
+</div>
+<div class="section" id="id151">
+<h5><a class="toc-backref" href="#id896">Semantics:</a><a class="headerlink" href="#id151" title="Permalink to this headline">¶</a></h5>
+<p>The elements of the two input vectors are numbered from left to right
+across both of the vectors. The shuffle mask operand specifies, for each
+element of the result vector, which element of the two input vectors the
+result element gets. The element selector may be undef (meaning “don’t
+care”) and the second operand may be undef if performing a shuffle from
+only one vector.</p>
+</div>
+<div class="section" id="id152">
+<h5><a class="toc-backref" href="#id897">Example:</a><a class="headerlink" href="#id152" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
+ <4 x i32> <i32 0, i32 4, i32 1, i32 5> ; yields <4 x i32>
+<result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
+ <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32> - Identity shuffle.
+<result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
+ <4 x i32> <i32 0, i32 1, i32 2, i32 3> ; yields <4 x i32>
+<result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
+ <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 > ; yields <8 x i32>
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="aggregate-operations">
+<h3><a class="toc-backref" href="#id898">Aggregate Operations</a><a class="headerlink" href="#aggregate-operations" title="Permalink to this headline">¶</a></h3>
+<p>LLVM supports several instructions for working with
+<a class="reference internal" href="#t-aggregate"><em>aggregate</em></a> values.</p>
+<div class="section" id="extractvalue-instruction">
+<span id="i-extractvalue"></span><h4><a class="toc-backref" href="#id899">‘<tt class="docutils literal"><span class="pre">extractvalue</span></tt>‘ Instruction</a><a class="headerlink" href="#extractvalue-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id153">
+<h5><a class="toc-backref" href="#id900">Syntax:</a><a class="headerlink" href="#id153" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
+</pre></div>
+</div>
+</div>
+<div class="section" id="id154">
+<h5><a class="toc-backref" href="#id901">Overview:</a><a class="headerlink" href="#id154" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">extractvalue</span></tt>‘ instruction extracts the value of a member field
+from an <a class="reference internal" href="#t-aggregate"><em>aggregate</em></a> value.</p>
+</div>
+<div class="section" id="id155">
+<h5><a class="toc-backref" href="#id902">Arguments:</a><a class="headerlink" href="#id155" title="Permalink to this headline">¶</a></h5>
+<p>The first operand of an ‘<tt class="docutils literal"><span class="pre">extractvalue</span></tt>‘ instruction is a value of
+<a class="reference internal" href="#t-struct"><em>struct</em></a> or <a class="reference internal" href="#t-array"><em>array</em></a> type. The operands are
+constant indices to specify which value to extract in a similar manner
+as indices in a ‘<tt class="docutils literal"><span class="pre">getelementptr</span></tt>‘ instruction.</p>
+<p>The major differences to <tt class="docutils literal"><span class="pre">getelementptr</span></tt> indexing are:</p>
+<ul class="simple">
+<li>Since the value being indexed is not a pointer, the first index is
+omitted and assumed to be zero.</li>
+<li>At least one index must be specified.</li>
+<li>Not only struct indices but also array indices must be in bounds.</li>
+</ul>
+</div>
+<div class="section" id="id156">
+<h5><a class="toc-backref" href="#id903">Semantics:</a><a class="headerlink" href="#id156" title="Permalink to this headline">¶</a></h5>
+<p>The result is the value at the position in the aggregate specified by
+the index operands.</p>
+</div>
+<div class="section" id="id157">
+<h5><a class="toc-backref" href="#id904">Example:</a><a class="headerlink" href="#id157" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = extractvalue {i32, float} %agg, 0 ; yields i32
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="insertvalue-instruction">
+<span id="i-insertvalue"></span><h4><a class="toc-backref" href="#id905">‘<tt class="docutils literal"><span class="pre">insertvalue</span></tt>‘ Instruction</a><a class="headerlink" href="#insertvalue-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id158">
+<h5><a class="toc-backref" href="#id906">Syntax:</a><a class="headerlink" href="#id158" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* ; yields <aggregate type>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id159">
+<h5><a class="toc-backref" href="#id907">Overview:</a><a class="headerlink" href="#id159" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">insertvalue</span></tt>‘ instruction inserts a value into a member field in
+an <a class="reference internal" href="#t-aggregate"><em>aggregate</em></a> value.</p>
+</div>
+<div class="section" id="id160">
+<h5><a class="toc-backref" href="#id908">Arguments:</a><a class="headerlink" href="#id160" title="Permalink to this headline">¶</a></h5>
+<p>The first operand of an ‘<tt class="docutils literal"><span class="pre">insertvalue</span></tt>‘ instruction is a value of
+<a class="reference internal" href="#t-struct"><em>struct</em></a> or <a class="reference internal" href="#t-array"><em>array</em></a> type. The second operand is
+a first-class value to insert. The following operands are constant
+indices indicating the position at which to insert the value in a
+similar manner as indices in a ‘<tt class="docutils literal"><span class="pre">extractvalue</span></tt>‘ instruction. The value
+to insert must have the same type as the value identified by the
+indices.</p>
+</div>
+<div class="section" id="id161">
+<h5><a class="toc-backref" href="#id909">Semantics:</a><a class="headerlink" href="#id161" title="Permalink to this headline">¶</a></h5>
+<p>The result is an aggregate of the same type as <tt class="docutils literal"><span class="pre">val</span></tt>. Its value is
+that of <tt class="docutils literal"><span class="pre">val</span></tt> except that the value at the position specified by the
+indices is that of <tt class="docutils literal"><span class="pre">elt</span></tt>.</p>
+</div>
+<div class="section" id="id162">
+<h5><a class="toc-backref" href="#id910">Example:</a><a class="headerlink" href="#id162" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%agg1</span> <span class="p">=</span> <span class="k">insertvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="kt">float</span><span class="p">}</span> <span class="k">undef</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="m">0</span> <span class="c">; yields {i32 1, float undef}</span>
+<span class="nv">%agg2</span> <span class="p">=</span> <span class="k">insertvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="kt">float</span><span class="p">}</span> <span class="nv">%agg1</span><span class="p">,</span> <span class="kt">float</span> <span class="nv">%val</span><span class="p">,</span> <span class="m">1</span> <span class="c">; yields {i32 1, float %val}</span>
+<span class="nv">%agg3</span> <span class="p">=</span> <span class="k">insertvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="p">{</span><span class="kt">float</span><span class="p">}}</span> <span class="k">undef</span><span class="p">,</span> <span class="kt">float</span> <span class="nv">%val</span><span class="p">,</span> <span class="m">1</span><span class="p">,</span> <span class="m">0</span> <span class="c">; yields {i32 undef, {float %val}}</span>
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="memory-access-and-addressing-operations">
+<span id="memoryops"></span><h3><a class="toc-backref" href="#id911">Memory Access and Addressing Operations</a><a class="headerlink" href="#memory-access-and-addressing-operations" title="Permalink to this headline">¶</a></h3>
+<p>A key design point of an SSA-based representation is how it represents
+memory. In LLVM, no memory locations are in SSA form, which makes things
+very simple. This section describes how to read, write, and allocate
+memory in LLVM.</p>
+<div class="section" id="alloca-instruction">
+<span id="i-alloca"></span><h4><a class="toc-backref" href="#id912">‘<tt class="docutils literal"><span class="pre">alloca</span></tt>‘ Instruction</a><a class="headerlink" href="#alloca-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id163">
+<h5><a class="toc-backref" href="#id913">Syntax:</a><a class="headerlink" href="#id163" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = alloca [inalloca] <type> [, <ty> <NumElements>] [, align <alignment>] ; yields type*:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id164">
+<h5><a class="toc-backref" href="#id914">Overview:</a><a class="headerlink" href="#id164" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">alloca</span></tt>‘ instruction allocates memory on the stack frame of the
+currently executing function, to be automatically released when this
+function returns to its caller. The object is always allocated in the
+generic address space (address space zero).</p>
+</div>
+<div class="section" id="id165">
+<h5><a class="toc-backref" href="#id915">Arguments:</a><a class="headerlink" href="#id165" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">alloca</span></tt>‘ instruction allocates <tt class="docutils literal"><span class="pre">sizeof(<type>)*NumElements</span></tt>
+bytes of memory on the runtime stack, returning a pointer of the
+appropriate type to the program. If “NumElements” is specified, it is
+the number of elements allocated, otherwise “NumElements” is defaulted
+to be one. If a constant alignment is specified, the value result of the
+allocation is guaranteed to be aligned to at least that boundary. The
+alignment may not be greater than <tt class="docutils literal"><span class="pre">1</span> <span class="pre"><<</span> <span class="pre">29</span></tt>. If not specified, or if
+zero, the target can choose to align the allocation on any convenient
+boundary compatible with the type.</p>
+<p>‘<tt class="docutils literal"><span class="pre">type</span></tt>‘ may be any sized type.</p>
+</div>
+<div class="section" id="id166">
+<h5><a class="toc-backref" href="#id916">Semantics:</a><a class="headerlink" href="#id166" title="Permalink to this headline">¶</a></h5>
+<p>Memory is allocated; a pointer is returned. The operation is undefined
+if there is insufficient stack space for the allocation. ‘<tt class="docutils literal"><span class="pre">alloca</span></tt>‘d
+memory is automatically released when the function returns. The
+‘<tt class="docutils literal"><span class="pre">alloca</span></tt>‘ instruction is commonly used to represent automatic
+variables that must have an address available. When the function returns
+(either with the <tt class="docutils literal"><span class="pre">ret</span></tt> or <tt class="docutils literal"><span class="pre">resume</span></tt> instructions), the memory is
+reclaimed. Allocating zero bytes is legal, but the result is undefined.
+The order in which memory is allocated (ie., which way the stack grows)
+is not specified.</p>
+</div>
+<div class="section" id="id167">
+<h5><a class="toc-backref" href="#id917">Example:</a><a class="headerlink" href="#id167" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%ptr</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i32</span> <span class="c">; yields i32*:ptr</span>
+<span class="nv">%ptr</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span> <span class="m">4</span> <span class="c">; yields i32*:ptr</span>
+<span class="nv">%ptr</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span> <span class="m">4</span><span class="p">,</span> <span class="k">align</span> <span class="m">1024</span> <span class="c">; yields i32*:ptr</span>
+<span class="nv">%ptr</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i32</span><span class="p">,</span> <span class="k">align</span> <span class="m">1024</span> <span class="c">; yields i32*:ptr</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="load-instruction">
+<span id="i-load"></span><h4><a class="toc-backref" href="#id918">‘<tt class="docutils literal"><span class="pre">load</span></tt>‘ Instruction</a><a class="headerlink" href="#load-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id168">
+<h5><a class="toc-backref" href="#id919">Syntax:</a><a class="headerlink" href="#id168" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = load [volatile] <ty>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>][, !invariant.load !<index>][, !nonnull !<index>][, !dereferenceable !<index>][, !dereferenceable_or_null !<index>]
+<result> = load atomic [volatile] <ty>* <pointer> [singlethread] <ordering>, align <alignment>
+!<index> = !{ i32 1 }
+</pre></div>
+</div>
+</div>
+<div class="section" id="id169">
+<h5><a class="toc-backref" href="#id920">Overview:</a><a class="headerlink" href="#id169" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">load</span></tt>‘ instruction is used to read from memory.</p>
+</div>
+<div class="section" id="id170">
+<h5><a class="toc-backref" href="#id921">Arguments:</a><a class="headerlink" href="#id170" title="Permalink to this headline">¶</a></h5>
+<p>The argument to the <tt class="docutils literal"><span class="pre">load</span></tt> instruction specifies the memory address
+from which to load. The type specified must be a <a class="reference internal" href="#t-firstclass"><em>first
+class</em></a> type. If the <tt class="docutils literal"><span class="pre">load</span></tt> is marked as <tt class="docutils literal"><span class="pre">volatile</span></tt>,
+then the optimizer is not allowed to modify the number or order of
+execution of this <tt class="docutils literal"><span class="pre">load</span></tt> with other <a class="reference internal" href="#volatile"><em>volatile
+operations</em></a>.</p>
+<p>If the <tt class="docutils literal"><span class="pre">load</span></tt> is marked as <tt class="docutils literal"><span class="pre">atomic</span></tt>, it takes an extra
+<a class="reference internal" href="#ordering"><em>ordering</em></a> and optional <tt class="docutils literal"><span class="pre">singlethread</span></tt> argument. The
+<tt class="docutils literal"><span class="pre">release</span></tt> and <tt class="docutils literal"><span class="pre">acq_rel</span></tt> orderings are not valid on <tt class="docutils literal"><span class="pre">load</span></tt>
+instructions. Atomic loads produce <a class="reference internal" href="#memmodel"><em>defined</em></a> results
+when they may see multiple atomic stores. The type of the pointee must
+be an integer type whose bit width is a power of two greater than or
+equal to eight and less than or equal to a target-specific size limit.
+<tt class="docutils literal"><span class="pre">align</span></tt> must be explicitly specified on atomic loads, and the load has
+undefined behavior if the alignment is not set to a value which is at
+least the size in bytes of the pointee. <tt class="docutils literal"><span class="pre">!nontemporal</span></tt> does not have
+any defined semantics for atomic loads.</p>
+<p>The optional constant <tt class="docutils literal"><span class="pre">align</span></tt> argument specifies the alignment of the
+operation (that is, the alignment of the memory address). A value of 0
+or an omitted <tt class="docutils literal"><span class="pre">align</span></tt> argument means that the operation has the ABI
+alignment for the target. It is the responsibility of the code emitter
+to ensure that the alignment information is correct. Overestimating the
+alignment results in undefined behavior. Underestimating the alignment
+may produce less efficient code. An alignment of 1 is always safe. The
+maximum possible alignment is <tt class="docutils literal"><span class="pre">1</span> <span class="pre"><<</span> <span class="pre">29</span></tt>.</p>
+<p>The optional <tt class="docutils literal"><span class="pre">!nontemporal</span></tt> metadata must reference a single
+metadata name <tt class="docutils literal"><span class="pre"><index></span></tt> corresponding to a metadata node with one
+<tt class="docutils literal"><span class="pre">i32</span></tt> entry of value 1. The existence of the <tt class="docutils literal"><span class="pre">!nontemporal</span></tt>
+metadata on the instruction tells the optimizer and code generator
+that this load is not expected to be reused in the cache. The code
+generator may select special instructions to save cache bandwidth, such
+as the <tt class="docutils literal"><span class="pre">MOVNT</span></tt> instruction on x86.</p>
+<p>The optional <tt class="docutils literal"><span class="pre">!invariant.load</span></tt> metadata must reference a single
+metadata name <tt class="docutils literal"><span class="pre"><index></span></tt> corresponding to a metadata node with no
+entries. The existence of the <tt class="docutils literal"><span class="pre">!invariant.load</span></tt> metadata on the
+instruction tells the optimizer and code generator that the address
+operand to this load points to memory which can be assumed unchanged.
+Being invariant does not imply that a location is dereferenceable,
+but it does imply that once the location is known dereferenceable
+its value is henceforth unchanging.</p>
+<p>The optional <tt class="docutils literal"><span class="pre">!nonnull</span></tt> metadata must reference a single
+metadata name <tt class="docutils literal"><span class="pre"><index></span></tt> corresponding to a metadata node with no
+entries. The existence of the <tt class="docutils literal"><span class="pre">!nonnull</span></tt> metadata on the
+instruction tells the optimizer that the value loaded is known to
+never be null. This is analogous to the ‘’nonnull’’ attribute
+on parameters and return values. This metadata can only be applied
+to loads of a pointer type.</p>
+<p>The optional <tt class="docutils literal"><span class="pre">!dereferenceable</span></tt> metadata must reference a single
+metadata name <tt class="docutils literal"><span class="pre"><index></span></tt> corresponding to a metadata node with one <tt class="docutils literal"><span class="pre">i64</span></tt>
+entry. The existence of the <tt class="docutils literal"><span class="pre">!dereferenceable</span></tt> metadata on the instruction
+tells the optimizer that the value loaded is known to be dereferenceable.
+The number of bytes known to be dereferenceable is specified by the integer
+value in the metadata node. This is analogous to the ‘’dereferenceable’’
+attribute on parameters and return values. This metadata can only be applied
+to loads of a pointer type.</p>
+<p>The optional <tt class="docutils literal"><span class="pre">!dereferenceable_or_null</span></tt> metadata must reference a single
+metadata name <tt class="docutils literal"><span class="pre"><index></span></tt> corresponding to a metadata node with one <tt class="docutils literal"><span class="pre">i64</span></tt>
+entry. The existence of the <tt class="docutils literal"><span class="pre">!dereferenceable_or_null</span></tt> metadata on the
+instruction tells the optimizer that the value loaded is known to be either
+dereferenceable or null.
+The number of bytes known to be dereferenceable is specified by the integer
+value in the metadata node. This is analogous to the ‘’dereferenceable_or_null’’
+attribute on parameters and return values. This metadata can only be applied
+to loads of a pointer type.</p>
+</div>
+<div class="section" id="id171">
+<h5><a class="toc-backref" href="#id922">Semantics:</a><a class="headerlink" href="#id171" title="Permalink to this headline">¶</a></h5>
+<p>The location of memory pointed to is loaded. If the value being loaded
+is of scalar type then the number of bytes read does not exceed the
+minimum number of bytes needed to hold all bits of the type. For
+example, loading an <tt class="docutils literal"><span class="pre">i24</span></tt> reads at most three bytes. When loading a
+value of a type like <tt class="docutils literal"><span class="pre">i20</span></tt> with a size that is not an integral number
+of bytes, the result is undefined if the value was not originally
+written using a store of the same type.</p>
+</div>
+<div class="section" id="id172">
+<h5><a class="toc-backref" href="#id923">Examples:</a><a class="headerlink" href="#id172" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%ptr</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i32</span> <span class="c">; yields i32*:ptr</span>
+<span class="k">store</span> <span class="k">i32</span> <span class="m">3</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%ptr</span> <span class="c">; yields void</span>
+<span class="nv">%val</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%ptr</span> <span class="c">; yields i32:val = i32 3</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="store-instruction">
+<span id="i-store"></span><h4><a class="toc-backref" href="#id924">‘<tt class="docutils literal"><span class="pre">store</span></tt>‘ Instruction</a><a class="headerlink" href="#store-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id173">
+<h5><a class="toc-backref" href="#id925">Syntax:</a><a class="headerlink" href="#id173" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>store [volatile] <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] ; yields void
+store atomic [volatile] <ty> <value>, <ty>* <pointer> [singlethread] <ordering>, align <alignment> ; yields void
+</pre></div>
+</div>
+</div>
+<div class="section" id="id174">
+<h5><a class="toc-backref" href="#id926">Overview:</a><a class="headerlink" href="#id174" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">store</span></tt>‘ instruction is used to write to memory.</p>
+</div>
+<div class="section" id="id175">
+<h5><a class="toc-backref" href="#id927">Arguments:</a><a class="headerlink" href="#id175" title="Permalink to this headline">¶</a></h5>
+<p>There are two arguments to the <tt class="docutils literal"><span class="pre">store</span></tt> instruction: a value to store
+and an address at which to store it. The type of the <tt class="docutils literal"><span class="pre"><pointer></span></tt>
+operand must be a pointer to the <a class="reference internal" href="#t-firstclass"><em>first class</em></a> type of
+the <tt class="docutils literal"><span class="pre"><value></span></tt> operand. If the <tt class="docutils literal"><span class="pre">store</span></tt> is marked as <tt class="docutils literal"><span class="pre">volatile</span></tt>,
+then the optimizer is not allowed to modify the number or order of
+execution of this <tt class="docutils literal"><span class="pre">store</span></tt> with other <a class="reference internal" href="#volatile"><em>volatile
+operations</em></a>.</p>
+<p>If the <tt class="docutils literal"><span class="pre">store</span></tt> is marked as <tt class="docutils literal"><span class="pre">atomic</span></tt>, it takes an extra
+<a class="reference internal" href="#ordering"><em>ordering</em></a> and optional <tt class="docutils literal"><span class="pre">singlethread</span></tt> argument. The
+<tt class="docutils literal"><span class="pre">acquire</span></tt> and <tt class="docutils literal"><span class="pre">acq_rel</span></tt> orderings aren’t valid on <tt class="docutils literal"><span class="pre">store</span></tt>
+instructions. Atomic loads produce <a class="reference internal" href="#memmodel"><em>defined</em></a> results
+when they may see multiple atomic stores. The type of the pointee must
+be an integer type whose bit width is a power of two greater than or
+equal to eight and less than or equal to a target-specific size limit.
+<tt class="docutils literal"><span class="pre">align</span></tt> must be explicitly specified on atomic stores, and the store
+has undefined behavior if the alignment is not set to a value which is
+at least the size in bytes of the pointee. <tt class="docutils literal"><span class="pre">!nontemporal</span></tt> does not
+have any defined semantics for atomic stores.</p>
+<p>The optional constant <tt class="docutils literal"><span class="pre">align</span></tt> argument specifies the alignment of the
+operation (that is, the alignment of the memory address). A value of 0
+or an omitted <tt class="docutils literal"><span class="pre">align</span></tt> argument means that the operation has the ABI
+alignment for the target. It is the responsibility of the code emitter
+to ensure that the alignment information is correct. Overestimating the
+alignment results in undefined behavior. Underestimating the
+alignment may produce less efficient code. An alignment of 1 is always
+safe. The maximum possible alignment is <tt class="docutils literal"><span class="pre">1</span> <span class="pre"><<</span> <span class="pre">29</span></tt>.</p>
+<p>The optional <tt class="docutils literal"><span class="pre">!nontemporal</span></tt> metadata must reference a single metadata
+name <tt class="docutils literal"><span class="pre"><index></span></tt> corresponding to a metadata node with one <tt class="docutils literal"><span class="pre">i32</span></tt> entry of
+value 1. The existence of the <tt class="docutils literal"><span class="pre">!nontemporal</span></tt> metadata on the instruction
+tells the optimizer and code generator that this load is not expected to
+be reused in the cache. The code generator may select special
+instructions to save cache bandwidth, such as the MOVNT instruction on
+x86.</p>
+</div>
+<div class="section" id="id176">
+<h5><a class="toc-backref" href="#id928">Semantics:</a><a class="headerlink" href="#id176" title="Permalink to this headline">¶</a></h5>
+<p>The contents of memory are updated to contain <tt class="docutils literal"><span class="pre"><value></span></tt> at the
+location specified by the <tt class="docutils literal"><span class="pre"><pointer></span></tt> operand. If <tt class="docutils literal"><span class="pre"><value></span></tt> is
+of scalar type then the number of bytes written does not exceed the
+minimum number of bytes needed to hold all bits of the type. For
+example, storing an <tt class="docutils literal"><span class="pre">i24</span></tt> writes at most three bytes. When writing a
+value of a type like <tt class="docutils literal"><span class="pre">i20</span></tt> with a size that is not an integral number
+of bytes, it is unspecified what happens to the extra bits that do not
+belong to the type, but they will typically be overwritten.</p>
+</div>
+<div class="section" id="id177">
+<h5><a class="toc-backref" href="#id929">Example:</a><a class="headerlink" href="#id177" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%ptr</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i32</span> <span class="c">; yields i32*:ptr</span>
+<span class="k">store</span> <span class="k">i32</span> <span class="m">3</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%ptr</span> <span class="c">; yields void</span>
+<span class="nv">%val</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%ptr</span> <span class="c">; yields i32:val = i32 3</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fence-instruction">
+<span id="i-fence"></span><h4><a class="toc-backref" href="#id930">‘<tt class="docutils literal"><span class="pre">fence</span></tt>‘ Instruction</a><a class="headerlink" href="#fence-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id178">
+<h5><a class="toc-backref" href="#id931">Syntax:</a><a class="headerlink" href="#id178" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>fence [singlethread] <ordering> ; yields void
+</pre></div>
+</div>
+</div>
+<div class="section" id="id179">
+<h5><a class="toc-backref" href="#id932">Overview:</a><a class="headerlink" href="#id179" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fence</span></tt>‘ instruction is used to introduce happens-before edges
+between operations.</p>
+</div>
+<div class="section" id="id180">
+<h5><a class="toc-backref" href="#id933">Arguments:</a><a class="headerlink" href="#id180" title="Permalink to this headline">¶</a></h5>
+<p>‘<tt class="docutils literal"><span class="pre">fence</span></tt>‘ instructions take an <a class="reference internal" href="#ordering"><em>ordering</em></a> argument which
+defines what <em>synchronizes-with</em> edges they add. They can only be given
+<tt class="docutils literal"><span class="pre">acquire</span></tt>, <tt class="docutils literal"><span class="pre">release</span></tt>, <tt class="docutils literal"><span class="pre">acq_rel</span></tt>, and <tt class="docutils literal"><span class="pre">seq_cst</span></tt> orderings.</p>
+</div>
+<div class="section" id="id181">
+<h5><a class="toc-backref" href="#id934">Semantics:</a><a class="headerlink" href="#id181" title="Permalink to this headline">¶</a></h5>
+<p>A fence A which has (at least) <tt class="docutils literal"><span class="pre">release</span></tt> ordering semantics
+<em>synchronizes with</em> a fence B with (at least) <tt class="docutils literal"><span class="pre">acquire</span></tt> ordering
+semantics if and only if there exist atomic operations X and Y, both
+operating on some atomic object M, such that A is sequenced before X, X
+modifies M (either directly or through some side effect of a sequence
+headed by X), Y is sequenced before B, and Y observes M. This provides a
+<em>happens-before</em> dependency between A and B. Rather than an explicit
+<tt class="docutils literal"><span class="pre">fence</span></tt>, one (but not both) of the atomic operations X or Y might
+provide a <tt class="docutils literal"><span class="pre">release</span></tt> or <tt class="docutils literal"><span class="pre">acquire</span></tt> (resp.) ordering constraint and
+still <em>synchronize-with</em> the explicit <tt class="docutils literal"><span class="pre">fence</span></tt> and establish the
+<em>happens-before</em> edge.</p>
+<p>A <tt class="docutils literal"><span class="pre">fence</span></tt> which has <tt class="docutils literal"><span class="pre">seq_cst</span></tt> ordering, in addition to having both
+<tt class="docutils literal"><span class="pre">acquire</span></tt> and <tt class="docutils literal"><span class="pre">release</span></tt> semantics specified above, participates in
+the global program order of other <tt class="docutils literal"><span class="pre">seq_cst</span></tt> operations and/or fences.</p>
+<p>The optional “<a class="reference internal" href="#singlethread"><em>singlethread</em></a>” argument specifies
+that the fence only synchronizes with other fences in the same thread.
+(This is useful for interacting with signal handlers.)</p>
+</div>
+<div class="section" id="id182">
+<h5><a class="toc-backref" href="#id935">Example:</a><a class="headerlink" href="#id182" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>fence acquire ; yields void
+fence singlethread seq_cst ; yields void
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="cmpxchg-instruction">
+<span id="i-cmpxchg"></span><h4><a class="toc-backref" href="#id936">‘<tt class="docutils literal"><span class="pre">cmpxchg</span></tt>‘ Instruction</a><a class="headerlink" href="#cmpxchg-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id183">
+<h5><a class="toc-backref" href="#id937">Syntax:</a><a class="headerlink" href="#id183" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>cmpxchg [weak] [volatile] <ty>* <pointer>, <ty> <cmp>, <ty> <new> [singlethread] <success ordering> <failure ordering> ; yields { ty, i1 }
+</pre></div>
+</div>
+</div>
+<div class="section" id="id184">
+<h5><a class="toc-backref" href="#id938">Overview:</a><a class="headerlink" href="#id184" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">cmpxchg</span></tt>‘ instruction is used to atomically modify memory. It
+loads a value in memory and compares it to a given value. If they are
+equal, it tries to store a new value into the memory.</p>
+</div>
+<div class="section" id="id185">
+<h5><a class="toc-backref" href="#id939">Arguments:</a><a class="headerlink" href="#id185" title="Permalink to this headline">¶</a></h5>
+<p>There are three arguments to the ‘<tt class="docutils literal"><span class="pre">cmpxchg</span></tt>‘ instruction: an address
+to operate on, a value to compare to the value currently be at that
+address, and a new value to place at that address if the compared values
+are equal. The type of ‘<cmp>’ must be an integer type whose bit width
+is a power of two greater than or equal to eight and less than or equal
+to a target-specific size limit. ‘<cmp>’ and ‘<new>’ must have the same
+type, and the type of ‘<pointer>’ must be a pointer to that type. If the
+<tt class="docutils literal"><span class="pre">cmpxchg</span></tt> is marked as <tt class="docutils literal"><span class="pre">volatile</span></tt>, then the optimizer is not allowed
+to modify the number or order of execution of this <tt class="docutils literal"><span class="pre">cmpxchg</span></tt> with
+other <a class="reference internal" href="#volatile"><em>volatile operations</em></a>.</p>
+<p>The success and failure <a class="reference internal" href="#ordering"><em>ordering</em></a> arguments specify how this
+<tt class="docutils literal"><span class="pre">cmpxchg</span></tt> synchronizes with other atomic operations. Both ordering parameters
+must be at least <tt class="docutils literal"><span class="pre">monotonic</span></tt>, the ordering constraint on failure must be no
+stronger than that on success, and the failure ordering cannot be either
+<tt class="docutils literal"><span class="pre">release</span></tt> or <tt class="docutils literal"><span class="pre">acq_rel</span></tt>.</p>
+<p>The optional “<tt class="docutils literal"><span class="pre">singlethread</span></tt>” argument declares that the <tt class="docutils literal"><span class="pre">cmpxchg</span></tt>
+is only atomic with respect to code (usually signal handlers) running in
+the same thread as the <tt class="docutils literal"><span class="pre">cmpxchg</span></tt>. Otherwise the cmpxchg is atomic with
+respect to all other code in the system.</p>
+<p>The pointer passed into cmpxchg must have alignment greater than or
+equal to the size in memory of the operand.</p>
+</div>
+<div class="section" id="id186">
+<h5><a class="toc-backref" href="#id940">Semantics:</a><a class="headerlink" href="#id186" title="Permalink to this headline">¶</a></h5>
+<p>The contents of memory at the location specified by the ‘<tt class="docutils literal"><span class="pre"><pointer></span></tt>‘ operand
+is read and compared to ‘<tt class="docutils literal"><span class="pre"><cmp></span></tt>‘; if the read value is the equal, the
+‘<tt class="docutils literal"><span class="pre"><new></span></tt>‘ is written. The original value at the location is returned, together
+with a flag indicating success (true) or failure (false).</p>
+<p>If the cmpxchg operation is marked as <tt class="docutils literal"><span class="pre">weak</span></tt> then a spurious failure is
+permitted: the operation may not write <tt class="docutils literal"><span class="pre"><new></span></tt> even if the comparison
+matched.</p>
+<p>If the cmpxchg operation is strong (the default), the i1 value is 1 if and only
+if the value loaded equals <tt class="docutils literal"><span class="pre">cmp</span></tt>.</p>
+<p>A successful <tt class="docutils literal"><span class="pre">cmpxchg</span></tt> is a read-modify-write instruction for the purpose of
+identifying release sequences. A failed <tt class="docutils literal"><span class="pre">cmpxchg</span></tt> is equivalent to an atomic
+load with an ordering parameter determined the second ordering parameter.</p>
+</div>
+<div class="section" id="id187">
+<h5><a class="toc-backref" href="#id941">Example:</a><a class="headerlink" href="#id187" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>entry:
+ %orig = atomic load i32, i32* %ptr unordered ; yields i32
+ br label %loop
+
+loop:
+ %cmp = phi i32 [ %orig, %entry ], [%old, %loop]
+ %squared = mul i32 %cmp, %cmp
+ %val_success = cmpxchg i32* %ptr, i32 %cmp, i32 %squared acq_rel monotonic ; yields { i32, i1 }
+ %value_loaded = extractvalue { i32, i1 } %val_success, 0
+ %success = extractvalue { i32, i1 } %val_success, 1
+ br i1 %success, label %done, label %loop
+
+done:
+ ...
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="atomicrmw-instruction">
+<span id="i-atomicrmw"></span><h4><a class="toc-backref" href="#id942">‘<tt class="docutils literal"><span class="pre">atomicrmw</span></tt>‘ Instruction</a><a class="headerlink" href="#atomicrmw-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id188">
+<h5><a class="toc-backref" href="#id943">Syntax:</a><a class="headerlink" href="#id188" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>atomicrmw [volatile] <operation> <ty>* <pointer>, <ty> <value> [singlethread] <ordering> ; yields ty
+</pre></div>
+</div>
+</div>
+<div class="section" id="id189">
+<h5><a class="toc-backref" href="#id944">Overview:</a><a class="headerlink" href="#id189" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">atomicrmw</span></tt>‘ instruction is used to atomically modify memory.</p>
+</div>
+<div class="section" id="id190">
+<h5><a class="toc-backref" href="#id945">Arguments:</a><a class="headerlink" href="#id190" title="Permalink to this headline">¶</a></h5>
+<p>There are three arguments to the ‘<tt class="docutils literal"><span class="pre">atomicrmw</span></tt>‘ instruction: an
+operation to apply, an address whose value to modify, an argument to the
+operation. The operation must be one of the following keywords:</p>
+<ul class="simple">
+<li>xchg</li>
+<li>add</li>
+<li>sub</li>
+<li>and</li>
+<li>nand</li>
+<li>or</li>
+<li>xor</li>
+<li>max</li>
+<li>min</li>
+<li>umax</li>
+<li>umin</li>
+</ul>
+<p>The type of ‘<value>’ must be an integer type whose bit width is a power
+of two greater than or equal to eight and less than or equal to a
+target-specific size limit. The type of the ‘<tt class="docutils literal"><span class="pre"><pointer></span></tt>‘ operand must
+be a pointer to that type. If the <tt class="docutils literal"><span class="pre">atomicrmw</span></tt> is marked as
+<tt class="docutils literal"><span class="pre">volatile</span></tt>, then the optimizer is not allowed to modify the number or
+order of execution of this <tt class="docutils literal"><span class="pre">atomicrmw</span></tt> with other <a class="reference internal" href="#volatile"><em>volatile
+operations</em></a>.</p>
+</div>
+<div class="section" id="id191">
+<h5><a class="toc-backref" href="#id946">Semantics:</a><a class="headerlink" href="#id191" title="Permalink to this headline">¶</a></h5>
+<p>The contents of memory at the location specified by the ‘<tt class="docutils literal"><span class="pre"><pointer></span></tt>‘
+operand are atomically read, modified, and written back. The original
+value at the location is returned. The modification is specified by the
+operation argument:</p>
+<ul class="simple">
+<li>xchg: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">val</span></tt></li>
+<li>add: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre">+</span> <span class="pre">val</span></tt></li>
+<li>sub: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre">-</span> <span class="pre">val</span></tt></li>
+<li>and: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre">&</span> <span class="pre">val</span></tt></li>
+<li>nand: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">~(*ptr</span> <span class="pre">&</span> <span class="pre">val)</span></tt></li>
+<li>or: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre">|</span> <span class="pre">val</span></tt></li>
+<li>xor: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre">^</span> <span class="pre">val</span></tt></li>
+<li>max: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre">></span> <span class="pre">val</span> <span class="pre">?</span> <span class="pre">*ptr</span> <span class="pre">:</span> <span class="pre">val</span></tt> (using a signed comparison)</li>
+<li>min: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre"><</span> <span class="pre">val</span> <span class="pre">?</span> <span class="pre">*ptr</span> <span class="pre">:</span> <span class="pre">val</span></tt> (using a signed comparison)</li>
+<li>umax: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre">></span> <span class="pre">val</span> <span class="pre">?</span> <span class="pre">*ptr</span> <span class="pre">:</span> <span class="pre">val</span></tt> (using an unsigned
+comparison)</li>
+<li>umin: <tt class="docutils literal"><span class="pre">*ptr</span> <span class="pre">=</span> <span class="pre">*ptr</span> <span class="pre"><</span> <span class="pre">val</span> <span class="pre">?</span> <span class="pre">*ptr</span> <span class="pre">:</span> <span class="pre">val</span></tt> (using an unsigned
+comparison)</li>
+</ul>
+</div>
+<div class="section" id="id192">
+<h5><a class="toc-backref" href="#id947">Example:</a><a class="headerlink" href="#id192" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>%old = atomicrmw add i32* %ptr, i32 1 acquire ; yields i32
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="getelementptr-instruction">
+<span id="i-getelementptr"></span><h4><a class="toc-backref" href="#id948">‘<tt class="docutils literal"><span class="pre">getelementptr</span></tt>‘ Instruction</a><a class="headerlink" href="#getelementptr-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id193">
+<h5><a class="toc-backref" href="#id949">Syntax:</a><a class="headerlink" href="#id193" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = getelementptr <ty>, <ty>* <ptrval>{, <ty> <idx>}*
+<result> = getelementptr inbounds <ty>, <ty>* <ptrval>{, <ty> <idx>}*
+<result> = getelementptr <ty>, <ptr vector> <ptrval>, <vector index type> <idx>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id194">
+<h5><a class="toc-backref" href="#id950">Overview:</a><a class="headerlink" href="#id194" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">getelementptr</span></tt>‘ instruction is used to get the address of a
+subelement of an <a class="reference internal" href="#t-aggregate"><em>aggregate</em></a> data structure. It performs
+address calculation only and does not access memory. The instruction can also
+be used to calculate a vector of such addresses.</p>
+</div>
+<div class="section" id="id195">
+<h5><a class="toc-backref" href="#id951">Arguments:</a><a class="headerlink" href="#id195" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is always a type used as the basis for the calculations.
+The second argument is always a pointer or a vector of pointers, and is the
+base address to start from. The remaining arguments are indices
+that indicate which of the elements of the aggregate object are indexed.
+The interpretation of each index is dependent on the type being indexed
+into. The first index always indexes the pointer value given as the
+first argument, the second index indexes a value of the type pointed to
+(not necessarily the value directly pointed to, since the first index
+can be non-zero), etc. The first type indexed into must be a pointer
+value, subsequent types can be arrays, vectors, and structs. Note that
+subsequent types being indexed into can never be pointers, since that
+would require loading the pointer before continuing calculation.</p>
+<p>The type of each index argument depends on the type it is indexing into.
+When indexing into a (optionally packed) structure, only <tt class="docutils literal"><span class="pre">i32</span></tt> integer
+<strong>constants</strong> are allowed (when using a vector of indices they must all
+be the <strong>same</strong> <tt class="docutils literal"><span class="pre">i32</span></tt> integer constant). When indexing into an array,
+pointer or vector, integers of any width are allowed, and they are not
+required to be constant. These integers are treated as signed values
+where relevant.</p>
+<p>For example, let’s consider a C code fragment and how it gets compiled
+to LLVM:</p>
+<div class="highlight-c"><div class="highlight"><pre><span class="k">struct</span> <span class="n">RT</span> <span class="p">{</span>
+ <span class="kt">char</span> <span class="n">A</span><span class="p">;</span>
+ <span class="kt">int</span> <span class="n">B</span><span class="p">[</span><span class="mi">10</span><span class="p">][</span><span class="mi">20</span><span class="p">];</span>
+ <span class="kt">char</span> <span class="n">C</span><span class="p">;</span>
+<span class="p">};</span>
+<span class="k">struct</span> <span class="n">ST</span> <span class="p">{</span>
+ <span class="kt">int</span> <span class="n">X</span><span class="p">;</span>
+ <span class="kt">double</span> <span class="n">Y</span><span class="p">;</span>
+ <span class="k">struct</span> <span class="n">RT</span> <span class="n">Z</span><span class="p">;</span>
+<span class="p">};</span>
+
+<span class="kt">int</span> <span class="o">*</span><span class="nf">foo</span><span class="p">(</span><span class="k">struct</span> <span class="n">ST</span> <span class="o">*</span><span class="n">s</span><span class="p">)</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="o">&</span><span class="n">s</span><span class="p">[</span><span class="mi">1</span><span class="p">].</span><span class="n">Z</span><span class="p">.</span><span class="n">B</span><span class="p">[</span><span class="mi">5</span><span class="p">][</span><span class="mi">13</span><span class="p">];</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>The LLVM code generated by Clang is:</p>
+<div class="highlight-llvm"><div class="highlight"><pre>%struct.RT = type { i8, [10 x [20 x i32]], i8 }
+%struct.ST = type { i32, double, %struct.RT }
+
+define i32* @foo(%struct.ST* %s) nounwind uwtable readnone optsize ssp {
+entry:
+ %arrayidx = getelementptr inbounds %struct.ST, %struct.ST* %s, i64 1, i32 2, i32 1, i64 5, i64 13
+ ret i32* %arrayidx
+}
+</pre></div>
+</div>
+</div>
+<div class="section" id="id196">
+<h5><a class="toc-backref" href="#id952">Semantics:</a><a class="headerlink" href="#id196" title="Permalink to this headline">¶</a></h5>
+<p>In the example above, the first index is indexing into the
+‘<tt class="docutils literal"><span class="pre">%struct.ST*</span></tt>‘ type, which is a pointer, yielding a ‘<tt class="docutils literal"><span class="pre">%struct.ST</span></tt>‘
+= ‘<tt class="docutils literal"><span class="pre">{</span> <span class="pre">i32,</span> <span class="pre">double,</span> <span class="pre">%struct.RT</span> <span class="pre">}</span></tt>‘ type, a structure. The second index
+indexes into the third element of the structure, yielding a
+‘<tt class="docutils literal"><span class="pre">%struct.RT</span></tt>‘ = ‘<tt class="docutils literal"><span class="pre">{</span> <span class="pre">i8</span> <span class="pre">,</span> <span class="pre">[10</span> <span class="pre">x</span> <span class="pre">[20</span> <span class="pre">x</span> <span class="pre">i32]],</span> <span class="pre">i8</span> <span class="pre">}</span></tt>‘ type, another
+structure. The third index indexes into the second element of the
+structure, yielding a ‘<tt class="docutils literal"><span class="pre">[10</span> <span class="pre">x</span> <span class="pre">[20</span> <span class="pre">x</span> <span class="pre">i32]]</span></tt>‘ type, an array. The two
+dimensions of the array are subscripted into, yielding an ‘<tt class="docutils literal"><span class="pre">i32</span></tt>‘
+type. The ‘<tt class="docutils literal"><span class="pre">getelementptr</span></tt>‘ instruction returns a pointer to this
+element, thus computing a value of ‘<tt class="docutils literal"><span class="pre">i32*</span></tt>‘ type.</p>
+<p>Note that it is perfectly legal to index partially through a structure,
+returning a pointer to an inner element. Because of this, the LLVM code
+for the given testcase is equivalent to:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">define</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@foo</span><span class="p">(</span><span class="nv">%struct.ST</span><span class="p">*</span> <span class="nv">%s</span><span class="p">)</span> <span class="p">{</span>
+ <span class="nv">%t1</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="nv">%struct.ST</span><span class="p">,</span> <span class="nv">%struct.ST</span><span class="p">*</span> <span class="nv">%s</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span> <span class="c">; yields %struct.ST*:%t1</span>
+ <span class="nv">%t2</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="nv">%struct.ST</span><span class="p">,</span> <span class="nv">%struct.ST</span><span class="p">*</span> <span class="nv">%t1</span><span class="p">,</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">2</span> <span class="c">; yields %struct.RT*:%t2</span>
+ <span class="nv">%t3</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="nv">%struct.RT</span><span class="p">,</span> <span class="nv">%struct.RT</span><span class="p">*</span> <span class="nv">%t2</span><span class="p">,</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span> <span class="c">; yields [10 x [20 x i32]]*:%t3</span>
+ <span class="nv">%t4</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="p">[</span><span class="m">10</span> <span class="k">x</span> <span class="p">[</span><span class="m">20</span> <span class="k">x</span> <span class="k">i32</span><span class="p">]],</span> <span class="p">[</span><span class="m">10</span> <span class="k">x</span> <span class="p">[</span><span class="m">20</span> <span class="k">x</span> <span class="k">i32</span><span class="p">]]*</span> <span class="nv">%t3</span><span class="p">,</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">5</span> <span class="c">; yields [20 x i32]*:%t4</span>
+ <span class="nv">%t5</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="p">[</span><span class="m">20</span> <span class="k">x</span> <span class="k">i32</span><span class="p">],</span> <span class="p">[</span><span class="m">20</span> <span class="k">x</span> <span class="k">i32</span><span class="p">]*</span> <span class="nv">%t4</span><span class="p">,</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">13</span> <span class="c">; yields i32*:%t5</span>
+ <span class="k">ret</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%t5</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>If the <tt class="docutils literal"><span class="pre">inbounds</span></tt> keyword is present, the result value of the
+<tt class="docutils literal"><span class="pre">getelementptr</span></tt> is a <a class="reference internal" href="#poisonvalues"><em>poison value</em></a> if the base
+pointer is not an <em>in bounds</em> address of an allocated object, or if any
+of the addresses that would be formed by successive addition of the
+offsets implied by the indices to the base address with infinitely
+precise signed arithmetic are not an <em>in bounds</em> address of that
+allocated object. The <em>in bounds</em> addresses for an allocated object are
+all the addresses that point into the object, plus the address one byte
+past the end. In cases where the base is a vector of pointers the
+<tt class="docutils literal"><span class="pre">inbounds</span></tt> keyword applies to each of the computations element-wise.</p>
+<p>If the <tt class="docutils literal"><span class="pre">inbounds</span></tt> keyword is not present, the offsets are added to the
+base address with silently-wrapping two’s complement arithmetic. If the
+offsets have a different width from the pointer, they are sign-extended
+or truncated to the width of the pointer. The result value of the
+<tt class="docutils literal"><span class="pre">getelementptr</span></tt> may be outside the object pointed to by the base
+pointer. The result value may not necessarily be used to access memory
+though, even if it happens to point into allocated storage. See the
+<a class="reference internal" href="#pointeraliasing"><em>Pointer Aliasing Rules</em></a> section for more
+information.</p>
+<p>The getelementptr instruction is often confusing. For some more insight
+into how it works, see <a class="reference internal" href="GetElementPtr.html"><em>the getelementptr FAQ</em></a>.</p>
+</div>
+<div class="section" id="id197">
+<h5><a class="toc-backref" href="#id953">Example:</a><a class="headerlink" href="#id197" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="c">; yields [12 x i8]*:aptr</span>
+<span class="nv">%aptr</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="p">[</span><span class="m">12</span> <span class="k">x</span> <span class="k">i8</span><span class="p">]},</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="p">[</span><span class="m">12</span> <span class="k">x</span> <span class="k">i8</span><span class="p">]}*</span> <span class="nv">%saptr</span><span class="p">,</span> <span class="k">i64</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span>
+<span class="c">; yields i8*:vptr</span>
+<span class="nv">%vptr</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="p"><</span><span class="m">2</span> <span class="k">x</span> <span class="k">i8</span><span class="p">>},</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="p"><</span><span class="m">2</span> <span class="k">x</span> <span class="k">i8</span><span class="p">>}*</span> <span class="nv">%svptr</span><span class="p">,</span> <span class="k">i64</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span>
+<span class="c">; yields i8*:eptr</span>
+<span class="nv">%eptr</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="p">[</span><span class="m">12</span> <span class="k">x</span> <span class="k">i8</span><span class="p">],</span> <span class="p">[</span><span class="m">12</span> <span class="k">x</span> <span class="k">i8</span><span class="p">]*</span> <span class="nv">%aptr</span><span class="p">,</span> <span class="k">i64</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span>
+<span class="c">; yields i32*:iptr</span>
+<span class="nv">%iptr</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="p">[</span><span class="m">10</span> <span class="k">x</span> <span class="k">i32</span><span class="p">],</span> <span class="p">[</span><span class="m">10</span> <span class="k">x</span> <span class="k">i32</span><span class="p">]*</span> <span class="vg">@arr</span><span class="p">,</span> <span class="k">i16</span> <span class="m">0</span><span class="p">,</span> <span class="k">i16</span> <span class="m">0</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="vector-of-pointers">
+<h5><a class="toc-backref" href="#id954">Vector of pointers:</a><a class="headerlink" href="#vector-of-pointers" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">getelementptr</span></tt> returns a vector of pointers, instead of a single address,
+when one or more of its arguments is a vector. In such cases, all vector
+arguments should have the same number of elements, and every scalar argument
+will be effectively broadcast into a vector during address calculation.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="c">; All arguments are vectors:</span>
+<span class="c">; A[i] = ptrs[i] + offsets[i]*sizeof(i8)</span>
+<span class="nv">%A</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="k">i8</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i8</span><span class="p">*></span> <span class="nv">%ptrs</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i64</span><span class="p">></span> <span class="nv">%offsets</span>
+
+<span class="c">; Add the same scalar offset to each pointer of a vector:</span>
+<span class="c">; A[i] = ptrs[i] + offset*sizeof(i8)</span>
+<span class="nv">%A</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="k">i8</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i8</span><span class="p">*></span> <span class="nv">%ptrs</span><span class="p">,</span> <span class="k">i64</span> <span class="nv">%offset</span>
+
+<span class="c">; Add distinct offsets to the same pointer:</span>
+<span class="c">; A[i] = ptr + offsets[i]*sizeof(i8)</span>
+<span class="nv">%A</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="k">i8</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%ptr</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i64</span><span class="p">></span> <span class="nv">%offsets</span>
+
+<span class="c">; In all cases described above the type of the result is <4 x i8*></span>
+</pre></div>
+</div>
+<p>The two following instructions are equivalent:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">getelementptr</span> <span class="nv">%struct.ST</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="nv">%struct.ST</span><span class="p">*></span> <span class="nv">%s</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i64</span><span class="p">></span> <span class="nv">%ind1</span><span class="p">,</span>
+ <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i32</span><span class="p">></span> <span class="p"><</span><span class="k">i32</span> <span class="m">2</span><span class="p">,</span> <span class="k">i32</span> <span class="m">2</span><span class="p">,</span> <span class="k">i32</span> <span class="m">2</span><span class="p">,</span> <span class="k">i32</span> <span class="m">2</span><span class="p">>,</span>
+ <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i32</span><span class="p">></span> <span class="p"><</span><span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span><span class="p">>,</span>
+ <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i32</span><span class="p">></span> <span class="nv">%ind4</span><span class="p">,</span>
+ <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i64</span><span class="p">></span> <span class="p"><</span><span class="k">i64</span> <span class="m">13</span><span class="p">,</span> <span class="k">i64</span> <span class="m">13</span><span class="p">,</span> <span class="k">i64</span> <span class="m">13</span><span class="p">,</span> <span class="k">i64</span> <span class="m">13</span><span class="p">></span>
+
+<span class="k">getelementptr</span> <span class="nv">%struct.ST</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="nv">%struct.ST</span><span class="p">*></span> <span class="nv">%s</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i64</span><span class="p">></span> <span class="nv">%ind1</span><span class="p">,</span>
+ <span class="k">i32</span> <span class="m">2</span><span class="p">,</span> <span class="k">i32</span> <span class="m">1</span><span class="p">,</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i32</span><span class="p">></span> <span class="nv">%ind4</span><span class="p">,</span> <span class="k">i64</span> <span class="m">13</span>
+</pre></div>
+</div>
+<p>Let’s look at the C code, where the vector version of <tt class="docutils literal"><span class="pre">getelementptr</span></tt>
+makes sense:</p>
+<div class="highlight-c"><div class="highlight"><pre><span class="c1">// Let's assume that we vectorize the following loop:</span>
+<span class="kt">double</span> <span class="o">*</span><span class="n">A</span><span class="p">,</span> <span class="n">B</span><span class="p">;</span> <span class="kt">int</span> <span class="o">*</span><span class="n">C</span><span class="p">;</span>
+<span class="k">for</span> <span class="p">(</span><span class="kt">int</span> <span class="n">i</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span> <span class="n">i</span> <span class="o"><</span> <span class="n">size</span><span class="p">;</span> <span class="o">++</span><span class="n">i</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">A</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="o">=</span> <span class="n">B</span><span class="p">[</span><span class="n">C</span><span class="p">[</span><span class="n">i</span><span class="p">]];</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="c">; get pointers for 8 elements from array B</span>
+<span class="nv">%ptrs</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="kt">double</span><span class="p">,</span> <span class="kt">double</span><span class="p">*</span> <span class="nv">%B</span><span class="p">,</span> <span class="p"><</span><span class="m">8</span> <span class="k">x</span> <span class="k">i32</span><span class="p">></span> <span class="nv">%C</span>
+<span class="c">; load 8 elements from array B into A</span>
+<span class="nv">%A</span> <span class="p">=</span> <span class="k">call</span> <span class="p"><</span><span class="m">8</span> <span class="k">x</span> <span class="kt">double</span><span class="p">></span> <span class="vg">@llvm.masked.gather.v8f64</span><span class="p">(<</span><span class="m">8</span> <span class="k">x</span> <span class="kt">double</span><span class="p">*></span> <span class="nv">%ptrs</span><span class="p">,</span>
+ <span class="k">i32</span> <span class="m">8</span><span class="p">,</span> <span class="p"><</span><span class="m">8</span> <span class="k">x</span> <span class="k">i1</span><span class="p">></span> <span class="nv">%mask</span><span class="p">,</span> <span class="p"><</span><span class="m">8</span> <span class="k">x</span> <span class="kt">double</span><span class="p">></span> <span class="nv">%passthru</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="conversion-operations">
+<h3><a class="toc-backref" href="#id955">Conversion Operations</a><a class="headerlink" href="#conversion-operations" title="Permalink to this headline">¶</a></h3>
+<p>The instructions in this category are the conversion instructions
+(casting) which all take a single operand and a type. They perform
+various bit conversions on the operand.</p>
+<div class="section" id="trunc-to-instruction">
+<h4><a class="toc-backref" href="#id956">‘<tt class="docutils literal"><span class="pre">trunc</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#trunc-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id198">
+<h5><a class="toc-backref" href="#id957">Syntax:</a><a class="headerlink" href="#id198" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = trunc <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id199">
+<h5><a class="toc-backref" href="#id958">Overview:</a><a class="headerlink" href="#id199" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">trunc</span></tt>‘ instruction truncates its operand to the type <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id200">
+<h5><a class="toc-backref" href="#id959">Arguments:</a><a class="headerlink" href="#id200" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">trunc</span></tt>‘ instruction takes a value to trunc, and a type to trunc
+it to. Both types must be of <a class="reference internal" href="#t-integer"><em>integer</em></a> types, or vectors
+of the same number of integers. The bit size of the <tt class="docutils literal"><span class="pre">value</span></tt> must be
+larger than the bit size of the destination type, <tt class="docutils literal"><span class="pre">ty2</span></tt>. Equal sized
+types are not allowed.</p>
+</div>
+<div class="section" id="id201">
+<h5><a class="toc-backref" href="#id960">Semantics:</a><a class="headerlink" href="#id201" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">trunc</span></tt>‘ instruction truncates the high order bits in <tt class="docutils literal"><span class="pre">value</span></tt>
+and converts the remaining bits to <tt class="docutils literal"><span class="pre">ty2</span></tt>. Since the source size must
+be larger than the destination size, <tt class="docutils literal"><span class="pre">trunc</span></tt> cannot be a <em>no-op cast</em>.
+It will always truncate bits.</p>
+</div>
+<div class="section" id="id202">
+<h5><a class="toc-backref" href="#id961">Example:</a><a class="headerlink" href="#id202" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">trunc</span> <span class="k">i32</span> <span class="m">257</span> <span class="k">to</span> <span class="k">i8</span> <span class="c">; yields i8:1</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">trunc</span> <span class="k">i32</span> <span class="m">123</span> <span class="k">to</span> <span class="k">i1</span> <span class="c">; yields i1:true</span>
+<span class="nv">%Z</span> <span class="p">=</span> <span class="k">trunc</span> <span class="k">i32</span> <span class="m">122</span> <span class="k">to</span> <span class="k">i1</span> <span class="c">; yields i1:false</span>
+<span class="nv">%W</span> <span class="p">=</span> <span class="k">trunc</span> <span class="p"><</span><span class="m">2</span> <span class="k">x</span> <span class="k">i16</span><span class="p">></span> <span class="p"><</span><span class="k">i16</span> <span class="m">8</span><span class="p">,</span> <span class="k">i16</span> <span class="m">7</span><span class="p">></span> <span class="k">to</span> <span class="p"><</span><span class="m">2</span> <span class="k">x</span> <span class="k">i8</span><span class="p">></span> <span class="c">; yields <i8 8, i8 7></span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="zext-to-instruction">
+<h4><a class="toc-backref" href="#id962">‘<tt class="docutils literal"><span class="pre">zext</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#zext-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id203">
+<h5><a class="toc-backref" href="#id963">Syntax:</a><a class="headerlink" href="#id203" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = zext <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id204">
+<h5><a class="toc-backref" href="#id964">Overview:</a><a class="headerlink" href="#id204" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">zext</span></tt>‘ instruction zero extends its operand to type <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id205">
+<h5><a class="toc-backref" href="#id965">Arguments:</a><a class="headerlink" href="#id205" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">zext</span></tt>‘ instruction takes a value to cast, and a type to cast it
+to. Both types must be of <a class="reference internal" href="#t-integer"><em>integer</em></a> types, or vectors of
+the same number of integers. The bit size of the <tt class="docutils literal"><span class="pre">value</span></tt> must be
+smaller than the bit size of the destination type, <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id206">
+<h5><a class="toc-backref" href="#id966">Semantics:</a><a class="headerlink" href="#id206" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">zext</span></tt> fills the high order bits of the <tt class="docutils literal"><span class="pre">value</span></tt> with zero bits
+until it reaches the size of the destination type, <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+<p>When zero extending from i1, the result will always be either 0 or 1.</p>
+</div>
+<div class="section" id="id207">
+<h5><a class="toc-backref" href="#id967">Example:</a><a class="headerlink" href="#id207" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">zext</span> <span class="k">i32</span> <span class="m">257</span> <span class="k">to</span> <span class="k">i64</span> <span class="c">; yields i64:257</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">zext</span> <span class="k">i1</span> <span class="k">true</span> <span class="k">to</span> <span class="k">i32</span> <span class="c">; yields i32:1</span>
+<span class="nv">%Z</span> <span class="p">=</span> <span class="k">zext</span> <span class="p"><</span><span class="m">2</span> <span class="k">x</span> <span class="k">i16</span><span class="p">></span> <span class="p"><</span><span class="k">i16</span> <span class="m">8</span><span class="p">,</span> <span class="k">i16</span> <span class="m">7</span><span class="p">></span> <span class="k">to</span> <span class="p"><</span><span class="m">2</span> <span class="k">x</span> <span class="k">i32</span><span class="p">></span> <span class="c">; yields <i32 8, i32 7></span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="sext-to-instruction">
+<h4><a class="toc-backref" href="#id968">‘<tt class="docutils literal"><span class="pre">sext</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#sext-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id208">
+<h5><a class="toc-backref" href="#id969">Syntax:</a><a class="headerlink" href="#id208" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = sext <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id209">
+<h5><a class="toc-backref" href="#id970">Overview:</a><a class="headerlink" href="#id209" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">sext</span></tt>‘ sign extends <tt class="docutils literal"><span class="pre">value</span></tt> to the type <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id210">
+<h5><a class="toc-backref" href="#id971">Arguments:</a><a class="headerlink" href="#id210" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">sext</span></tt>‘ instruction takes a value to cast, and a type to cast it
+to. Both types must be of <a class="reference internal" href="#t-integer"><em>integer</em></a> types, or vectors of
+the same number of integers. The bit size of the <tt class="docutils literal"><span class="pre">value</span></tt> must be
+smaller than the bit size of the destination type, <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id211">
+<h5><a class="toc-backref" href="#id972">Semantics:</a><a class="headerlink" href="#id211" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">sext</span></tt>‘ instruction performs a sign extension by copying the sign
+bit (highest order bit) of the <tt class="docutils literal"><span class="pre">value</span></tt> until it reaches the bit size
+of the type <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+<p>When sign extending from i1, the extension always results in -1 or 0.</p>
+</div>
+<div class="section" id="id212">
+<h5><a class="toc-backref" href="#id973">Example:</a><a class="headerlink" href="#id212" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">sext</span> <span class="k">i8</span> <span class="m">-1</span> <span class="k">to</span> <span class="k">i16</span> <span class="c">; yields i16 :65535</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">sext</span> <span class="k">i1</span> <span class="k">true</span> <span class="k">to</span> <span class="k">i32</span> <span class="c">; yields i32:-1</span>
+<span class="nv">%Z</span> <span class="p">=</span> <span class="k">sext</span> <span class="p"><</span><span class="m">2</span> <span class="k">x</span> <span class="k">i16</span><span class="p">></span> <span class="p"><</span><span class="k">i16</span> <span class="m">8</span><span class="p">,</span> <span class="k">i16</span> <span class="m">7</span><span class="p">></span> <span class="k">to</span> <span class="p"><</span><span class="m">2</span> <span class="k">x</span> <span class="k">i32</span><span class="p">></span> <span class="c">; yields <i32 8, i32 7></span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fptrunc-to-instruction">
+<h4><a class="toc-backref" href="#id974">‘<tt class="docutils literal"><span class="pre">fptrunc</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#fptrunc-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id213">
+<h5><a class="toc-backref" href="#id975">Syntax:</a><a class="headerlink" href="#id213" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fptrunc <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id214">
+<h5><a class="toc-backref" href="#id976">Overview:</a><a class="headerlink" href="#id214" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptrunc</span></tt>‘ instruction truncates <tt class="docutils literal"><span class="pre">value</span></tt> to type <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id215">
+<h5><a class="toc-backref" href="#id977">Arguments:</a><a class="headerlink" href="#id215" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptrunc</span></tt>‘ instruction takes a <a class="reference internal" href="#t-floating"><em>floating point</em></a>
+value to cast and a <a class="reference internal" href="#t-floating"><em>floating point</em></a> type to cast it to.
+The size of <tt class="docutils literal"><span class="pre">value</span></tt> must be larger than the size of <tt class="docutils literal"><span class="pre">ty2</span></tt>. This
+implies that <tt class="docutils literal"><span class="pre">fptrunc</span></tt> cannot be used to make a <em>no-op cast</em>.</p>
+</div>
+<div class="section" id="id216">
+<h5><a class="toc-backref" href="#id978">Semantics:</a><a class="headerlink" href="#id216" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptrunc</span></tt>‘ instruction truncates a <tt class="docutils literal"><span class="pre">value</span></tt> from a larger
+<a class="reference internal" href="#t-floating"><em>floating point</em></a> type to a smaller <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> type. If the value cannot fit within the
+destination type, <tt class="docutils literal"><span class="pre">ty2</span></tt>, then the results are undefined.</p>
+</div>
+<div class="section" id="id217">
+<h5><a class="toc-backref" href="#id979">Example:</a><a class="headerlink" href="#id217" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">fptrunc</span> <span class="kt">double</span> <span class="m">123.0</span> <span class="k">to</span> <span class="kt">float</span> <span class="c">; yields float:123.0</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">fptrunc</span> <span class="kt">double</span> <span class="m">1.0E+300</span> <span class="k">to</span> <span class="kt">float</span> <span class="c">; yields undefined</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fpext-to-instruction">
+<h4><a class="toc-backref" href="#id980">‘<tt class="docutils literal"><span class="pre">fpext</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#fpext-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id218">
+<h5><a class="toc-backref" href="#id981">Syntax:</a><a class="headerlink" href="#id218" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fpext <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id219">
+<h5><a class="toc-backref" href="#id982">Overview:</a><a class="headerlink" href="#id219" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fpext</span></tt>‘ extends a floating point <tt class="docutils literal"><span class="pre">value</span></tt> to a larger floating
+point value.</p>
+</div>
+<div class="section" id="id220">
+<h5><a class="toc-backref" href="#id983">Arguments:</a><a class="headerlink" href="#id220" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fpext</span></tt>‘ instruction takes a <a class="reference internal" href="#t-floating"><em>floating point</em></a>
+<tt class="docutils literal"><span class="pre">value</span></tt> to cast, and a <a class="reference internal" href="#t-floating"><em>floating point</em></a> type to cast it
+to. The source type must be smaller than the destination type.</p>
+</div>
+<div class="section" id="id221">
+<h5><a class="toc-backref" href="#id984">Semantics:</a><a class="headerlink" href="#id221" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fpext</span></tt>‘ instruction extends the <tt class="docutils literal"><span class="pre">value</span></tt> from a smaller
+<a class="reference internal" href="#t-floating"><em>floating point</em></a> type to a larger <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> type. The <tt class="docutils literal"><span class="pre">fpext</span></tt> cannot be used to make a
+<em>no-op cast</em> because it always changes bits. Use <tt class="docutils literal"><span class="pre">bitcast</span></tt> to make a
+<em>no-op cast</em> for a floating point cast.</p>
+</div>
+<div class="section" id="id222">
+<h5><a class="toc-backref" href="#id985">Example:</a><a class="headerlink" href="#id222" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">fpext</span> <span class="kt">float</span> <span class="m">3.125</span> <span class="k">to</span> <span class="kt">double</span> <span class="c">; yields double:3.125000e+00</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">fpext</span> <span class="kt">double</span> <span class="nv">%X</span> <span class="k">to</span> <span class="kt">fp128</span> <span class="c">; yields fp128:0xL00000000000000004000900000000000</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fptoui-to-instruction">
+<h4><a class="toc-backref" href="#id986">‘<tt class="docutils literal"><span class="pre">fptoui</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#fptoui-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id223">
+<h5><a class="toc-backref" href="#id987">Syntax:</a><a class="headerlink" href="#id223" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fptoui <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id224">
+<h5><a class="toc-backref" href="#id988">Overview:</a><a class="headerlink" href="#id224" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptoui</span></tt>‘ converts a floating point <tt class="docutils literal"><span class="pre">value</span></tt> to its unsigned
+integer equivalent of type <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id225">
+<h5><a class="toc-backref" href="#id989">Arguments:</a><a class="headerlink" href="#id225" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptoui</span></tt>‘ instruction takes a value to cast, which must be a
+scalar or vector <a class="reference internal" href="#t-floating"><em>floating point</em></a> value, and a type to
+cast it to <tt class="docutils literal"><span class="pre">ty2</span></tt>, which must be an <a class="reference internal" href="#t-integer"><em>integer</em></a> type. If
+<tt class="docutils literal"><span class="pre">ty</span></tt> is a vector floating point type, <tt class="docutils literal"><span class="pre">ty2</span></tt> must be a vector integer
+type with the same number of elements as <tt class="docutils literal"><span class="pre">ty</span></tt></p>
+</div>
+<div class="section" id="id226">
+<h5><a class="toc-backref" href="#id990">Semantics:</a><a class="headerlink" href="#id226" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptoui</span></tt>‘ instruction converts its <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> operand into the nearest (rounding towards zero)
+unsigned integer value. If the value cannot fit in <tt class="docutils literal"><span class="pre">ty2</span></tt>, the results
+are undefined.</p>
+</div>
+<div class="section" id="id227">
+<h5><a class="toc-backref" href="#id991">Example:</a><a class="headerlink" href="#id227" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>%X = fptoui double 123.0 to i32 ; yields i32:123
+%Y = fptoui float 1.0E+300 to i1 ; yields undefined:1
+%Z = fptoui float 1.04E+17 to i8 ; yields undefined:1
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="fptosi-to-instruction">
+<h4><a class="toc-backref" href="#id992">‘<tt class="docutils literal"><span class="pre">fptosi</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#fptosi-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id228">
+<h5><a class="toc-backref" href="#id993">Syntax:</a><a class="headerlink" href="#id228" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fptosi <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id229">
+<h5><a class="toc-backref" href="#id994">Overview:</a><a class="headerlink" href="#id229" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptosi</span></tt>‘ instruction converts <a class="reference internal" href="#t-floating"><em>floating point</em></a>
+<tt class="docutils literal"><span class="pre">value</span></tt> to type <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id230">
+<h5><a class="toc-backref" href="#id995">Arguments:</a><a class="headerlink" href="#id230" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptosi</span></tt>‘ instruction takes a value to cast, which must be a
+scalar or vector <a class="reference internal" href="#t-floating"><em>floating point</em></a> value, and a type to
+cast it to <tt class="docutils literal"><span class="pre">ty2</span></tt>, which must be an <a class="reference internal" href="#t-integer"><em>integer</em></a> type. If
+<tt class="docutils literal"><span class="pre">ty</span></tt> is a vector floating point type, <tt class="docutils literal"><span class="pre">ty2</span></tt> must be a vector integer
+type with the same number of elements as <tt class="docutils literal"><span class="pre">ty</span></tt></p>
+</div>
+<div class="section" id="id231">
+<h5><a class="toc-backref" href="#id996">Semantics:</a><a class="headerlink" href="#id231" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fptosi</span></tt>‘ instruction converts its <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> operand into the nearest (rounding towards zero)
+signed integer value. If the value cannot fit in <tt class="docutils literal"><span class="pre">ty2</span></tt>, the results
+are undefined.</p>
+</div>
+<div class="section" id="id232">
+<h5><a class="toc-backref" href="#id997">Example:</a><a class="headerlink" href="#id232" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>%X = fptosi double -123.0 to i32 ; yields i32:-123
+%Y = fptosi float 1.0E-247 to i1 ; yields undefined:1
+%Z = fptosi float 1.04E+17 to i8 ; yields undefined:1
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="uitofp-to-instruction">
+<h4><a class="toc-backref" href="#id998">‘<tt class="docutils literal"><span class="pre">uitofp</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#uitofp-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id233">
+<h5><a class="toc-backref" href="#id999">Syntax:</a><a class="headerlink" href="#id233" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = uitofp <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id234">
+<h5><a class="toc-backref" href="#id1000">Overview:</a><a class="headerlink" href="#id234" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">uitofp</span></tt>‘ instruction regards <tt class="docutils literal"><span class="pre">value</span></tt> as an unsigned integer
+and converts that value to the <tt class="docutils literal"><span class="pre">ty2</span></tt> type.</p>
+</div>
+<div class="section" id="id235">
+<h5><a class="toc-backref" href="#id1001">Arguments:</a><a class="headerlink" href="#id235" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">uitofp</span></tt>‘ instruction takes a value to cast, which must be a
+scalar or vector <a class="reference internal" href="#t-integer"><em>integer</em></a> value, and a type to cast it to
+<tt class="docutils literal"><span class="pre">ty2</span></tt>, which must be an <a class="reference internal" href="#t-floating"><em>floating point</em></a> type. If
+<tt class="docutils literal"><span class="pre">ty</span></tt> is a vector integer type, <tt class="docutils literal"><span class="pre">ty2</span></tt> must be a vector floating point
+type with the same number of elements as <tt class="docutils literal"><span class="pre">ty</span></tt></p>
+</div>
+<div class="section" id="id236">
+<h5><a class="toc-backref" href="#id1002">Semantics:</a><a class="headerlink" href="#id236" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">uitofp</span></tt>‘ instruction interprets its operand as an unsigned
+integer quantity and converts it to the corresponding floating point
+value. If the value cannot fit in the floating point value, the results
+are undefined.</p>
+</div>
+<div class="section" id="id237">
+<h5><a class="toc-backref" href="#id1003">Example:</a><a class="headerlink" href="#id237" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">uitofp</span> <span class="k">i32</span> <span class="m">257</span> <span class="k">to</span> <span class="kt">float</span> <span class="c">; yields float:257.0</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">uitofp</span> <span class="k">i8</span> <span class="m">-1</span> <span class="k">to</span> <span class="kt">double</span> <span class="c">; yields double:255.0</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="sitofp-to-instruction">
+<h4><a class="toc-backref" href="#id1004">‘<tt class="docutils literal"><span class="pre">sitofp</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#sitofp-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id238">
+<h5><a class="toc-backref" href="#id1005">Syntax:</a><a class="headerlink" href="#id238" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = sitofp <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id239">
+<h5><a class="toc-backref" href="#id1006">Overview:</a><a class="headerlink" href="#id239" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">sitofp</span></tt>‘ instruction regards <tt class="docutils literal"><span class="pre">value</span></tt> as a signed integer and
+converts that value to the <tt class="docutils literal"><span class="pre">ty2</span></tt> type.</p>
+</div>
+<div class="section" id="id240">
+<h5><a class="toc-backref" href="#id1007">Arguments:</a><a class="headerlink" href="#id240" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">sitofp</span></tt>‘ instruction takes a value to cast, which must be a
+scalar or vector <a class="reference internal" href="#t-integer"><em>integer</em></a> value, and a type to cast it to
+<tt class="docutils literal"><span class="pre">ty2</span></tt>, which must be an <a class="reference internal" href="#t-floating"><em>floating point</em></a> type. If
+<tt class="docutils literal"><span class="pre">ty</span></tt> is a vector integer type, <tt class="docutils literal"><span class="pre">ty2</span></tt> must be a vector floating point
+type with the same number of elements as <tt class="docutils literal"><span class="pre">ty</span></tt></p>
+</div>
+<div class="section" id="id241">
+<h5><a class="toc-backref" href="#id1008">Semantics:</a><a class="headerlink" href="#id241" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">sitofp</span></tt>‘ instruction interprets its operand as a signed integer
+quantity and converts it to the corresponding floating point value. If
+the value cannot fit in the floating point value, the results are
+undefined.</p>
+</div>
+<div class="section" id="id242">
+<h5><a class="toc-backref" href="#id1009">Example:</a><a class="headerlink" href="#id242" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">sitofp</span> <span class="k">i32</span> <span class="m">257</span> <span class="k">to</span> <span class="kt">float</span> <span class="c">; yields float:257.0</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">sitofp</span> <span class="k">i8</span> <span class="m">-1</span> <span class="k">to</span> <span class="kt">double</span> <span class="c">; yields double:-1.0</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="ptrtoint-to-instruction">
+<span id="i-ptrtoint"></span><h4><a class="toc-backref" href="#id1010">‘<tt class="docutils literal"><span class="pre">ptrtoint</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#ptrtoint-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id243">
+<h5><a class="toc-backref" href="#id1011">Syntax:</a><a class="headerlink" href="#id243" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = ptrtoint <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id244">
+<h5><a class="toc-backref" href="#id1012">Overview:</a><a class="headerlink" href="#id244" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">ptrtoint</span></tt>‘ instruction converts the pointer or a vector of
+pointers <tt class="docutils literal"><span class="pre">value</span></tt> to the integer (or vector of integers) type <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id245">
+<h5><a class="toc-backref" href="#id1013">Arguments:</a><a class="headerlink" href="#id245" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">ptrtoint</span></tt>‘ instruction takes a <tt class="docutils literal"><span class="pre">value</span></tt> to cast, which must be
+a value of type <a class="reference internal" href="#t-pointer"><em>pointer</em></a> or a vector of pointers, and a
+type to cast it to <tt class="docutils literal"><span class="pre">ty2</span></tt>, which must be an <a class="reference internal" href="#t-integer"><em>integer</em></a> or
+a vector of integers type.</p>
+</div>
+<div class="section" id="id246">
+<h5><a class="toc-backref" href="#id1014">Semantics:</a><a class="headerlink" href="#id246" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">ptrtoint</span></tt>‘ instruction converts <tt class="docutils literal"><span class="pre">value</span></tt> to integer type
+<tt class="docutils literal"><span class="pre">ty2</span></tt> by interpreting the pointer value as an integer and either
+truncating or zero extending that value to the size of the integer type.
+If <tt class="docutils literal"><span class="pre">value</span></tt> is smaller than <tt class="docutils literal"><span class="pre">ty2</span></tt> then a zero extension is done. If
+<tt class="docutils literal"><span class="pre">value</span></tt> is larger than <tt class="docutils literal"><span class="pre">ty2</span></tt> then a truncation is done. If they are
+the same size, then nothing is done (<em>no-op cast</em>) other than a type
+change.</p>
+</div>
+<div class="section" id="id247">
+<h5><a class="toc-backref" href="#id1015">Example:</a><a class="headerlink" href="#id247" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">ptrtoint</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%P</span> <span class="k">to</span> <span class="k">i8</span> <span class="c">; yields truncation on 32-bit architecture</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">ptrtoint</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%P</span> <span class="k">to</span> <span class="k">i64</span> <span class="c">; yields zero extension on 32-bit architecture</span>
+<span class="nv">%Z</span> <span class="p">=</span> <span class="k">ptrtoint</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i32</span><span class="p">*></span> <span class="nv">%P</span> <span class="k">to</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i64</span><span class="p">></span><span class="c">; yields vector zero extension for a vector of addresses on 32-bit architecture</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="inttoptr-to-instruction">
+<span id="i-inttoptr"></span><h4><a class="toc-backref" href="#id1016">‘<tt class="docutils literal"><span class="pre">inttoptr</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#inttoptr-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id248">
+<h5><a class="toc-backref" href="#id1017">Syntax:</a><a class="headerlink" href="#id248" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = inttoptr <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id249">
+<h5><a class="toc-backref" href="#id1018">Overview:</a><a class="headerlink" href="#id249" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">inttoptr</span></tt>‘ instruction converts an integer <tt class="docutils literal"><span class="pre">value</span></tt> to a
+pointer type, <tt class="docutils literal"><span class="pre">ty2</span></tt>.</p>
+</div>
+<div class="section" id="id250">
+<h5><a class="toc-backref" href="#id1019">Arguments:</a><a class="headerlink" href="#id250" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">inttoptr</span></tt>‘ instruction takes an <a class="reference internal" href="#t-integer"><em>integer</em></a> value to
+cast, and a type to cast it to, which must be a <a class="reference internal" href="#t-pointer"><em>pointer</em></a>
+type.</p>
+</div>
+<div class="section" id="id251">
+<h5><a class="toc-backref" href="#id1020">Semantics:</a><a class="headerlink" href="#id251" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">inttoptr</span></tt>‘ instruction converts <tt class="docutils literal"><span class="pre">value</span></tt> to type <tt class="docutils literal"><span class="pre">ty2</span></tt> by
+applying either a zero extension or a truncation depending on the size
+of the integer <tt class="docutils literal"><span class="pre">value</span></tt>. If <tt class="docutils literal"><span class="pre">value</span></tt> is larger than the size of a
+pointer then a truncation is done. If <tt class="docutils literal"><span class="pre">value</span></tt> is smaller than the size
+of a pointer then a zero extension is done. If they are the same size,
+nothing is done (<em>no-op cast</em>).</p>
+</div>
+<div class="section" id="id252">
+<h5><a class="toc-backref" href="#id1021">Example:</a><a class="headerlink" href="#id252" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">inttoptr</span> <span class="k">i32</span> <span class="m">255</span> <span class="k">to</span> <span class="k">i32</span><span class="p">*</span> <span class="c">; yields zero extension on 64-bit architecture</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">inttoptr</span> <span class="k">i32</span> <span class="m">255</span> <span class="k">to</span> <span class="k">i32</span><span class="p">*</span> <span class="c">; yields no-op on 32-bit architecture</span>
+<span class="nv">%Z</span> <span class="p">=</span> <span class="k">inttoptr</span> <span class="k">i64</span> <span class="m">0</span> <span class="k">to</span> <span class="k">i32</span><span class="p">*</span> <span class="c">; yields truncation on 32-bit architecture</span>
+<span class="nv">%Z</span> <span class="p">=</span> <span class="k">inttoptr</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i32</span><span class="p">></span> <span class="nv">%G</span> <span class="k">to</span> <span class="p"><</span><span class="m">4</span> <span class="k">x</span> <span class="k">i8</span><span class="p">*></span><span class="c">; yields truncation of vector G to four pointers</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="bitcast-to-instruction">
+<span id="i-bitcast"></span><h4><a class="toc-backref" href="#id1022">‘<tt class="docutils literal"><span class="pre">bitcast</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#bitcast-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id253">
+<h5><a class="toc-backref" href="#id1023">Syntax:</a><a class="headerlink" href="#id253" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = bitcast <ty> <value> to <ty2> ; yields ty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id254">
+<h5><a class="toc-backref" href="#id1024">Overview:</a><a class="headerlink" href="#id254" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">bitcast</span></tt>‘ instruction converts <tt class="docutils literal"><span class="pre">value</span></tt> to type <tt class="docutils literal"><span class="pre">ty2</span></tt> without
+changing any bits.</p>
+</div>
+<div class="section" id="id255">
+<h5><a class="toc-backref" href="#id1025">Arguments:</a><a class="headerlink" href="#id255" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">bitcast</span></tt>‘ instruction takes a value to cast, which must be a
+non-aggregate first class value, and a type to cast it to, which must
+also be a non-aggregate <a class="reference internal" href="#t-firstclass"><em>first class</em></a> type. The
+bit sizes of <tt class="docutils literal"><span class="pre">value</span></tt> and the destination type, <tt class="docutils literal"><span class="pre">ty2</span></tt>, must be
+identical. If the source type is a pointer, the destination type must
+also be a pointer of the same size. This instruction supports bitwise
+conversion of vectors to integers and to vectors of other types (as
+long as they have the same size).</p>
+</div>
+<div class="section" id="id256">
+<h5><a class="toc-backref" href="#id1026">Semantics:</a><a class="headerlink" href="#id256" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">bitcast</span></tt>‘ instruction converts <tt class="docutils literal"><span class="pre">value</span></tt> to type <tt class="docutils literal"><span class="pre">ty2</span></tt>. It
+is always a <em>no-op cast</em> because no bits change with this
+conversion. The conversion is done as if the <tt class="docutils literal"><span class="pre">value</span></tt> had been stored
+to memory and read back as type <tt class="docutils literal"><span class="pre">ty2</span></tt>. Pointer (or vector of
+pointers) types may only be converted to other pointer (or vector of
+pointers) types with the same address space through this instruction.
+To convert pointers to other types, use the <a class="reference internal" href="#i-inttoptr"><em>inttoptr</em></a>
+or <a class="reference internal" href="#i-ptrtoint"><em>ptrtoint</em></a> instructions first.</p>
+</div>
+<div class="section" id="id257">
+<h5><a class="toc-backref" href="#id1027">Example:</a><a class="headerlink" href="#id257" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>%X = bitcast i8 255 to i8 ; yields i8 :-1
+%Y = bitcast i32* %x to sint* ; yields sint*:%x
+%Z = bitcast <2 x int> %V to i64; ; yields i64: %V
+%Z = bitcast <2 x i32*> %V to <2 x i64*> ; yields <2 x i64*>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="addrspacecast-to-instruction">
+<span id="i-addrspacecast"></span><h4><a class="toc-backref" href="#id1028">‘<tt class="docutils literal"><span class="pre">addrspacecast</span> <span class="pre">..</span> <span class="pre">to</span></tt>‘ Instruction</a><a class="headerlink" href="#addrspacecast-to-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id258">
+<h5><a class="toc-backref" href="#id1029">Syntax:</a><a class="headerlink" href="#id258" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = addrspacecast <pty> <ptrval> to <pty2> ; yields pty2
+</pre></div>
+</div>
+</div>
+<div class="section" id="id259">
+<h5><a class="toc-backref" href="#id1030">Overview:</a><a class="headerlink" href="#id259" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">addrspacecast</span></tt>‘ instruction converts <tt class="docutils literal"><span class="pre">ptrval</span></tt> from <tt class="docutils literal"><span class="pre">pty</span></tt> in
+address space <tt class="docutils literal"><span class="pre">n</span></tt> to type <tt class="docutils literal"><span class="pre">pty2</span></tt> in address space <tt class="docutils literal"><span class="pre">m</span></tt>.</p>
+</div>
+<div class="section" id="id260">
+<h5><a class="toc-backref" href="#id1031">Arguments:</a><a class="headerlink" href="#id260" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">addrspacecast</span></tt>‘ instruction takes a pointer or vector of pointer value
+to cast and a pointer type to cast it to, which must have a different
+address space.</p>
+</div>
+<div class="section" id="id261">
+<h5><a class="toc-backref" href="#id1032">Semantics:</a><a class="headerlink" href="#id261" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">addrspacecast</span></tt>‘ instruction converts the pointer value
+<tt class="docutils literal"><span class="pre">ptrval</span></tt> to type <tt class="docutils literal"><span class="pre">pty2</span></tt>. It can be a <em>no-op cast</em> or a complex
+value modification, depending on the target and the address space
+pair. Pointer conversions within the same address space must be
+performed with the <tt class="docutils literal"><span class="pre">bitcast</span></tt> instruction. Note that if the address space
+conversion is legal then both result and operand refer to the same memory
+location.</p>
+</div>
+<div class="section" id="id262">
+<h5><a class="toc-backref" href="#id1033">Example:</a><a class="headerlink" href="#id262" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>%X = addrspacecast i32* %x to i32 addrspace(1)* ; yields i32 addrspace(1)*:%x
+%Y = addrspacecast i32 addrspace(1)* %y to i64 addrspace(2)* ; yields i64 addrspace(2)*:%y
+%Z = addrspacecast <4 x i32*> %z to <4 x float addrspace(3)*> ; yields <4 x float addrspace(3)*>:%z
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="other-operations">
+<span id="otherops"></span><h3><a class="toc-backref" href="#id1034">Other Operations</a><a class="headerlink" href="#other-operations" title="Permalink to this headline">¶</a></h3>
+<p>The instructions in this category are the “miscellaneous” instructions,
+which defy better classification.</p>
+<div class="section" id="icmp-instruction">
+<span id="i-icmp"></span><h4><a class="toc-backref" href="#id1035">‘<tt class="docutils literal"><span class="pre">icmp</span></tt>‘ Instruction</a><a class="headerlink" href="#icmp-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id263">
+<h5><a class="toc-backref" href="#id1036">Syntax:</a><a class="headerlink" href="#id263" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = icmp <cond> <ty> <op1>, <op2> ; yields i1 or <N x i1>:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id264">
+<h5><a class="toc-backref" href="#id1037">Overview:</a><a class="headerlink" href="#id264" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">icmp</span></tt>‘ instruction returns a boolean value or a vector of
+boolean values based on comparison of its two integer, integer vector,
+pointer, or pointer vector operands.</p>
+</div>
+<div class="section" id="id265">
+<h5><a class="toc-backref" href="#id1038">Arguments:</a><a class="headerlink" href="#id265" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">icmp</span></tt>‘ instruction takes three operands. The first operand is
+the condition code indicating the kind of comparison to perform. It is
+not a value, just a keyword. The possible condition code are:</p>
+<ol class="arabic simple">
+<li><tt class="docutils literal"><span class="pre">eq</span></tt>: equal</li>
+<li><tt class="docutils literal"><span class="pre">ne</span></tt>: not equal</li>
+<li><tt class="docutils literal"><span class="pre">ugt</span></tt>: unsigned greater than</li>
+<li><tt class="docutils literal"><span class="pre">uge</span></tt>: unsigned greater or equal</li>
+<li><tt class="docutils literal"><span class="pre">ult</span></tt>: unsigned less than</li>
+<li><tt class="docutils literal"><span class="pre">ule</span></tt>: unsigned less or equal</li>
+<li><tt class="docutils literal"><span class="pre">sgt</span></tt>: signed greater than</li>
+<li><tt class="docutils literal"><span class="pre">sge</span></tt>: signed greater or equal</li>
+<li><tt class="docutils literal"><span class="pre">slt</span></tt>: signed less than</li>
+<li><tt class="docutils literal"><span class="pre">sle</span></tt>: signed less or equal</li>
+</ol>
+<p>The remaining two arguments must be <a class="reference internal" href="#t-integer"><em>integer</em></a> or
+<a class="reference internal" href="#t-pointer"><em>pointer</em></a> or integer <a class="reference internal" href="#t-vector"><em>vector</em></a> typed. They
+must also be identical types.</p>
+</div>
+<div class="section" id="id266">
+<h5><a class="toc-backref" href="#id1039">Semantics:</a><a class="headerlink" href="#id266" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">icmp</span></tt>‘ compares <tt class="docutils literal"><span class="pre">op1</span></tt> and <tt class="docutils literal"><span class="pre">op2</span></tt> according to the condition
+code given as <tt class="docutils literal"><span class="pre">cond</span></tt>. The comparison performed always yields either an
+<a class="reference internal" href="#t-integer"><em>i1</em></a> or vector of <tt class="docutils literal"><span class="pre">i1</span></tt> result, as follows:</p>
+<ol class="arabic simple">
+<li><tt class="docutils literal"><span class="pre">eq</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if the operands are equal, <tt class="docutils literal"><span class="pre">false</span></tt>
+otherwise. No sign interpretation is necessary or performed.</li>
+<li><tt class="docutils literal"><span class="pre">ne</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if the operands are unequal, <tt class="docutils literal"><span class="pre">false</span></tt>
+otherwise. No sign interpretation is necessary or performed.</li>
+<li><tt class="docutils literal"><span class="pre">ugt</span></tt>: interprets the operands as unsigned values and yields
+<tt class="docutils literal"><span class="pre">true</span></tt> if <tt class="docutils literal"><span class="pre">op1</span></tt> is greater than <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">uge</span></tt>: interprets the operands as unsigned values and yields
+<tt class="docutils literal"><span class="pre">true</span></tt> if <tt class="docutils literal"><span class="pre">op1</span></tt> is greater than or equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ult</span></tt>: interprets the operands as unsigned values and yields
+<tt class="docutils literal"><span class="pre">true</span></tt> if <tt class="docutils literal"><span class="pre">op1</span></tt> is less than <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ule</span></tt>: interprets the operands as unsigned values and yields
+<tt class="docutils literal"><span class="pre">true</span></tt> if <tt class="docutils literal"><span class="pre">op1</span></tt> is less than or equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">sgt</span></tt>: interprets the operands as signed values and yields <tt class="docutils literal"><span class="pre">true</span></tt>
+if <tt class="docutils literal"><span class="pre">op1</span></tt> is greater than <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">sge</span></tt>: interprets the operands as signed values and yields <tt class="docutils literal"><span class="pre">true</span></tt>
+if <tt class="docutils literal"><span class="pre">op1</span></tt> is greater than or equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">slt</span></tt>: interprets the operands as signed values and yields <tt class="docutils literal"><span class="pre">true</span></tt>
+if <tt class="docutils literal"><span class="pre">op1</span></tt> is less than <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">sle</span></tt>: interprets the operands as signed values and yields <tt class="docutils literal"><span class="pre">true</span></tt>
+if <tt class="docutils literal"><span class="pre">op1</span></tt> is less than or equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+</ol>
+<p>If the operands are <a class="reference internal" href="#t-pointer"><em>pointer</em></a> typed, the pointer values
+are compared as if they were integers.</p>
+<p>If the operands are integer vectors, then they are compared element by
+element. The result is an <tt class="docutils literal"><span class="pre">i1</span></tt> vector with the same number of elements
+as the values being compared. Otherwise, the result is an <tt class="docutils literal"><span class="pre">i1</span></tt>.</p>
+</div>
+<div class="section" id="id267">
+<h5><a class="toc-backref" href="#id1040">Example:</a><a class="headerlink" href="#id267" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = icmp eq i32 4, 5 ; yields: result=false
+<result> = icmp ne float* %X, %X ; yields: result=false
+<result> = icmp ult i16 4, 5 ; yields: result=true
+<result> = icmp sgt i16 4, 5 ; yields: result=false
+<result> = icmp ule i16 -4, 5 ; yields: result=false
+<result> = icmp sge i16 4, 5 ; yields: result=false
+</pre></div>
+</div>
+<p>Note that the code generator does not yet support vector types with the
+<tt class="docutils literal"><span class="pre">icmp</span></tt> instruction.</p>
+</div>
+</div>
+<div class="section" id="fcmp-instruction">
+<span id="i-fcmp"></span><h4><a class="toc-backref" href="#id1041">‘<tt class="docutils literal"><span class="pre">fcmp</span></tt>‘ Instruction</a><a class="headerlink" href="#fcmp-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id268">
+<h5><a class="toc-backref" href="#id1042">Syntax:</a><a class="headerlink" href="#id268" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = fcmp [fast-math flags]* <cond> <ty> <op1>, <op2> ; yields i1 or <N x i1>:result
+</pre></div>
+</div>
+</div>
+<div class="section" id="id269">
+<h5><a class="toc-backref" href="#id1043">Overview:</a><a class="headerlink" href="#id269" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fcmp</span></tt>‘ instruction returns a boolean value or vector of boolean
+values based on comparison of its operands.</p>
+<p>If the operands are floating point scalars, then the result type is a
+boolean (<a class="reference internal" href="#t-integer"><em>i1</em></a>).</p>
+<p>If the operands are floating point vectors, then the result type is a
+vector of boolean with the same number of elements as the operands being
+compared.</p>
+</div>
+<div class="section" id="id270">
+<h5><a class="toc-backref" href="#id1044">Arguments:</a><a class="headerlink" href="#id270" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fcmp</span></tt>‘ instruction takes three operands. The first operand is
+the condition code indicating the kind of comparison to perform. It is
+not a value, just a keyword. The possible condition code are:</p>
+<ol class="arabic simple">
+<li><tt class="docutils literal"><span class="pre">false</span></tt>: no comparison, always returns false</li>
+<li><tt class="docutils literal"><span class="pre">oeq</span></tt>: ordered and equal</li>
+<li><tt class="docutils literal"><span class="pre">ogt</span></tt>: ordered and greater than</li>
+<li><tt class="docutils literal"><span class="pre">oge</span></tt>: ordered and greater than or equal</li>
+<li><tt class="docutils literal"><span class="pre">olt</span></tt>: ordered and less than</li>
+<li><tt class="docutils literal"><span class="pre">ole</span></tt>: ordered and less than or equal</li>
+<li><tt class="docutils literal"><span class="pre">one</span></tt>: ordered and not equal</li>
+<li><tt class="docutils literal"><span class="pre">ord</span></tt>: ordered (no nans)</li>
+<li><tt class="docutils literal"><span class="pre">ueq</span></tt>: unordered or equal</li>
+<li><tt class="docutils literal"><span class="pre">ugt</span></tt>: unordered or greater than</li>
+<li><tt class="docutils literal"><span class="pre">uge</span></tt>: unordered or greater than or equal</li>
+<li><tt class="docutils literal"><span class="pre">ult</span></tt>: unordered or less than</li>
+<li><tt class="docutils literal"><span class="pre">ule</span></tt>: unordered or less than or equal</li>
+<li><tt class="docutils literal"><span class="pre">une</span></tt>: unordered or not equal</li>
+<li><tt class="docutils literal"><span class="pre">uno</span></tt>: unordered (either nans)</li>
+<li><tt class="docutils literal"><span class="pre">true</span></tt>: no comparison, always returns true</li>
+</ol>
+<p><em>Ordered</em> means that neither operand is a QNAN while <em>unordered</em> means
+that either operand may be a QNAN.</p>
+<p>Each of <tt class="docutils literal"><span class="pre">val1</span></tt> and <tt class="docutils literal"><span class="pre">val2</span></tt> arguments must be either a <a class="reference internal" href="#t-floating"><em>floating
+point</em></a> type or a <a class="reference internal" href="#t-vector"><em>vector</em></a> of floating point
+type. They must have identical types.</p>
+</div>
+<div class="section" id="id271">
+<h5><a class="toc-backref" href="#id1045">Semantics:</a><a class="headerlink" href="#id271" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">fcmp</span></tt>‘ instruction compares <tt class="docutils literal"><span class="pre">op1</span></tt> and <tt class="docutils literal"><span class="pre">op2</span></tt> according to the
+condition code given as <tt class="docutils literal"><span class="pre">cond</span></tt>. If the operands are vectors, then the
+vectors are compared element by element. Each comparison performed
+always yields an <a class="reference internal" href="#t-integer"><em>i1</em></a> result, as follows:</p>
+<ol class="arabic simple">
+<li><tt class="docutils literal"><span class="pre">false</span></tt>: always yields <tt class="docutils literal"><span class="pre">false</span></tt>, regardless of operands.</li>
+<li><tt class="docutils literal"><span class="pre">oeq</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if both operands are not a QNAN and <tt class="docutils literal"><span class="pre">op1</span></tt>
+is equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ogt</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if both operands are not a QNAN and <tt class="docutils literal"><span class="pre">op1</span></tt>
+is greater than <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">oge</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if both operands are not a QNAN and <tt class="docutils literal"><span class="pre">op1</span></tt>
+is greater than or equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">olt</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if both operands are not a QNAN and <tt class="docutils literal"><span class="pre">op1</span></tt>
+is less than <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ole</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if both operands are not a QNAN and <tt class="docutils literal"><span class="pre">op1</span></tt>
+is less than or equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">one</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if both operands are not a QNAN and <tt class="docutils literal"><span class="pre">op1</span></tt>
+is not equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ord</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if both operands are not a QNAN.</li>
+<li><tt class="docutils literal"><span class="pre">ueq</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if either operand is a QNAN or <tt class="docutils literal"><span class="pre">op1</span></tt> is
+equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ugt</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if either operand is a QNAN or <tt class="docutils literal"><span class="pre">op1</span></tt> is
+greater than <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">uge</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if either operand is a QNAN or <tt class="docutils literal"><span class="pre">op1</span></tt> is
+greater than or equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ult</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if either operand is a QNAN or <tt class="docutils literal"><span class="pre">op1</span></tt> is
+less than <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">ule</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if either operand is a QNAN or <tt class="docutils literal"><span class="pre">op1</span></tt> is
+less than or equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">une</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if either operand is a QNAN or <tt class="docutils literal"><span class="pre">op1</span></tt> is
+not equal to <tt class="docutils literal"><span class="pre">op2</span></tt>.</li>
+<li><tt class="docutils literal"><span class="pre">uno</span></tt>: yields <tt class="docutils literal"><span class="pre">true</span></tt> if either operand is a QNAN.</li>
+<li><tt class="docutils literal"><span class="pre">true</span></tt>: always yields <tt class="docutils literal"><span class="pre">true</span></tt>, regardless of operands.</li>
+</ol>
+<p>The <tt class="docutils literal"><span class="pre">fcmp</span></tt> instruction can also optionally take any number of
+<a class="reference internal" href="#fastmath"><em>fast-math flags</em></a>, which are optimization hints to enable
+otherwise unsafe floating point optimizations.</p>
+<p>Any set of fast-math flags are legal on an <tt class="docutils literal"><span class="pre">fcmp</span></tt> instruction, but the
+only flags that have any effect on its semantics are those that allow
+assumptions to be made about the values of input arguments; namely
+<tt class="docutils literal"><span class="pre">nnan</span></tt>, <tt class="docutils literal"><span class="pre">ninf</span></tt>, and <tt class="docutils literal"><span class="pre">nsz</span></tt>. See <a class="reference internal" href="#fastmath"><em>Fast-Math Flags</em></a> for more information.</p>
+</div>
+<div class="section" id="id272">
+<h5><a class="toc-backref" href="#id1046">Example:</a><a class="headerlink" href="#id272" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><result> = fcmp oeq float 4.0, 5.0 ; yields: result=false
+<result> = fcmp one float 4.0, 5.0 ; yields: result=true
+<result> = fcmp olt float 4.0, 5.0 ; yields: result=true
+<result> = fcmp ueq double 1.0, 2.0 ; yields: result=false
+</pre></div>
+</div>
+<p>Note that the code generator does not yet support vector types with the
+<tt class="docutils literal"><span class="pre">fcmp</span></tt> instruction.</p>
+</div>
+</div>
+<div class="section" id="phi-instruction">
+<span id="i-phi"></span><h4><a class="toc-backref" href="#id1047">‘<tt class="docutils literal"><span class="pre">phi</span></tt>‘ Instruction</a><a class="headerlink" href="#phi-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id273">
+<h5><a class="toc-backref" href="#id1048">Syntax:</a><a class="headerlink" href="#id273" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = phi <ty> [ <val0>, <label0>], ...
+</pre></div>
+</div>
+</div>
+<div class="section" id="id274">
+<h5><a class="toc-backref" href="#id1049">Overview:</a><a class="headerlink" href="#id274" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">phi</span></tt>‘ instruction is used to implement the Ï node in the SSA
+graph representing the function.</p>
+</div>
+<div class="section" id="id275">
+<h5><a class="toc-backref" href="#id1050">Arguments:</a><a class="headerlink" href="#id275" title="Permalink to this headline">¶</a></h5>
+<p>The type of the incoming values is specified with the first type field.
+After this, the ‘<tt class="docutils literal"><span class="pre">phi</span></tt>‘ instruction takes a list of pairs as
+arguments, with one pair for each predecessor basic block of the current
+block. Only values of <a class="reference internal" href="#t-firstclass"><em>first class</em></a> type may be used as
+the value arguments to the PHI node. Only labels may be used as the
+label arguments.</p>
+<p>There must be no non-phi instructions between the start of a basic block
+and the PHI instructions: i.e. PHI instructions must be first in a basic
+block.</p>
+<p>For the purposes of the SSA form, the use of each incoming value is
+deemed to occur on the edge from the corresponding predecessor block to
+the current block (but after any definition of an ‘<tt class="docutils literal"><span class="pre">invoke</span></tt>‘
+instruction’s return value on the same edge).</p>
+</div>
+<div class="section" id="id276">
+<h5><a class="toc-backref" href="#id1051">Semantics:</a><a class="headerlink" href="#id276" title="Permalink to this headline">¶</a></h5>
+<p>At runtime, the ‘<tt class="docutils literal"><span class="pre">phi</span></tt>‘ instruction logically takes on the value
+specified by the pair corresponding to the predecessor basic block that
+executed just prior to the current block.</p>
+</div>
+<div class="section" id="id277">
+<h5><a class="toc-backref" href="#id1052">Example:</a><a class="headerlink" href="#id277" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nl">Loop:</span> <span class="c">; Infinite loop that counts from 0 on up...</span>
+ <span class="nv">%indvar</span> <span class="p">=</span> <span class="k">phi</span> <span class="k">i32</span> <span class="p">[</span> <span class="m">0</span><span class="p">,</span> <span class="nv">%LoopHeader</span> <span class="p">],</span> <span class="p">[</span> <span class="nv">%nextindvar</span><span class="p">,</span> <span class="nv">%Loop</span> <span class="p">]</span>
+ <span class="nv">%nextindvar</span> <span class="p">=</span> <span class="k">add</span> <span class="k">i32</span> <span class="nv">%indvar</span><span class="p">,</span> <span class="m">1</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%Loop</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="select-instruction">
+<span id="i-select"></span><h4><a class="toc-backref" href="#id1053">‘<tt class="docutils literal"><span class="pre">select</span></tt>‘ Instruction</a><a class="headerlink" href="#select-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id278">
+<h5><a class="toc-backref" href="#id1054">Syntax:</a><a class="headerlink" href="#id278" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = select selty <cond>, <ty> <val1>, <ty> <val2> ; yields ty
+
+selty is either i1 or {<N x i1>}
+</pre></div>
+</div>
+</div>
+<div class="section" id="id279">
+<h5><a class="toc-backref" href="#id1055">Overview:</a><a class="headerlink" href="#id279" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">select</span></tt>‘ instruction is used to choose one value based on a
+condition, without IR-level branching.</p>
+</div>
+<div class="section" id="id280">
+<h5><a class="toc-backref" href="#id1056">Arguments:</a><a class="headerlink" href="#id280" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">select</span></tt>‘ instruction requires an ‘i1’ value or a vector of ‘i1’
+values indicating the condition, and two values of the same <a class="reference internal" href="#t-firstclass"><em>first
+class</em></a> type.</p>
+</div>
+<div class="section" id="id281">
+<h5><a class="toc-backref" href="#id1057">Semantics:</a><a class="headerlink" href="#id281" title="Permalink to this headline">¶</a></h5>
+<p>If the condition is an i1 and it evaluates to 1, the instruction returns
+the first value argument; otherwise, it returns the second value
+argument.</p>
+<p>If the condition is a vector of i1, then the value arguments must be
+vectors of the same size, and the selection is done element by element.</p>
+<p>If the condition is an i1 and the value arguments are vectors of the
+same size, then an entire vector is selected.</p>
+</div>
+<div class="section" id="id282">
+<h5><a class="toc-backref" href="#id1058">Example:</a><a class="headerlink" href="#id282" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%X</span> <span class="p">=</span> <span class="k">select</span> <span class="k">i1</span> <span class="k">true</span><span class="p">,</span> <span class="k">i8</span> <span class="m">17</span><span class="p">,</span> <span class="k">i8</span> <span class="m">42</span> <span class="c">; yields i8:17</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="call-instruction">
+<span id="i-call"></span><h4><a class="toc-backref" href="#id1059">‘<tt class="docutils literal"><span class="pre">call</span></tt>‘ Instruction</a><a class="headerlink" href="#call-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id283">
+<h5><a class="toc-backref" href="#id1060">Syntax:</a><a class="headerlink" href="#id283" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><result> = [tail | musttail] call [cconv] [ret attrs] <ty> [<fnty>*] <fnptrval>(<function args>) [fn attrs]
+</pre></div>
+</div>
+</div>
+<div class="section" id="id284">
+<h5><a class="toc-backref" href="#id1061">Overview:</a><a class="headerlink" href="#id284" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">call</span></tt>‘ instruction represents a simple function call.</p>
+</div>
+<div class="section" id="id285">
+<h5><a class="toc-backref" href="#id1062">Arguments:</a><a class="headerlink" href="#id285" title="Permalink to this headline">¶</a></h5>
+<p>This instruction requires several arguments:</p>
+<ol class="arabic">
+<li><p class="first">The optional <tt class="docutils literal"><span class="pre">tail</span></tt> and <tt class="docutils literal"><span class="pre">musttail</span></tt> markers indicate that the optimizers
+should perform tail call optimization. The <tt class="docutils literal"><span class="pre">tail</span></tt> marker is a hint that
+<a class="reference external" href="CodeGenerator.html#sibcallopt">can be ignored</a>. The <tt class="docutils literal"><span class="pre">musttail</span></tt> marker
+means that the call must be tail call optimized in order for the program to
+be correct. The <tt class="docutils literal"><span class="pre">musttail</span></tt> marker provides these guarantees:</p>
+<ol class="arabic simple">
+<li>The call will not cause unbounded stack growth if it is part of a
+recursive cycle in the call graph.</li>
+<li>Arguments with the <a class="reference internal" href="#attr-inalloca"><em>inalloca</em></a> attribute are
+forwarded in place.</li>
+</ol>
+<p>Both markers imply that the callee does not access allocas or varargs from
+the caller. Calls marked <tt class="docutils literal"><span class="pre">musttail</span></tt> must obey the following additional
+rules:</p>
+<ul class="simple">
+<li>The call must immediately precede a <a class="reference internal" href="#i-ret"><em>ret</em></a> instruction,
+or a pointer bitcast followed by a ret instruction.</li>
+<li>The ret instruction must return the (possibly bitcasted) value
+produced by the call or void.</li>
+<li>The caller and callee prototypes must match. Pointer types of
+parameters or return types may differ in pointee type, but not
+in address space.</li>
+<li>The calling conventions of the caller and callee must match.</li>
+<li>All ABI-impacting function attributes, such as sret, byval, inreg,
+returned, and inalloca, must match.</li>
+<li>The callee must be varargs iff the caller is varargs. Bitcasting a
+non-varargs function to the appropriate varargs type is legal so
+long as the non-varargs prefixes obey the other rules.</li>
+</ul>
+<p>Tail call optimization for calls marked <tt class="docutils literal"><span class="pre">tail</span></tt> is guaranteed to occur if
+the following conditions are met:</p>
+<ul class="simple">
+<li>Caller and callee both have the calling convention <tt class="docutils literal"><span class="pre">fastcc</span></tt>.</li>
+<li>The call is in tail position (ret immediately follows call and ret
+uses value of call or is void).</li>
+<li>Option <tt class="docutils literal"><span class="pre">-tailcallopt</span></tt> is enabled, or
+<tt class="docutils literal"><span class="pre">llvm::GuaranteedTailCallOpt</span></tt> is <tt class="docutils literal"><span class="pre">true</span></tt>.</li>
+<li><a class="reference external" href="CodeGenerator.html#tailcallopt">Platform-specific constraints are
+met.</a></li>
+</ul>
+</li>
+<li><p class="first">The optional “cconv” marker indicates which <a class="reference internal" href="#callingconv"><em>calling
+convention</em></a> the call should use. If none is
+specified, the call defaults to using C calling conventions. The
+calling convention of the call must match the calling convention of
+the target function, or else the behavior is undefined.</p>
+</li>
+<li><p class="first">The optional <a class="reference internal" href="#paramattrs"><em>Parameter Attributes</em></a> list for return
+values. Only ‘<tt class="docutils literal"><span class="pre">zeroext</span></tt>‘, ‘<tt class="docutils literal"><span class="pre">signext</span></tt>‘, and ‘<tt class="docutils literal"><span class="pre">inreg</span></tt>‘ attributes
+are valid here.</p>
+</li>
+<li><p class="first">‘<tt class="docutils literal"><span class="pre">ty</span></tt>‘: the type of the call instruction itself which is also the
+type of the return value. Functions that return no value are marked
+<tt class="docutils literal"><span class="pre">void</span></tt>.</p>
+</li>
+<li><p class="first">‘<tt class="docutils literal"><span class="pre">fnty</span></tt>‘: shall be the signature of the pointer to function value
+being invoked. The argument types must match the types implied by
+this signature. This type can be omitted if the function is not
+varargs and if the function type does not return a pointer to a
+function.</p>
+</li>
+<li><p class="first">‘<tt class="docutils literal"><span class="pre">fnptrval</span></tt>‘: An LLVM value containing a pointer to a function to
+be invoked. In most cases, this is a direct function invocation, but
+indirect <tt class="docutils literal"><span class="pre">call</span></tt>‘s are just as possible, calling an arbitrary pointer
+to function value.</p>
+</li>
+<li><p class="first">‘<tt class="docutils literal"><span class="pre">function</span> <span class="pre">args</span></tt>‘: argument list whose types match the function
+signature argument types and parameter attributes. All arguments must
+be of <a class="reference internal" href="#t-firstclass"><em>first class</em></a> type. If the function signature
+indicates the function accepts a variable number of arguments, the
+extra arguments can be specified.</p>
+</li>
+<li><p class="first">The optional <a class="reference internal" href="#fnattrs"><em>function attributes</em></a> list. Only
+‘<tt class="docutils literal"><span class="pre">noreturn</span></tt>‘, ‘<tt class="docutils literal"><span class="pre">nounwind</span></tt>‘, ‘<tt class="docutils literal"><span class="pre">readonly</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">readnone</span></tt>‘
+attributes are valid here.</p>
+</li>
+</ol>
+</div>
+<div class="section" id="id286">
+<h5><a class="toc-backref" href="#id1063">Semantics:</a><a class="headerlink" href="#id286" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">call</span></tt>‘ instruction is used to cause control flow to transfer to
+a specified function, with its incoming arguments bound to the specified
+values. Upon a ‘<tt class="docutils literal"><span class="pre">ret</span></tt>‘ instruction in the called function, control
+flow continues with the instruction after the function call, and the
+return value of the function is bound to the result argument.</p>
+</div>
+<div class="section" id="id287">
+<h5><a class="toc-backref" href="#id1064">Example:</a><a class="headerlink" href="#id287" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%retval</span> <span class="p">=</span> <span class="k">call</span> <span class="k">i32</span> <span class="vg">@test</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%argc</span><span class="p">)</span>
+<span class="k">call</span> <span class="k">i32</span> <span class="p">(</span><span class="k">i8</span><span class="p">*,</span> <span class="p">...)*</span> <span class="vg">@printf</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%msg</span><span class="p">,</span> <span class="k">i32</span> <span class="m">12</span><span class="p">,</span> <span class="k">i8</span> <span class="m">42</span><span class="p">)</span> <span class="c">; yields i32</span>
+<span class="nv">%X</span> <span class="p">=</span> <span class="k">tail</span> <span class="k">call</span> <span class="k">i32</span> <span class="vg">@foo</span><span class="p">()</span> <span class="c">; yields i32</span>
+<span class="nv">%Y</span> <span class="p">=</span> <span class="k">tail</span> <span class="k">call</span> <span class="k">fastcc</span> <span class="k">i32</span> <span class="vg">@foo</span><span class="p">()</span> <span class="c">; yields i32</span>
+<span class="k">call</span> <span class="kt">void</span> <span class="nv">%foo</span><span class="p">(</span><span class="k">i8</span> <span class="m">97</span> <span class="k">signext</span><span class="p">)</span>
+
+<span class="nv">%struct.A</span> <span class="p">=</span> <span class="k">type</span> <span class="p">{</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i8</span> <span class="p">}</span>
+<span class="nv">%r</span> <span class="p">=</span> <span class="k">call</span> <span class="nv">%struct.A</span> <span class="vg">@foo</span><span class="p">()</span> <span class="c">; yields { i32, i8 }</span>
+<span class="nv">%gr</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="nv">%struct.A</span> <span class="nv">%r</span><span class="p">,</span> <span class="m">0</span> <span class="c">; yields i32</span>
+<span class="nv">%gr1</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="nv">%struct.A</span> <span class="nv">%r</span><span class="p">,</span> <span class="m">1</span> <span class="c">; yields i8</span>
+<span class="nv">%Z</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">void</span> <span class="vg">@foo</span><span class="p">()</span> <span class="k">noreturn</span> <span class="c">; indicates that %foo never returns normally</span>
+<span class="nv">%ZZ</span> <span class="p">=</span> <span class="k">call</span> <span class="k">zeroext</span> <span class="k">i32</span> <span class="vg">@bar</span><span class="p">()</span> <span class="c">; Return value is %zero extended</span>
+</pre></div>
+</div>
+<p>llvm treats calls to some functions with names and arguments that match
+the standard C99 library as being the C99 library functions, and may
+perform optimizations or generate code for them under that assumption.
+This is something we’d like to change in the future to provide better
+support for freestanding environments and non-C-based languages.</p>
+</div>
+</div>
+<div class="section" id="va-arg-instruction">
+<span id="i-va-arg"></span><h4><a class="toc-backref" href="#id1065">‘<tt class="docutils literal"><span class="pre">va_arg</span></tt>‘ Instruction</a><a class="headerlink" href="#va-arg-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id288">
+<h5><a class="toc-backref" href="#id1066">Syntax:</a><a class="headerlink" href="#id288" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><resultval> = va_arg <va_list*> <arglist>, <argty>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id289">
+<h5><a class="toc-backref" href="#id1067">Overview:</a><a class="headerlink" href="#id289" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">va_arg</span></tt>‘ instruction is used to access arguments passed through
+the “variable argument” area of a function call. It is used to implement
+the <tt class="docutils literal"><span class="pre">va_arg</span></tt> macro in C.</p>
+</div>
+<div class="section" id="id290">
+<h5><a class="toc-backref" href="#id1068">Arguments:</a><a class="headerlink" href="#id290" title="Permalink to this headline">¶</a></h5>
+<p>This instruction takes a <tt class="docutils literal"><span class="pre">va_list*</span></tt> value and the type of the
+argument. It returns a value of the specified argument type and
+increments the <tt class="docutils literal"><span class="pre">va_list</span></tt> to point to the next argument. The actual
+type of <tt class="docutils literal"><span class="pre">va_list</span></tt> is target specific.</p>
+</div>
+<div class="section" id="id291">
+<h5><a class="toc-backref" href="#id1069">Semantics:</a><a class="headerlink" href="#id291" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">va_arg</span></tt>‘ instruction loads an argument of the specified type
+from the specified <tt class="docutils literal"><span class="pre">va_list</span></tt> and causes the <tt class="docutils literal"><span class="pre">va_list</span></tt> to point to
+the next argument. For more information, see the variable argument
+handling <a class="reference internal" href="#int-varargs"><em>Intrinsic Functions</em></a>.</p>
+<p>It is legal for this instruction to be called in a function which does
+not take a variable number of arguments, for example, the <tt class="docutils literal"><span class="pre">vfprintf</span></tt>
+function.</p>
+<p><tt class="docutils literal"><span class="pre">va_arg</span></tt> is an LLVM instruction instead of an <a class="reference internal" href="#intrinsics"><em>intrinsic
+function</em></a> because it takes a type as an argument.</p>
+</div>
+<div class="section" id="id292">
+<h5><a class="toc-backref" href="#id1070">Example:</a><a class="headerlink" href="#id292" title="Permalink to this headline">¶</a></h5>
+<p>See the <a class="reference internal" href="#int-varargs"><em>variable argument processing</em></a> section.</p>
+<p>Note that the code generator does not yet fully support va_arg on many
+targets. Also, it does not currently support va_arg with aggregate
+types on any target.</p>
+</div>
+</div>
+<div class="section" id="landingpad-instruction">
+<span id="i-landingpad"></span><h4><a class="toc-backref" href="#id1071">‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ Instruction</a><a class="headerlink" href="#landingpad-instruction" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id293">
+<h5><a class="toc-backref" href="#id1072">Syntax:</a><a class="headerlink" href="#id293" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre><resultval> = landingpad <resultty> <clause>+
+<resultval> = landingpad <resultty> cleanup <clause>*
+
+<clause> := catch <type> <value>
+<clause> := filter <array constant type> <array constant>
+</pre></div>
+</div>
+</div>
+<div class="section" id="id294">
+<h5><a class="toc-backref" href="#id1073">Overview:</a><a class="headerlink" href="#id294" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ instruction is used by <a class="reference external" href="ExceptionHandling.html#overview">LLVM’s exception handling
+system</a> to specify that a basic block
+is a landing pad — one where the exception lands, and corresponds to the
+code found in the <tt class="docutils literal"><span class="pre">catch</span></tt> portion of a <tt class="docutils literal"><span class="pre">try</span></tt>/<tt class="docutils literal"><span class="pre">catch</span></tt> sequence. It
+defines values supplied by the <a class="reference internal" href="#personalityfn"><em>personality function</em></a> upon
+re-entry to the function. The <tt class="docutils literal"><span class="pre">resultval</span></tt> has the type <tt class="docutils literal"><span class="pre">resultty</span></tt>.</p>
+</div>
+<div class="section" id="id295">
+<h5><a class="toc-backref" href="#id1074">Arguments:</a><a class="headerlink" href="#id295" title="Permalink to this headline">¶</a></h5>
+<p>The optional
+<tt class="docutils literal"><span class="pre">cleanup</span></tt> flag indicates that the landing pad block is a cleanup.</p>
+<p>A <tt class="docutils literal"><span class="pre">clause</span></tt> begins with the clause type — <tt class="docutils literal"><span class="pre">catch</span></tt> or <tt class="docutils literal"><span class="pre">filter</span></tt> — and
+contains the global variable representing the “type” that may be caught
+or filtered respectively. Unlike the <tt class="docutils literal"><span class="pre">catch</span></tt> clause, the <tt class="docutils literal"><span class="pre">filter</span></tt>
+clause takes an array constant as its argument. Use
+“<tt class="docutils literal"><span class="pre">[0</span> <span class="pre">x</span> <span class="pre">i8**]</span> <span class="pre">undef</span></tt>” for a filter which cannot throw. The
+‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ instruction must contain <em>at least</em> one <tt class="docutils literal"><span class="pre">clause</span></tt> or
+the <tt class="docutils literal"><span class="pre">cleanup</span></tt> flag.</p>
+</div>
+<div class="section" id="id296">
+<h5><a class="toc-backref" href="#id1075">Semantics:</a><a class="headerlink" href="#id296" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ instruction defines the values which are set by the
+<a class="reference internal" href="#personalityfn"><em>personality function</em></a> upon re-entry to the function, and
+therefore the “result type” of the <tt class="docutils literal"><span class="pre">landingpad</span></tt> instruction. As with
+calling conventions, how the personality function results are
+represented in LLVM IR is target specific.</p>
+<p>The clauses are applied in order from top to bottom. If two
+<tt class="docutils literal"><span class="pre">landingpad</span></tt> instructions are merged together through inlining, the
+clauses from the calling function are appended to the list of clauses.
+When the call stack is being unwound due to an exception being thrown,
+the exception is compared against each <tt class="docutils literal"><span class="pre">clause</span></tt> in turn. If it doesn’t
+match any of the clauses, and the <tt class="docutils literal"><span class="pre">cleanup</span></tt> flag is not set, then
+unwinding continues further up the call stack.</p>
+<p>The <tt class="docutils literal"><span class="pre">landingpad</span></tt> instruction has several restrictions:</p>
+<ul class="simple">
+<li>A landing pad block is a basic block which is the unwind destination
+of an ‘<tt class="docutils literal"><span class="pre">invoke</span></tt>‘ instruction.</li>
+<li>A landing pad block must have a ‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ instruction as its
+first non-PHI instruction.</li>
+<li>There can be only one ‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ instruction within the landing
+pad block.</li>
+<li>A basic block that is not a landing pad block may not include a
+‘<tt class="docutils literal"><span class="pre">landingpad</span></tt>‘ instruction.</li>
+</ul>
+</div>
+<div class="section" id="id297">
+<h5><a class="toc-backref" href="#id1076">Example:</a><a class="headerlink" href="#id297" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre>;; A landing pad which can catch an integer.
+%res = landingpad { i8*, i32 }
+ catch i8** @_ZTIi
+;; A landing pad that is a cleanup.
+%res = landingpad { i8*, i32 }
+ cleanup
+;; A landing pad which can catch an integer and can only throw a double.
+%res = landingpad { i8*, i32 }
+ catch i8** @_ZTIi
+ filter [1 x i8**] [@_ZTId]
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="intrinsic-functions">
+<span id="intrinsics"></span><h2><a class="toc-backref" href="#id1077">Intrinsic Functions</a><a class="headerlink" href="#intrinsic-functions" title="Permalink to this headline">¶</a></h2>
+<p>LLVM supports the notion of an “intrinsic function”. These functions
+have well known names and semantics and are required to follow certain
+restrictions. Overall, these intrinsics represent an extension mechanism
+for the LLVM language that does not require changing all of the
+transformations in LLVM when adding to the language (or the bitcode
+reader/writer, the parser, etc...).</p>
+<p>Intrinsic function names must all start with an “<tt class="docutils literal"><span class="pre">llvm.</span></tt>” prefix. This
+prefix is reserved in LLVM for intrinsic names; thus, function names may
+not begin with this prefix. Intrinsic functions must always be external
+functions: you cannot define the body of intrinsic functions. Intrinsic
+functions may only be used in call or invoke instructions: it is illegal
+to take the address of an intrinsic function. Additionally, because
+intrinsic functions are part of the LLVM language, it is required if any
+are added that they be documented here.</p>
+<p>Some intrinsic functions can be overloaded, i.e., the intrinsic
+represents a family of functions that perform the same operation but on
+different data types. Because LLVM can represent over 8 million
+different integer types, overloading is used commonly to allow an
+intrinsic function to operate on any integer type. One or more of the
+argument types or the result type can be overloaded to accept any
+integer type. Argument types may also be defined as exactly matching a
+previous argument’s type or the result type. This allows an intrinsic
+function which accepts multiple arguments, but needs all of them to be
+of the same type, to only be overloaded with respect to a single
+argument or the result.</p>
+<p>Overloaded intrinsics will have the names of its overloaded argument
+types encoded into its function name, each preceded by a period. Only
+those types which are overloaded result in a name suffix. Arguments
+whose type is matched against another type do not. For example, the
+<tt class="docutils literal"><span class="pre">llvm.ctpop</span></tt> function can take an integer of any width and returns an
+integer of exactly the same integer width. This leads to a family of
+functions such as <tt class="docutils literal"><span class="pre">i8</span> <span class="pre">@llvm.ctpop.i8(i8</span> <span class="pre">%val)</span></tt> and
+<tt class="docutils literal"><span class="pre">i29</span> <span class="pre">@llvm.ctpop.i29(i29</span> <span class="pre">%val)</span></tt>. Only one type, the return type, is
+overloaded, and only one type suffix is required. Because the argument’s
+type is matched against the return type, it does not require its own
+name suffix.</p>
+<p>To learn how to add an intrinsic function, please see the <a class="reference external" href="ExtendingLLVM.html">Extending
+LLVM Guide</a>.</p>
+<div class="section" id="variable-argument-handling-intrinsics">
+<span id="int-varargs"></span><h3><a class="toc-backref" href="#id1078">Variable Argument Handling Intrinsics</a><a class="headerlink" href="#variable-argument-handling-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>Variable argument support is defined in LLVM with the
+<a class="reference internal" href="#i-va-arg"><em>va_arg</em></a> instruction and these three intrinsic
+functions. These functions are related to the similarly named macros
+defined in the <tt class="docutils literal"><span class="pre"><stdarg.h></span></tt> header file.</p>
+<p>All of these functions operate on arguments that use a target-specific
+value type “<tt class="docutils literal"><span class="pre">va_list</span></tt>”. The LLVM assembly language reference manual
+does not define what this type is, so all transformations should be
+prepared to handle these functions regardless of the type used.</p>
+<p>This example shows how the <a class="reference internal" href="#i-va-arg"><em>va_arg</em></a> instruction and the
+variable argument handling intrinsic functions are used.</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="c">; This struct is different for every platform. For most platforms,</span>
+<span class="c">; it is merely an i8*.</span>
+<span class="nv">%struct.va_list</span> <span class="p">=</span> <span class="k">type</span> <span class="p">{</span> <span class="k">i8</span><span class="p">*</span> <span class="p">}</span>
+
+<span class="c">; For Unix x86_64 platforms, va_list is the following struct:</span>
+<span class="c">; %struct.va_list = type { i32, i32, i8*, i8* }</span>
+
+<span class="k">define</span> <span class="k">i32</span> <span class="vg">@test</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%X</span><span class="p">,</span> <span class="p">...)</span> <span class="p">{</span>
+ <span class="c">; Initialize variable argument processing</span>
+ <span class="nv">%ap</span> <span class="p">=</span> <span class="k">alloca</span> <span class="nv">%struct.va_list</span>
+ <span class="nv">%ap2</span> <span class="p">=</span> <span class="k">bitcast</span> <span class="nv">%struct.va_list</span><span class="p">*</span> <span class="nv">%ap</span> <span class="k">to</span> <span class="k">i8</span><span class="p">*</span>
+ <span class="k">call</span> <span class="kt">void</span> <span class="vg">@llvm.va_start</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%ap2</span><span class="p">)</span>
+
+ <span class="c">; Read a single integer argument</span>
+ <span class="nv">%tmp</span> <span class="p">=</span> <span class="k">va_arg</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%ap2</span><span class="p">,</span> <span class="k">i32</span>
+
+ <span class="c">; Demonstrate usage of llvm.va_copy and llvm.va_end</span>
+ <span class="nv">%aq</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i8</span><span class="p">*</span>
+ <span class="nv">%aq2</span> <span class="p">=</span> <span class="k">bitcast</span> <span class="k">i8</span><span class="p">**</span> <span class="nv">%aq</span> <span class="k">to</span> <span class="k">i8</span><span class="p">*</span>
+ <span class="k">call</span> <span class="kt">void</span> <span class="vg">@llvm.va_copy</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%aq2</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%ap2</span><span class="p">)</span>
+ <span class="k">call</span> <span class="kt">void</span> <span class="vg">@llvm.va_end</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%aq2</span><span class="p">)</span>
+
+ <span class="c">; Stop processing of arguments.</span>
+ <span class="k">call</span> <span class="kt">void</span> <span class="vg">@llvm.va_end</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%ap2</span><span class="p">)</span>
+ <span class="k">ret</span> <span class="k">i32</span> <span class="nv">%tmp</span>
+<span class="p">}</span>
+
+<span class="k">declare</span> <span class="kt">void</span> <span class="vg">@llvm.va_start</span><span class="p">(</span><span class="k">i8</span><span class="p">*)</span>
+<span class="k">declare</span> <span class="kt">void</span> <span class="vg">@llvm.va_copy</span><span class="p">(</span><span class="k">i8</span><span class="p">*,</span> <span class="k">i8</span><span class="p">*)</span>
+<span class="k">declare</span> <span class="kt">void</span> <span class="vg">@llvm.va_end</span><span class="p">(</span><span class="k">i8</span><span class="p">*)</span>
+</pre></div>
+</div>
+<div class="section" id="llvm-va-start-intrinsic">
+<span id="int-va-start"></span><h4><a class="toc-backref" href="#id1079">‘<tt class="docutils literal"><span class="pre">llvm.va_start</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-va-start-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id298">
+<h5><a class="toc-backref" href="#id1080">Syntax:</a><a class="headerlink" href="#id298" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.va_start(i8* <arglist>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id299">
+<h5><a class="toc-backref" href="#id1081">Overview:</a><a class="headerlink" href="#id299" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.va_start</span></tt>‘ intrinsic initializes <tt class="docutils literal"><span class="pre">*<arglist></span></tt> for
+subsequent use by <tt class="docutils literal"><span class="pre">va_arg</span></tt>.</p>
+</div>
+<div class="section" id="id300">
+<h5><a class="toc-backref" href="#id1082">Arguments:</a><a class="headerlink" href="#id300" title="Permalink to this headline">¶</a></h5>
+<p>The argument is a pointer to a <tt class="docutils literal"><span class="pre">va_list</span></tt> element to initialize.</p>
+</div>
+<div class="section" id="id301">
+<h5><a class="toc-backref" href="#id1083">Semantics:</a><a class="headerlink" href="#id301" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.va_start</span></tt>‘ intrinsic works just like the <tt class="docutils literal"><span class="pre">va_start</span></tt> macro
+available in C. In a target-dependent way, it initializes the
+<tt class="docutils literal"><span class="pre">va_list</span></tt> element to which the argument points, so that the next call
+to <tt class="docutils literal"><span class="pre">va_arg</span></tt> will produce the first variable argument passed to the
+function. Unlike the C <tt class="docutils literal"><span class="pre">va_start</span></tt> macro, this intrinsic does not need
+to know the last argument of the function as the compiler can figure
+that out.</p>
+</div>
+</div>
+<div class="section" id="llvm-va-end-intrinsic">
+<h4><a class="toc-backref" href="#id1084">‘<tt class="docutils literal"><span class="pre">llvm.va_end</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-va-end-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id302">
+<h5><a class="toc-backref" href="#id1085">Syntax:</a><a class="headerlink" href="#id302" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.va_end(i8* <arglist>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id303">
+<h5><a class="toc-backref" href="#id1086">Overview:</a><a class="headerlink" href="#id303" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.va_end</span></tt>‘ intrinsic destroys <tt class="docutils literal"><span class="pre">*<arglist></span></tt>, which has been
+initialized previously with <tt class="docutils literal"><span class="pre">llvm.va_start</span></tt> or <tt class="docutils literal"><span class="pre">llvm.va_copy</span></tt>.</p>
+</div>
+<div class="section" id="id304">
+<h5><a class="toc-backref" href="#id1087">Arguments:</a><a class="headerlink" href="#id304" title="Permalink to this headline">¶</a></h5>
+<p>The argument is a pointer to a <tt class="docutils literal"><span class="pre">va_list</span></tt> to destroy.</p>
+</div>
+<div class="section" id="id305">
+<h5><a class="toc-backref" href="#id1088">Semantics:</a><a class="headerlink" href="#id305" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.va_end</span></tt>‘ intrinsic works just like the <tt class="docutils literal"><span class="pre">va_end</span></tt> macro
+available in C. In a target-dependent way, it destroys the <tt class="docutils literal"><span class="pre">va_list</span></tt>
+element to which the argument points. Calls to
+<a class="reference internal" href="#int-va-start"><em>llvm.va_start</em></a> and
+<a class="reference internal" href="#int-va-copy"><em>llvm.va_copy</em></a> must be matched exactly with calls to
+<tt class="docutils literal"><span class="pre">llvm.va_end</span></tt>.</p>
+</div>
+</div>
+<div class="section" id="llvm-va-copy-intrinsic">
+<span id="int-va-copy"></span><h4><a class="toc-backref" href="#id1089">‘<tt class="docutils literal"><span class="pre">llvm.va_copy</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-va-copy-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id306">
+<h5><a class="toc-backref" href="#id1090">Syntax:</a><a class="headerlink" href="#id306" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id307">
+<h5><a class="toc-backref" href="#id1091">Overview:</a><a class="headerlink" href="#id307" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.va_copy</span></tt>‘ intrinsic copies the current argument position
+from the source argument list to the destination argument list.</p>
+</div>
+<div class="section" id="id308">
+<h5><a class="toc-backref" href="#id1092">Arguments:</a><a class="headerlink" href="#id308" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a pointer to a <tt class="docutils literal"><span class="pre">va_list</span></tt> element to initialize.
+The second argument is a pointer to a <tt class="docutils literal"><span class="pre">va_list</span></tt> element to copy from.</p>
+</div>
+<div class="section" id="id309">
+<h5><a class="toc-backref" href="#id1093">Semantics:</a><a class="headerlink" href="#id309" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.va_copy</span></tt>‘ intrinsic works just like the <tt class="docutils literal"><span class="pre">va_copy</span></tt> macro
+available in C. In a target-dependent way, it copies the source
+<tt class="docutils literal"><span class="pre">va_list</span></tt> element into the destination <tt class="docutils literal"><span class="pre">va_list</span></tt> element. This
+intrinsic is necessary because the `` llvm.va_start`` intrinsic may be
+arbitrarily complex and require, for example, memory allocation.</p>
+</div>
+</div>
+</div>
+<div class="section" id="accurate-garbage-collection-intrinsics">
+<h3><a class="toc-backref" href="#id1094">Accurate Garbage Collection Intrinsics</a><a class="headerlink" href="#accurate-garbage-collection-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>LLVM’s support for <a class="reference external" href="GarbageCollection.html">Accurate Garbage Collection</a>
+(GC) requires the frontend to generate code containing appropriate intrinsic
+calls and select an appropriate GC strategy which knows how to lower these
+intrinsics in a manner which is appropriate for the target collector.</p>
+<p>These intrinsics allow identification of <a class="reference internal" href="#int-gcroot"><em>GC roots on the
+stack</em></a>, as well as garbage collector implementations that
+require <a class="reference internal" href="#int-gcread"><em>read</em></a> and <a class="reference internal" href="#int-gcwrite"><em>write</em></a> barriers.
+Frontends for type-safe garbage collected languages should generate
+these intrinsics to make use of the LLVM garbage collectors. For more
+details, see <a class="reference external" href="GarbageCollection.html">Garbage Collection with LLVM</a>.</p>
+<div class="section" id="experimental-statepoint-intrinsics">
+<h4><a class="toc-backref" href="#id1095">Experimental Statepoint Intrinsics</a><a class="headerlink" href="#experimental-statepoint-intrinsics" title="Permalink to this headline">¶</a></h4>
+<p>LLVM provides an second experimental set of intrinsics for describing garbage
+collection safepoints in compiled code. These intrinsics are an alternative
+to the <tt class="docutils literal"><span class="pre">llvm.gcroot</span></tt> intrinsics, but are compatible with the ones for
+<a class="reference internal" href="#int-gcread"><em>read</em></a> and <a class="reference internal" href="#int-gcwrite"><em>write</em></a> barriers. The
+differences in approach are covered in the <a class="reference external" href="GarbageCollection.html">Garbage Collection with LLVM</a> documentation. The intrinsics themselves are
+described in <a class="reference internal" href="Statepoints.html"><em>Garbage Collection Safepoints in LLVM</em></a>.</p>
+</div>
+<div class="section" id="llvm-gcroot-intrinsic">
+<span id="int-gcroot"></span><h4><a class="toc-backref" href="#id1096">‘<tt class="docutils literal"><span class="pre">llvm.gcroot</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-gcroot-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id311">
+<h5><a class="toc-backref" href="#id1097">Syntax:</a><a class="headerlink" href="#id311" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id312">
+<h5><a class="toc-backref" href="#id1098">Overview:</a><a class="headerlink" href="#id312" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.gcroot</span></tt>‘ intrinsic declares the existence of a GC root to
+the code generator, and allows some metadata to be associated with it.</p>
+</div>
+<div class="section" id="id313">
+<h5><a class="toc-backref" href="#id1099">Arguments:</a><a class="headerlink" href="#id313" title="Permalink to this headline">¶</a></h5>
+<p>The first argument specifies the address of a stack object that contains
+the root pointer. The second pointer (which must be either a constant or
+a global value address) contains the meta-data to be associated with the
+root.</p>
+</div>
+<div class="section" id="id314">
+<h5><a class="toc-backref" href="#id1100">Semantics:</a><a class="headerlink" href="#id314" title="Permalink to this headline">¶</a></h5>
+<p>At runtime, a call to this intrinsic stores a null pointer into the
+“ptrloc” location. At compile-time, the code generator generates
+information to allow the runtime to find the pointer at GC safe points.
+The ‘<tt class="docutils literal"><span class="pre">llvm.gcroot</span></tt>‘ intrinsic may only be used in a function which
+<a class="reference internal" href="#gc"><em>specifies a GC algorithm</em></a>.</p>
+</div>
+</div>
+<div class="section" id="llvm-gcread-intrinsic">
+<span id="int-gcread"></span><h4><a class="toc-backref" href="#id1101">‘<tt class="docutils literal"><span class="pre">llvm.gcread</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-gcread-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id315">
+<h5><a class="toc-backref" href="#id1102">Syntax:</a><a class="headerlink" href="#id315" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id316">
+<h5><a class="toc-backref" href="#id1103">Overview:</a><a class="headerlink" href="#id316" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.gcread</span></tt>‘ intrinsic identifies reads of references from heap
+locations, allowing garbage collector implementations that require read
+barriers.</p>
+</div>
+<div class="section" id="id317">
+<h5><a class="toc-backref" href="#id1104">Arguments:</a><a class="headerlink" href="#id317" title="Permalink to this headline">¶</a></h5>
+<p>The second argument is the address to read from, which should be an
+address allocated from the garbage collector. The first object is a
+pointer to the start of the referenced object, if needed by the language
+runtime (otherwise null).</p>
+</div>
+<div class="section" id="id318">
+<h5><a class="toc-backref" href="#id1105">Semantics:</a><a class="headerlink" href="#id318" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.gcread</span></tt>‘ intrinsic has the same semantics as a load
+instruction, but may be replaced with substantially more complex code by
+the garbage collector runtime, as needed. The ‘<tt class="docutils literal"><span class="pre">llvm.gcread</span></tt>‘
+intrinsic may only be used in a function which <a class="reference internal" href="#gc"><em>specifies a GC
+algorithm</em></a>.</p>
+</div>
+</div>
+<div class="section" id="llvm-gcwrite-intrinsic">
+<span id="int-gcwrite"></span><h4><a class="toc-backref" href="#id1106">‘<tt class="docutils literal"><span class="pre">llvm.gcwrite</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-gcwrite-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id319">
+<h5><a class="toc-backref" href="#id1107">Syntax:</a><a class="headerlink" href="#id319" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id320">
+<h5><a class="toc-backref" href="#id1108">Overview:</a><a class="headerlink" href="#id320" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.gcwrite</span></tt>‘ intrinsic identifies writes of references to heap
+locations, allowing garbage collector implementations that require write
+barriers (such as generational or reference counting collectors).</p>
+</div>
+<div class="section" id="id321">
+<h5><a class="toc-backref" href="#id1109">Arguments:</a><a class="headerlink" href="#id321" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is the reference to store, the second is the start of
+the object to store it to, and the third is the address of the field of
+Obj to store to. If the runtime does not require a pointer to the
+object, Obj may be null.</p>
+</div>
+<div class="section" id="id322">
+<h5><a class="toc-backref" href="#id1110">Semantics:</a><a class="headerlink" href="#id322" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.gcwrite</span></tt>‘ intrinsic has the same semantics as a store
+instruction, but may be replaced with substantially more complex code by
+the garbage collector runtime, as needed. The ‘<tt class="docutils literal"><span class="pre">llvm.gcwrite</span></tt>‘
+intrinsic may only be used in a function which <a class="reference internal" href="#gc"><em>specifies a GC
+algorithm</em></a>.</p>
+</div>
+</div>
+</div>
+<div class="section" id="code-generator-intrinsics">
+<h3><a class="toc-backref" href="#id1111">Code Generator Intrinsics</a><a class="headerlink" href="#code-generator-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>These intrinsics are provided by LLVM to expose special features that
+may only be implemented with code generator support.</p>
+<div class="section" id="llvm-returnaddress-intrinsic">
+<h4><a class="toc-backref" href="#id1112">‘<tt class="docutils literal"><span class="pre">llvm.returnaddress</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-returnaddress-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id323">
+<h5><a class="toc-backref" href="#id1113">Syntax:</a><a class="headerlink" href="#id323" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i8 *@llvm.returnaddress(i32 <level>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id324">
+<h5><a class="toc-backref" href="#id1114">Overview:</a><a class="headerlink" href="#id324" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.returnaddress</span></tt>‘ intrinsic attempts to compute a
+target-specific value indicating the return address of the current
+function or one of its callers.</p>
+</div>
+<div class="section" id="id325">
+<h5><a class="toc-backref" href="#id1115">Arguments:</a><a class="headerlink" href="#id325" title="Permalink to this headline">¶</a></h5>
+<p>The argument to this intrinsic indicates which function to return the
+address for. Zero indicates the calling function, one indicates its
+caller, etc. The argument is <strong>required</strong> to be a constant integer
+value.</p>
+</div>
+<div class="section" id="id326">
+<h5><a class="toc-backref" href="#id1116">Semantics:</a><a class="headerlink" href="#id326" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.returnaddress</span></tt>‘ intrinsic either returns a pointer
+indicating the return address of the specified call frame, or zero if it
+cannot be identified. The value returned by this intrinsic is likely to
+be incorrect or 0 for arguments other than zero, so it should only be
+used for debugging purposes.</p>
+<p>Note that calling this intrinsic does not prevent function inlining or
+other aggressive transformations, so the value returned may not be that
+of the obvious source-language caller.</p>
+</div>
+</div>
+<div class="section" id="llvm-frameaddress-intrinsic">
+<h4><a class="toc-backref" href="#id1117">‘<tt class="docutils literal"><span class="pre">llvm.frameaddress</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-frameaddress-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id327">
+<h5><a class="toc-backref" href="#id1118">Syntax:</a><a class="headerlink" href="#id327" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i8* @llvm.frameaddress(i32 <level>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id328">
+<h5><a class="toc-backref" href="#id1119">Overview:</a><a class="headerlink" href="#id328" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.frameaddress</span></tt>‘ intrinsic attempts to return the
+target-specific frame pointer value for the specified stack frame.</p>
+</div>
+<div class="section" id="id329">
+<h5><a class="toc-backref" href="#id1120">Arguments:</a><a class="headerlink" href="#id329" title="Permalink to this headline">¶</a></h5>
+<p>The argument to this intrinsic indicates which function to return the
+frame pointer for. Zero indicates the calling function, one indicates
+its caller, etc. The argument is <strong>required</strong> to be a constant integer
+value.</p>
+</div>
+<div class="section" id="id330">
+<h5><a class="toc-backref" href="#id1121">Semantics:</a><a class="headerlink" href="#id330" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.frameaddress</span></tt>‘ intrinsic either returns a pointer
+indicating the frame address of the specified call frame, or zero if it
+cannot be identified. The value returned by this intrinsic is likely to
+be incorrect or 0 for arguments other than zero, so it should only be
+used for debugging purposes.</p>
+<p>Note that calling this intrinsic does not prevent function inlining or
+other aggressive transformations, so the value returned may not be that
+of the obvious source-language caller.</p>
+</div>
+</div>
+<div class="section" id="llvm-localescape-and-llvm-localrecover-intrinsics">
+<h4><a class="toc-backref" href="#id1122">‘<tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">llvm.localrecover</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-localescape-and-llvm-localrecover-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id331">
+<h5><a class="toc-backref" href="#id1123">Syntax:</a><a class="headerlink" href="#id331" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.localescape(...)
+declare i8* @llvm.localrecover(i8* %func, i8* %fp, i32 %idx)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id332">
+<h5><a class="toc-backref" href="#id1124">Overview:</a><a class="headerlink" href="#id332" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>‘ intrinsic escapes offsets of a collection of static
+allocas, and the ‘<tt class="docutils literal"><span class="pre">llvm.localrecover</span></tt>‘ intrinsic applies those offsets to a
+live frame pointer to recover the address of the allocation. The offset is
+computed during frame layout of the caller of <tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>.</p>
+</div>
+<div class="section" id="id333">
+<h5><a class="toc-backref" href="#id1125">Arguments:</a><a class="headerlink" href="#id333" title="Permalink to this headline">¶</a></h5>
+<p>All arguments to ‘<tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>‘ must be pointers to static allocas or
+casts of static allocas. Each function can only call ‘<tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>‘
+once, and it can only do so from the entry block.</p>
+<p>The <tt class="docutils literal"><span class="pre">func</span></tt> argument to ‘<tt class="docutils literal"><span class="pre">llvm.localrecover</span></tt>‘ must be a constant
+bitcasted pointer to a function defined in the current module. The code
+generator cannot determine the frame allocation offset of functions defined in
+other modules.</p>
+<p>The <tt class="docutils literal"><span class="pre">fp</span></tt> argument to ‘<tt class="docutils literal"><span class="pre">llvm.localrecover</span></tt>‘ must be a frame pointer of a
+call frame that is currently live. The return value of ‘<tt class="docutils literal"><span class="pre">llvm.localaddress</span></tt>‘
+is one way to produce such a value, but various runtimes also expose a suitable
+pointer in platform-specific ways.</p>
+<p>The <tt class="docutils literal"><span class="pre">idx</span></tt> argument to ‘<tt class="docutils literal"><span class="pre">llvm.localrecover</span></tt>‘ indicates which alloca passed to
+‘<tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>‘ to recover. It is zero-indexed.</p>
+</div>
+<div class="section" id="id334">
+<h5><a class="toc-backref" href="#id1126">Semantics:</a><a class="headerlink" href="#id334" title="Permalink to this headline">¶</a></h5>
+<p>These intrinsics allow a group of functions to share access to a set of local
+stack allocations of a one parent function. The parent function may call the
+‘<tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>‘ intrinsic once from the function entry block, and the
+child functions can use ‘<tt class="docutils literal"><span class="pre">llvm.localrecover</span></tt>‘ to access the escaped allocas.
+The ‘<tt class="docutils literal"><span class="pre">llvm.localescape</span></tt>‘ intrinsic blocks inlining, as inlining changes where
+the escaped allocas are allocated, which would break attempts to use
+‘<tt class="docutils literal"><span class="pre">llvm.localrecover</span></tt>‘.</p>
+</div>
+</div>
+<div class="section" id="llvm-read-register-and-llvm-write-register-intrinsics">
+<span id="int-write-register"></span><span id="int-read-register"></span><h4><a class="toc-backref" href="#id1127">‘<tt class="docutils literal"><span class="pre">llvm.read_register</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">llvm.write_register</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-read-register-and-llvm-write-register-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id335">
+<h5><a class="toc-backref" href="#id1128">Syntax:</a><a class="headerlink" href="#id335" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i32 @llvm.read_register.i32(metadata)
+declare i64 @llvm.read_register.i64(metadata)
+declare void @llvm.write_register.i32(metadata, i32 @value)
+declare void @llvm.write_register.i64(metadata, i64 @value)
+!0 = !{!"sp\00"}
+</pre></div>
+</div>
+</div>
+<div class="section" id="id336">
+<h5><a class="toc-backref" href="#id1129">Overview:</a><a class="headerlink" href="#id336" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.read_register</span></tt>‘ and ‘<tt class="docutils literal"><span class="pre">llvm.write_register</span></tt>‘ intrinsics
+provides access to the named register. The register must be valid on
+the architecture being compiled to. The type needs to be compatible
+with the register being read.</p>
+</div>
+<div class="section" id="id337">
+<h5><a class="toc-backref" href="#id1130">Semantics:</a><a class="headerlink" href="#id337" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.read_register</span></tt>‘ intrinsic returns the current value of the
+register, where possible. The ‘<tt class="docutils literal"><span class="pre">llvm.write_register</span></tt>‘ intrinsic sets
+the current value of the register, where possible.</p>
+<p>This is useful to implement named register global variables that need
+to always be mapped to a specific register, as is common practice on
+bare-metal programs including OS kernels.</p>
+<p>The compiler doesn’t check for register availability or use of the used
+register in surrounding code, including inline assembly. Because of that,
+allocatable registers are not supported.</p>
+<p>Warning: So far it only works with the stack pointer on selected
+architectures (ARM, AArch64, PowerPC and x86_64). Significant amount of
+work is needed to support other registers and even more so, allocatable
+registers.</p>
+</div>
+</div>
+<div class="section" id="llvm-stacksave-intrinsic">
+<span id="int-stacksave"></span><h4><a class="toc-backref" href="#id1131">‘<tt class="docutils literal"><span class="pre">llvm.stacksave</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-stacksave-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id338">
+<h5><a class="toc-backref" href="#id1132">Syntax:</a><a class="headerlink" href="#id338" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i8* @llvm.stacksave()
+</pre></div>
+</div>
+</div>
+<div class="section" id="id339">
+<h5><a class="toc-backref" href="#id1133">Overview:</a><a class="headerlink" href="#id339" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.stacksave</span></tt>‘ intrinsic is used to remember the current state
+of the function stack, for use with
+<a class="reference internal" href="#int-stackrestore"><em>llvm.stackrestore</em></a>. This is useful for
+implementing language features like scoped automatic variable sized
+arrays in C99.</p>
+</div>
+<div class="section" id="id340">
+<h5><a class="toc-backref" href="#id1134">Semantics:</a><a class="headerlink" href="#id340" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic returns a opaque pointer value that can be passed to
+<a class="reference internal" href="#int-stackrestore"><em>llvm.stackrestore</em></a>. When an
+<tt class="docutils literal"><span class="pre">llvm.stackrestore</span></tt> intrinsic is executed with a value saved from
+<tt class="docutils literal"><span class="pre">llvm.stacksave</span></tt>, it effectively restores the state of the stack to
+the state it was in when the <tt class="docutils literal"><span class="pre">llvm.stacksave</span></tt> intrinsic executed. In
+practice, this pops any <a class="reference internal" href="#i-alloca"><em>alloca</em></a> blocks from the stack that
+were allocated after the <tt class="docutils literal"><span class="pre">llvm.stacksave</span></tt> was executed.</p>
+</div>
+</div>
+<div class="section" id="llvm-stackrestore-intrinsic">
+<span id="int-stackrestore"></span><h4><a class="toc-backref" href="#id1135">‘<tt class="docutils literal"><span class="pre">llvm.stackrestore</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-stackrestore-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id341">
+<h5><a class="toc-backref" href="#id1136">Syntax:</a><a class="headerlink" href="#id341" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.stackrestore(i8* %ptr)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id342">
+<h5><a class="toc-backref" href="#id1137">Overview:</a><a class="headerlink" href="#id342" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.stackrestore</span></tt>‘ intrinsic is used to restore the state of
+the function stack to the state it was in when the corresponding
+<a class="reference internal" href="#int-stacksave"><em>llvm.stacksave</em></a> intrinsic executed. This is
+useful for implementing language features like scoped automatic variable
+sized arrays in C99.</p>
+</div>
+<div class="section" id="id343">
+<h5><a class="toc-backref" href="#id1138">Semantics:</a><a class="headerlink" href="#id343" title="Permalink to this headline">¶</a></h5>
+<p>See the description for <a class="reference internal" href="#int-stacksave"><em>llvm.stacksave</em></a>.</p>
+</div>
+</div>
+<div class="section" id="llvm-prefetch-intrinsic">
+<h4><a class="toc-backref" href="#id1139">‘<tt class="docutils literal"><span class="pre">llvm.prefetch</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-prefetch-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id344">
+<h5><a class="toc-backref" href="#id1140">Syntax:</a><a class="headerlink" href="#id344" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id345">
+<h5><a class="toc-backref" href="#id1141">Overview:</a><a class="headerlink" href="#id345" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.prefetch</span></tt>‘ intrinsic is a hint to the code generator to
+insert a prefetch instruction if supported; otherwise, it is a noop.
+Prefetches have no effect on the behavior of the program but can change
+its performance characteristics.</p>
+</div>
+<div class="section" id="id346">
+<h5><a class="toc-backref" href="#id1142">Arguments:</a><a class="headerlink" href="#id346" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">address</span></tt> is the address to be prefetched, <tt class="docutils literal"><span class="pre">rw</span></tt> is the specifier
+determining if the fetch should be for a read (0) or write (1), and
+<tt class="docutils literal"><span class="pre">locality</span></tt> is a temporal locality specifier ranging from (0) - no
+locality, to (3) - extremely local keep in cache. The <tt class="docutils literal"><span class="pre">cache</span> <span class="pre">type</span></tt>
+specifies whether the prefetch is performed on the data (1) or
+instruction (0) cache. The <tt class="docutils literal"><span class="pre">rw</span></tt>, <tt class="docutils literal"><span class="pre">locality</span></tt> and <tt class="docutils literal"><span class="pre">cache</span> <span class="pre">type</span></tt>
+arguments must be constant integers.</p>
+</div>
+<div class="section" id="id347">
+<h5><a class="toc-backref" href="#id1143">Semantics:</a><a class="headerlink" href="#id347" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic does not modify the behavior of the program. In
+particular, prefetches cannot trap and do not produce a value. On
+targets that support this intrinsic, the prefetch can provide hints to
+the processor cache for better performance.</p>
+</div>
+</div>
+<div class="section" id="llvm-pcmarker-intrinsic">
+<h4><a class="toc-backref" href="#id1144">‘<tt class="docutils literal"><span class="pre">llvm.pcmarker</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-pcmarker-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id348">
+<h5><a class="toc-backref" href="#id1145">Syntax:</a><a class="headerlink" href="#id348" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.pcmarker(i32 <id>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id349">
+<h5><a class="toc-backref" href="#id1146">Overview:</a><a class="headerlink" href="#id349" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.pcmarker</span></tt>‘ intrinsic is a method to export a Program
+Counter (PC) in a region of code to simulators and other tools. The
+method is target specific, but it is expected that the marker will use
+exported symbols to transmit the PC of the marker. The marker makes no
+guarantees that it will remain with any specific instruction after
+optimizations. It is possible that the presence of a marker will inhibit
+optimizations. The intended use is to be inserted after optimizations to
+allow correlations of simulation runs.</p>
+</div>
+<div class="section" id="id350">
+<h5><a class="toc-backref" href="#id1147">Arguments:</a><a class="headerlink" href="#id350" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">id</span></tt> is a numerical id identifying the marker.</p>
+</div>
+<div class="section" id="id351">
+<h5><a class="toc-backref" href="#id1148">Semantics:</a><a class="headerlink" href="#id351" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic does not modify the behavior of the program. Backends
+that do not support this intrinsic may ignore it.</p>
+</div>
+</div>
+<div class="section" id="llvm-readcyclecounter-intrinsic">
+<h4><a class="toc-backref" href="#id1149">‘<tt class="docutils literal"><span class="pre">llvm.readcyclecounter</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-readcyclecounter-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id352">
+<h5><a class="toc-backref" href="#id1150">Syntax:</a><a class="headerlink" href="#id352" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i64 @llvm.readcyclecounter()
+</pre></div>
+</div>
+</div>
+<div class="section" id="id353">
+<h5><a class="toc-backref" href="#id1151">Overview:</a><a class="headerlink" href="#id353" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.readcyclecounter</span></tt>‘ intrinsic provides access to the cycle
+counter register (or similar low latency, high accuracy clocks) on those
+targets that support it. On X86, it should map to RDTSC. On Alpha, it
+should map to RPCC. As the backing counters overflow quickly (on the
+order of 9 seconds on alpha), this should only be used for small
+timings.</p>
+</div>
+<div class="section" id="id354">
+<h5><a class="toc-backref" href="#id1152">Semantics:</a><a class="headerlink" href="#id354" title="Permalink to this headline">¶</a></h5>
+<p>When directly supported, reading the cycle counter should not modify any
+memory. Implementations are allowed to either return a application
+specific value or a system wide value. On backends without support, this
+is lowered to a constant 0.</p>
+<p>Note that runtime support may be conditional on the privilege-level code is
+running at and the host platform.</p>
+</div>
+</div>
+<div class="section" id="llvm-clear-cache-intrinsic">
+<h4><a class="toc-backref" href="#id1153">‘<tt class="docutils literal"><span class="pre">llvm.clear_cache</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-clear-cache-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id355">
+<h5><a class="toc-backref" href="#id1154">Syntax:</a><a class="headerlink" href="#id355" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.clear_cache(i8*, i8*)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id356">
+<h5><a class="toc-backref" href="#id1155">Overview:</a><a class="headerlink" href="#id356" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.clear_cache</span></tt>‘ intrinsic ensures visibility of modifications
+in the specified range to the execution unit of the processor. On
+targets with non-unified instruction and data cache, the implementation
+flushes the instruction cache.</p>
+</div>
+<div class="section" id="id357">
+<h5><a class="toc-backref" href="#id1156">Semantics:</a><a class="headerlink" href="#id357" title="Permalink to this headline">¶</a></h5>
+<p>On platforms with coherent instruction and data caches (e.g. x86), this
+intrinsic is a nop. On platforms with non-coherent instruction and data
+cache (e.g. ARM, MIPS), the intrinsic is lowered either to appropriate
+instructions or a system call, if cache flushing requires special
+privileges.</p>
+<p>The default behavior is to emit a call to <tt class="docutils literal"><span class="pre">__clear_cache</span></tt> from the run
+time library.</p>
+<p>This instrinsic does <em>not</em> empty the instruction pipeline. Modifications
+of the current function are outside the scope of the intrinsic.</p>
+</div>
+</div>
+<div class="section" id="llvm-instrprof-increment-intrinsic">
+<h4><a class="toc-backref" href="#id1157">‘<tt class="docutils literal"><span class="pre">llvm.instrprof_increment</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-instrprof-increment-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id358">
+<h5><a class="toc-backref" href="#id1158">Syntax:</a><a class="headerlink" href="#id358" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.instrprof_increment(i8* <name>, i64 <hash>,
+ i32 <num-counters>, i32 <index>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id359">
+<h5><a class="toc-backref" href="#id1159">Overview:</a><a class="headerlink" href="#id359" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.instrprof_increment</span></tt>‘ intrinsic can be emitted by a
+frontend for use with instrumentation based profiling. These will be
+lowered by the <tt class="docutils literal"><span class="pre">-instrprof</span></tt> pass to generate execution counts of a
+program at runtime.</p>
+</div>
+<div class="section" id="id360">
+<h5><a class="toc-backref" href="#id1160">Arguments:</a><a class="headerlink" href="#id360" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a pointer to a global variable containing the
+name of the entity being instrumented. This should generally be the
+(mangled) function name for a set of counters.</p>
+<p>The second argument is a hash value that can be used by the consumer
+of the profile data to detect changes to the instrumented source, and
+the third is the number of counters associated with <tt class="docutils literal"><span class="pre">name</span></tt>. It is an
+error if <tt class="docutils literal"><span class="pre">hash</span></tt> or <tt class="docutils literal"><span class="pre">num-counters</span></tt> differ between two instances of
+<tt class="docutils literal"><span class="pre">instrprof_increment</span></tt> that refer to the same name.</p>
+<p>The last argument refers to which of the counters for <tt class="docutils literal"><span class="pre">name</span></tt> should
+be incremented. It should be a value between 0 and <tt class="docutils literal"><span class="pre">num-counters</span></tt>.</p>
+</div>
+<div class="section" id="id361">
+<h5><a class="toc-backref" href="#id1161">Semantics:</a><a class="headerlink" href="#id361" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic represents an increment of a profiling counter. It will
+cause the <tt class="docutils literal"><span class="pre">-instrprof</span></tt> pass to generate the appropriate data
+structures and the code to increment the appropriate value, in a
+format that can be written out by a compiler runtime and consumed via
+the <tt class="docutils literal"><span class="pre">llvm-profdata</span></tt> tool.</p>
+</div>
+</div>
+</div>
+<div class="section" id="standard-c-library-intrinsics">
+<h3><a class="toc-backref" href="#id1162">Standard C Library Intrinsics</a><a class="headerlink" href="#standard-c-library-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>LLVM provides intrinsics for a few important standard C library
+functions. These intrinsics allow source-language front-ends to pass
+information about the alignment of the pointer arguments to the code
+generator, providing opportunity for more efficient code generation.</p>
+<div class="section" id="llvm-memcpy-intrinsic">
+<span id="int-memcpy"></span><h4><a class="toc-backref" href="#id1163">‘<tt class="docutils literal"><span class="pre">llvm.memcpy</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-memcpy-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id362">
+<h5><a class="toc-backref" href="#id1164">Syntax:</a><a class="headerlink" href="#id362" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.memcpy</span></tt> on any
+integer bit width and for different address spaces. Not all targets
+support all bit widths however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
+ i32 <len>, i32 <align>, i1 <isvolatile>)
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
+ i64 <len>, i32 <align>, i1 <isvolatile>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id363">
+<h5><a class="toc-backref" href="#id1165">Overview:</a><a class="headerlink" href="#id363" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.memcpy.*</span></tt>‘ intrinsics copy a block of memory from the
+source location to the destination location.</p>
+<p>Note that, unlike the standard libc function, the <tt class="docutils literal"><span class="pre">llvm.memcpy.*</span></tt>
+intrinsics do not return a value, takes extra alignment/isvolatile
+arguments and the pointers can be in specified address spaces.</p>
+</div>
+<div class="section" id="id364">
+<h5><a class="toc-backref" href="#id1166">Arguments:</a><a class="headerlink" href="#id364" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a pointer to the destination, the second is a
+pointer to the source. The third argument is an integer argument
+specifying the number of bytes to copy, the fourth argument is the
+alignment of the source and destination locations, and the fifth is a
+boolean indicating a volatile access.</p>
+<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
+then the caller guarantees that both the source and destination pointers
+are aligned to that boundary.</p>
+<p>If the <tt class="docutils literal"><span class="pre">isvolatile</span></tt> parameter is <tt class="docutils literal"><span class="pre">true</span></tt>, the <tt class="docutils literal"><span class="pre">llvm.memcpy</span></tt> call is
+a <a class="reference internal" href="#volatile"><em>volatile operation</em></a>. The detailed access behavior is not
+very cleanly specified and it is unwise to depend on it.</p>
+</div>
+<div class="section" id="id365">
+<h5><a class="toc-backref" href="#id1167">Semantics:</a><a class="headerlink" href="#id365" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.memcpy.*</span></tt>‘ intrinsics copy a block of memory from the
+source location to the destination location, which are not allowed to
+overlap. It copies “len” bytes of memory over. If the argument is known
+to be aligned to some boundary, this can be specified as the fourth
+argument, otherwise it should be set to 0 or 1 (both meaning no alignment).</p>
+</div>
+</div>
+<div class="section" id="llvm-memmove-intrinsic">
+<h4><a class="toc-backref" href="#id1168">‘<tt class="docutils literal"><span class="pre">llvm.memmove</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-memmove-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id366">
+<h5><a class="toc-backref" href="#id1169">Syntax:</a><a class="headerlink" href="#id366" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer
+bit width and for different address space. Not all targets support all
+bit widths however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
+ i32 <len>, i32 <align>, i1 <isvolatile>)
+declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
+ i64 <len>, i32 <align>, i1 <isvolatile>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id367">
+<h5><a class="toc-backref" href="#id1170">Overview:</a><a class="headerlink" href="#id367" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.memmove.*</span></tt>‘ intrinsics move a block of memory from the
+source location to the destination location. It is similar to the
+‘<tt class="docutils literal"><span class="pre">llvm.memcpy</span></tt>‘ intrinsic but allows the two memory locations to
+overlap.</p>
+<p>Note that, unlike the standard libc function, the <tt class="docutils literal"><span class="pre">llvm.memmove.*</span></tt>
+intrinsics do not return a value, takes extra alignment/isvolatile
+arguments and the pointers can be in specified address spaces.</p>
+</div>
+<div class="section" id="id368">
+<h5><a class="toc-backref" href="#id1171">Arguments:</a><a class="headerlink" href="#id368" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a pointer to the destination, the second is a
+pointer to the source. The third argument is an integer argument
+specifying the number of bytes to copy, the fourth argument is the
+alignment of the source and destination locations, and the fifth is a
+boolean indicating a volatile access.</p>
+<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
+then the caller guarantees that the source and destination pointers are
+aligned to that boundary.</p>
+<p>If the <tt class="docutils literal"><span class="pre">isvolatile</span></tt> parameter is <tt class="docutils literal"><span class="pre">true</span></tt>, the <tt class="docutils literal"><span class="pre">llvm.memmove</span></tt> call
+is a <a class="reference internal" href="#volatile"><em>volatile operation</em></a>. The detailed access behavior is
+not very cleanly specified and it is unwise to depend on it.</p>
+</div>
+<div class="section" id="id369">
+<h5><a class="toc-backref" href="#id1172">Semantics:</a><a class="headerlink" href="#id369" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.memmove.*</span></tt>‘ intrinsics copy a block of memory from the
+source location to the destination location, which may overlap. It
+copies “len” bytes of memory over. If the argument is known to be
+aligned to some boundary, this can be specified as the fourth argument,
+otherwise it should be set to 0 or 1 (both meaning no alignment).</p>
+</div>
+</div>
+<div class="section" id="llvm-memset-intrinsics">
+<h4><a class="toc-backref" href="#id1173">‘<tt class="docutils literal"><span class="pre">llvm.memset.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-memset-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id370">
+<h5><a class="toc-backref" href="#id1174">Syntax:</a><a class="headerlink" href="#id370" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use llvm.memset on any integer
+bit width and for different address spaces. However, not all targets
+support all bit widths.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
+ i32 <len>, i32 <align>, i1 <isvolatile>)
+declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
+ i64 <len>, i32 <align>, i1 <isvolatile>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id371">
+<h5><a class="toc-backref" href="#id1175">Overview:</a><a class="headerlink" href="#id371" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.memset.*</span></tt>‘ intrinsics fill a block of memory with a
+particular byte value.</p>
+<p>Note that, unlike the standard libc function, the <tt class="docutils literal"><span class="pre">llvm.memset</span></tt>
+intrinsic does not return a value and takes extra alignment/volatile
+arguments. Also, the destination can be in an arbitrary address space.</p>
+</div>
+<div class="section" id="id372">
+<h5><a class="toc-backref" href="#id1176">Arguments:</a><a class="headerlink" href="#id372" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a pointer to the destination to fill, the second
+is the byte value with which to fill it, the third argument is an
+integer argument specifying the number of bytes to fill, and the fourth
+argument is the known alignment of the destination location.</p>
+<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
+then the caller guarantees that the destination pointer is aligned to
+that boundary.</p>
+<p>If the <tt class="docutils literal"><span class="pre">isvolatile</span></tt> parameter is <tt class="docutils literal"><span class="pre">true</span></tt>, the <tt class="docutils literal"><span class="pre">llvm.memset</span></tt> call is
+a <a class="reference internal" href="#volatile"><em>volatile operation</em></a>. The detailed access behavior is not
+very cleanly specified and it is unwise to depend on it.</p>
+</div>
+<div class="section" id="id373">
+<h5><a class="toc-backref" href="#id1177">Semantics:</a><a class="headerlink" href="#id373" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.memset.*</span></tt>‘ intrinsics fill “len” bytes of memory starting
+at the destination location. If the argument is known to be aligned to
+some boundary, this can be specified as the fourth argument, otherwise
+it should be set to 0 or 1 (both meaning no alignment).</p>
+</div>
+</div>
+<div class="section" id="llvm-sqrt-intrinsic">
+<h4><a class="toc-backref" href="#id1178">‘<tt class="docutils literal"><span class="pre">llvm.sqrt.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-sqrt-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id374">
+<h5><a class="toc-backref" href="#id1179">Syntax:</a><a class="headerlink" href="#id374" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.sqrt</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.sqrt.f32(float %Val)
+declare double @llvm.sqrt.f64(double %Val)
+declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
+declare fp128 @llvm.sqrt.f128(fp128 %Val)
+declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id375">
+<h5><a class="toc-backref" href="#id1180">Overview:</a><a class="headerlink" href="#id375" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.sqrt</span></tt>‘ intrinsics return the sqrt of the specified operand,
+returning the same value as the libm ‘<tt class="docutils literal"><span class="pre">sqrt</span></tt>‘ functions would. Unlike
+<tt class="docutils literal"><span class="pre">sqrt</span></tt> in libm, however, <tt class="docutils literal"><span class="pre">llvm.sqrt</span></tt> has undefined behavior for
+negative numbers other than -0.0 (which allows for better optimization,
+because there is no need to worry about errno being set).
+<tt class="docutils literal"><span class="pre">llvm.sqrt(-0.0)</span></tt> is defined to return -0.0 like IEEE sqrt.</p>
+</div>
+<div class="section" id="id376">
+<h5><a class="toc-backref" href="#id1181">Arguments:</a><a class="headerlink" href="#id376" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id377">
+<h5><a class="toc-backref" href="#id1182">Semantics:</a><a class="headerlink" href="#id377" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the sqrt of the specified operand if it is a
+nonnegative floating point number.</p>
+</div>
+</div>
+<div class="section" id="llvm-powi-intrinsic">
+<h4><a class="toc-backref" href="#id1183">‘<tt class="docutils literal"><span class="pre">llvm.powi.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-powi-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id378">
+<h5><a class="toc-backref" href="#id1184">Syntax:</a><a class="headerlink" href="#id378" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.powi</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.powi.f32(float %Val, i32 %power)
+declare double @llvm.powi.f64(double %Val, i32 %power)
+declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
+declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
+declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id379">
+<h5><a class="toc-backref" href="#id1185">Overview:</a><a class="headerlink" href="#id379" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.powi.*</span></tt>‘ intrinsics return the first operand raised to the
+specified (positive or negative) power. The order of evaluation of
+multiplications is not defined. When a vector of floating point type is
+used, the second argument remains a scalar integer value.</p>
+</div>
+<div class="section" id="id380">
+<h5><a class="toc-backref" href="#id1186">Arguments:</a><a class="headerlink" href="#id380" title="Permalink to this headline">¶</a></h5>
+<p>The second argument is an integer power, and the first is a value to
+raise to that power.</p>
+</div>
+<div class="section" id="id381">
+<h5><a class="toc-backref" href="#id1187">Semantics:</a><a class="headerlink" href="#id381" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the first value raised to the second power with an
+unspecified sequence of rounding operations.</p>
+</div>
+</div>
+<div class="section" id="llvm-sin-intrinsic">
+<h4><a class="toc-backref" href="#id1188">‘<tt class="docutils literal"><span class="pre">llvm.sin.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-sin-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id382">
+<h5><a class="toc-backref" href="#id1189">Syntax:</a><a class="headerlink" href="#id382" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.sin</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.sin.f32(float %Val)
+declare double @llvm.sin.f64(double %Val)
+declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
+declare fp128 @llvm.sin.f128(fp128 %Val)
+declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id383">
+<h5><a class="toc-backref" href="#id1190">Overview:</a><a class="headerlink" href="#id383" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.sin.*</span></tt>‘ intrinsics return the sine of the operand.</p>
+</div>
+<div class="section" id="id384">
+<h5><a class="toc-backref" href="#id1191">Arguments:</a><a class="headerlink" href="#id384" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id385">
+<h5><a class="toc-backref" href="#id1192">Semantics:</a><a class="headerlink" href="#id385" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the sine of the specified operand, returning the
+same values as the libm <tt class="docutils literal"><span class="pre">sin</span></tt> functions would, and handles error
+conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-cos-intrinsic">
+<h4><a class="toc-backref" href="#id1193">‘<tt class="docutils literal"><span class="pre">llvm.cos.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-cos-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id386">
+<h5><a class="toc-backref" href="#id1194">Syntax:</a><a class="headerlink" href="#id386" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.cos</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.cos.f32(float %Val)
+declare double @llvm.cos.f64(double %Val)
+declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
+declare fp128 @llvm.cos.f128(fp128 %Val)
+declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id387">
+<h5><a class="toc-backref" href="#id1195">Overview:</a><a class="headerlink" href="#id387" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.cos.*</span></tt>‘ intrinsics return the cosine of the operand.</p>
+</div>
+<div class="section" id="id388">
+<h5><a class="toc-backref" href="#id1196">Arguments:</a><a class="headerlink" href="#id388" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id389">
+<h5><a class="toc-backref" href="#id1197">Semantics:</a><a class="headerlink" href="#id389" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the cosine of the specified operand, returning the
+same values as the libm <tt class="docutils literal"><span class="pre">cos</span></tt> functions would, and handles error
+conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-pow-intrinsic">
+<h4><a class="toc-backref" href="#id1198">‘<tt class="docutils literal"><span class="pre">llvm.pow.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-pow-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id390">
+<h5><a class="toc-backref" href="#id1199">Syntax:</a><a class="headerlink" href="#id390" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.pow</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.pow.f32(float %Val, float %Power)
+declare double @llvm.pow.f64(double %Val, double %Power)
+declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
+declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
+declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id391">
+<h5><a class="toc-backref" href="#id1200">Overview:</a><a class="headerlink" href="#id391" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.pow.*</span></tt>‘ intrinsics return the first operand raised to the
+specified (positive or negative) power.</p>
+</div>
+<div class="section" id="id392">
+<h5><a class="toc-backref" href="#id1201">Arguments:</a><a class="headerlink" href="#id392" title="Permalink to this headline">¶</a></h5>
+<p>The second argument is a floating point power, and the first is a value
+to raise to that power.</p>
+</div>
+<div class="section" id="id393">
+<h5><a class="toc-backref" href="#id1202">Semantics:</a><a class="headerlink" href="#id393" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the first value raised to the second power,
+returning the same values as the libm <tt class="docutils literal"><span class="pre">pow</span></tt> functions would, and
+handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-exp-intrinsic">
+<h4><a class="toc-backref" href="#id1203">‘<tt class="docutils literal"><span class="pre">llvm.exp.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-exp-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id394">
+<h5><a class="toc-backref" href="#id1204">Syntax:</a><a class="headerlink" href="#id394" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.exp</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.exp.f32(float %Val)
+declare double @llvm.exp.f64(double %Val)
+declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
+declare fp128 @llvm.exp.f128(fp128 %Val)
+declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id395">
+<h5><a class="toc-backref" href="#id1205">Overview:</a><a class="headerlink" href="#id395" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.exp.*</span></tt>‘ intrinsics perform the exp function.</p>
+</div>
+<div class="section" id="id396">
+<h5><a class="toc-backref" href="#id1206">Arguments:</a><a class="headerlink" href="#id396" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id397">
+<h5><a class="toc-backref" href="#id1207">Semantics:</a><a class="headerlink" href="#id397" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">exp</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-exp2-intrinsic">
+<h4><a class="toc-backref" href="#id1208">‘<tt class="docutils literal"><span class="pre">llvm.exp2.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-exp2-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id398">
+<h5><a class="toc-backref" href="#id1209">Syntax:</a><a class="headerlink" href="#id398" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.exp2</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.exp2.f32(float %Val)
+declare double @llvm.exp2.f64(double %Val)
+declare x86_fp80 @llvm.exp2.f80(x86_fp80 %Val)
+declare fp128 @llvm.exp2.f128(fp128 %Val)
+declare ppc_fp128 @llvm.exp2.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id399">
+<h5><a class="toc-backref" href="#id1210">Overview:</a><a class="headerlink" href="#id399" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.exp2.*</span></tt>‘ intrinsics perform the exp2 function.</p>
+</div>
+<div class="section" id="id400">
+<h5><a class="toc-backref" href="#id1211">Arguments:</a><a class="headerlink" href="#id400" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id401">
+<h5><a class="toc-backref" href="#id1212">Semantics:</a><a class="headerlink" href="#id401" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">exp2</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-log-intrinsic">
+<h4><a class="toc-backref" href="#id1213">‘<tt class="docutils literal"><span class="pre">llvm.log.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-log-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id402">
+<h5><a class="toc-backref" href="#id1214">Syntax:</a><a class="headerlink" href="#id402" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.log</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.log.f32(float %Val)
+declare double @llvm.log.f64(double %Val)
+declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
+declare fp128 @llvm.log.f128(fp128 %Val)
+declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id403">
+<h5><a class="toc-backref" href="#id1215">Overview:</a><a class="headerlink" href="#id403" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.log.*</span></tt>‘ intrinsics perform the log function.</p>
+</div>
+<div class="section" id="id404">
+<h5><a class="toc-backref" href="#id1216">Arguments:</a><a class="headerlink" href="#id404" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id405">
+<h5><a class="toc-backref" href="#id1217">Semantics:</a><a class="headerlink" href="#id405" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">log</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-log10-intrinsic">
+<h4><a class="toc-backref" href="#id1218">‘<tt class="docutils literal"><span class="pre">llvm.log10.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-log10-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id406">
+<h5><a class="toc-backref" href="#id1219">Syntax:</a><a class="headerlink" href="#id406" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.log10</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.log10.f32(float %Val)
+declare double @llvm.log10.f64(double %Val)
+declare x86_fp80 @llvm.log10.f80(x86_fp80 %Val)
+declare fp128 @llvm.log10.f128(fp128 %Val)
+declare ppc_fp128 @llvm.log10.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id407">
+<h5><a class="toc-backref" href="#id1220">Overview:</a><a class="headerlink" href="#id407" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.log10.*</span></tt>‘ intrinsics perform the log10 function.</p>
+</div>
+<div class="section" id="id408">
+<h5><a class="toc-backref" href="#id1221">Arguments:</a><a class="headerlink" href="#id408" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id409">
+<h5><a class="toc-backref" href="#id1222">Semantics:</a><a class="headerlink" href="#id409" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">log10</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-log2-intrinsic">
+<h4><a class="toc-backref" href="#id1223">‘<tt class="docutils literal"><span class="pre">llvm.log2.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-log2-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id410">
+<h5><a class="toc-backref" href="#id1224">Syntax:</a><a class="headerlink" href="#id410" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.log2</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.log2.f32(float %Val)
+declare double @llvm.log2.f64(double %Val)
+declare x86_fp80 @llvm.log2.f80(x86_fp80 %Val)
+declare fp128 @llvm.log2.f128(fp128 %Val)
+declare ppc_fp128 @llvm.log2.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id411">
+<h5><a class="toc-backref" href="#id1225">Overview:</a><a class="headerlink" href="#id411" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.log2.*</span></tt>‘ intrinsics perform the log2 function.</p>
+</div>
+<div class="section" id="id412">
+<h5><a class="toc-backref" href="#id1226">Arguments:</a><a class="headerlink" href="#id412" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id413">
+<h5><a class="toc-backref" href="#id1227">Semantics:</a><a class="headerlink" href="#id413" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">log2</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-fma-intrinsic">
+<h4><a class="toc-backref" href="#id1228">‘<tt class="docutils literal"><span class="pre">llvm.fma.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-fma-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id414">
+<h5><a class="toc-backref" href="#id1229">Syntax:</a><a class="headerlink" href="#id414" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.fma</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.fma.f32(float %a, float %b, float %c)
+declare double @llvm.fma.f64(double %a, double %b, double %c)
+declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
+declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
+declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id415">
+<h5><a class="toc-backref" href="#id1230">Overview:</a><a class="headerlink" href="#id415" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.fma.*</span></tt>‘ intrinsics perform the fused multiply-add
+operation.</p>
+</div>
+<div class="section" id="id416">
+<h5><a class="toc-backref" href="#id1231">Arguments:</a><a class="headerlink" href="#id416" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id417">
+<h5><a class="toc-backref" href="#id1232">Semantics:</a><a class="headerlink" href="#id417" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">fma</span></tt> functions
+would, and does not set errno.</p>
+</div>
+</div>
+<div class="section" id="llvm-fabs-intrinsic">
+<h4><a class="toc-backref" href="#id1233">‘<tt class="docutils literal"><span class="pre">llvm.fabs.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-fabs-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id418">
+<h5><a class="toc-backref" href="#id1234">Syntax:</a><a class="headerlink" href="#id418" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.fabs</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.fabs.f32(float %Val)
+declare double @llvm.fabs.f64(double %Val)
+declare x86_fp80 @llvm.fabs.f80(x86_fp80 %Val)
+declare fp128 @llvm.fabs.f128(fp128 %Val)
+declare ppc_fp128 @llvm.fabs.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id419">
+<h5><a class="toc-backref" href="#id1235">Overview:</a><a class="headerlink" href="#id419" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.fabs.*</span></tt>‘ intrinsics return the absolute value of the
+operand.</p>
+</div>
+<div class="section" id="id420">
+<h5><a class="toc-backref" href="#id1236">Arguments:</a><a class="headerlink" href="#id420" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id421">
+<h5><a class="toc-backref" href="#id1237">Semantics:</a><a class="headerlink" href="#id421" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">fabs</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-minnum-intrinsic">
+<h4><a class="toc-backref" href="#id1238">‘<tt class="docutils literal"><span class="pre">llvm.minnum.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-minnum-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id422">
+<h5><a class="toc-backref" href="#id1239">Syntax:</a><a class="headerlink" href="#id422" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.minnum</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.minnum.f32(float %Val0, float %Val1)
+declare double @llvm.minnum.f64(double %Val0, double %Val1)
+declare x86_fp80 @llvm.minnum.f80(x86_fp80 %Val0, x86_fp80 %Val1)
+declare fp128 @llvm.minnum.f128(fp128 %Val0, fp128 %Val1)
+declare ppc_fp128 @llvm.minnum.ppcf128(ppc_fp128 %Val0, ppc_fp128 %Val1)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id423">
+<h5><a class="toc-backref" href="#id1240">Overview:</a><a class="headerlink" href="#id423" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.minnum.*</span></tt>‘ intrinsics return the minimum of the two
+arguments.</p>
+</div>
+<div class="section" id="id424">
+<h5><a class="toc-backref" href="#id1241">Arguments:</a><a class="headerlink" href="#id424" title="Permalink to this headline">¶</a></h5>
+<p>The arguments and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id425">
+<h5><a class="toc-backref" href="#id1242">Semantics:</a><a class="headerlink" href="#id425" title="Permalink to this headline">¶</a></h5>
+<p>Follows the IEEE-754 semantics for minNum, which also match for libm’s
+fmin.</p>
+<p>If either operand is a NaN, returns the other non-NaN operand. Returns
+NaN only if both operands are NaN. If the operands compare equal,
+returns a value that compares equal to both operands. This means that
+fmin(+/-0.0, +/-0.0) could return either -0.0 or 0.0.</p>
+</div>
+</div>
+<div class="section" id="llvm-maxnum-intrinsic">
+<h4><a class="toc-backref" href="#id1243">‘<tt class="docutils literal"><span class="pre">llvm.maxnum.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-maxnum-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id426">
+<h5><a class="toc-backref" href="#id1244">Syntax:</a><a class="headerlink" href="#id426" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.maxnum</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.maxnum.f32(float %Val0, float %Val1l)
+declare double @llvm.maxnum.f64(double %Val0, double %Val1)
+declare x86_fp80 @llvm.maxnum.f80(x86_fp80 %Val0, x86_fp80 %Val1)
+declare fp128 @llvm.maxnum.f128(fp128 %Val0, fp128 %Val1)
+declare ppc_fp128 @llvm.maxnum.ppcf128(ppc_fp128 %Val0, ppc_fp128 %Val1)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id427">
+<h5><a class="toc-backref" href="#id1245">Overview:</a><a class="headerlink" href="#id427" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.maxnum.*</span></tt>‘ intrinsics return the maximum of the two
+arguments.</p>
+</div>
+<div class="section" id="id428">
+<h5><a class="toc-backref" href="#id1246">Arguments:</a><a class="headerlink" href="#id428" title="Permalink to this headline">¶</a></h5>
+<p>The arguments and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id429">
+<h5><a class="toc-backref" href="#id1247">Semantics:</a><a class="headerlink" href="#id429" title="Permalink to this headline">¶</a></h5>
+<p>Follows the IEEE-754 semantics for maxNum, which also match for libm’s
+fmax.</p>
+<p>If either operand is a NaN, returns the other non-NaN operand. Returns
+NaN only if both operands are NaN. If the operands compare equal,
+returns a value that compares equal to both operands. This means that
+fmax(+/-0.0, +/-0.0) could return either -0.0 or 0.0.</p>
+</div>
+</div>
+<div class="section" id="llvm-copysign-intrinsic">
+<h4><a class="toc-backref" href="#id1248">‘<tt class="docutils literal"><span class="pre">llvm.copysign.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-copysign-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id430">
+<h5><a class="toc-backref" href="#id1249">Syntax:</a><a class="headerlink" href="#id430" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.copysign</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.copysign.f32(float %Mag, float %Sgn)
+declare double @llvm.copysign.f64(double %Mag, double %Sgn)
+declare x86_fp80 @llvm.copysign.f80(x86_fp80 %Mag, x86_fp80 %Sgn)
+declare fp128 @llvm.copysign.f128(fp128 %Mag, fp128 %Sgn)
+declare ppc_fp128 @llvm.copysign.ppcf128(ppc_fp128 %Mag, ppc_fp128 %Sgn)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id431">
+<h5><a class="toc-backref" href="#id1250">Overview:</a><a class="headerlink" href="#id431" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.copysign.*</span></tt>‘ intrinsics return a value with the magnitude of the
+first operand and the sign of the second operand.</p>
+</div>
+<div class="section" id="id432">
+<h5><a class="toc-backref" href="#id1251">Arguments:</a><a class="headerlink" href="#id432" title="Permalink to this headline">¶</a></h5>
+<p>The arguments and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id433">
+<h5><a class="toc-backref" href="#id1252">Semantics:</a><a class="headerlink" href="#id433" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">copysign</span></tt>
+functions would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-floor-intrinsic">
+<h4><a class="toc-backref" href="#id1253">‘<tt class="docutils literal"><span class="pre">llvm.floor.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-floor-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id434">
+<h5><a class="toc-backref" href="#id1254">Syntax:</a><a class="headerlink" href="#id434" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.floor</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.floor.f32(float %Val)
+declare double @llvm.floor.f64(double %Val)
+declare x86_fp80 @llvm.floor.f80(x86_fp80 %Val)
+declare fp128 @llvm.floor.f128(fp128 %Val)
+declare ppc_fp128 @llvm.floor.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id435">
+<h5><a class="toc-backref" href="#id1255">Overview:</a><a class="headerlink" href="#id435" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.floor.*</span></tt>‘ intrinsics return the floor of the operand.</p>
+</div>
+<div class="section" id="id436">
+<h5><a class="toc-backref" href="#id1256">Arguments:</a><a class="headerlink" href="#id436" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id437">
+<h5><a class="toc-backref" href="#id1257">Semantics:</a><a class="headerlink" href="#id437" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">floor</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-ceil-intrinsic">
+<h4><a class="toc-backref" href="#id1258">‘<tt class="docutils literal"><span class="pre">llvm.ceil.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-ceil-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id438">
+<h5><a class="toc-backref" href="#id1259">Syntax:</a><a class="headerlink" href="#id438" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.ceil</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.ceil.f32(float %Val)
+declare double @llvm.ceil.f64(double %Val)
+declare x86_fp80 @llvm.ceil.f80(x86_fp80 %Val)
+declare fp128 @llvm.ceil.f128(fp128 %Val)
+declare ppc_fp128 @llvm.ceil.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id439">
+<h5><a class="toc-backref" href="#id1260">Overview:</a><a class="headerlink" href="#id439" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.ceil.*</span></tt>‘ intrinsics return the ceiling of the operand.</p>
+</div>
+<div class="section" id="id440">
+<h5><a class="toc-backref" href="#id1261">Arguments:</a><a class="headerlink" href="#id440" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id441">
+<h5><a class="toc-backref" href="#id1262">Semantics:</a><a class="headerlink" href="#id441" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">ceil</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-trunc-intrinsic">
+<h4><a class="toc-backref" href="#id1263">‘<tt class="docutils literal"><span class="pre">llvm.trunc.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-trunc-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id442">
+<h5><a class="toc-backref" href="#id1264">Syntax:</a><a class="headerlink" href="#id442" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.trunc</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.trunc.f32(float %Val)
+declare double @llvm.trunc.f64(double %Val)
+declare x86_fp80 @llvm.trunc.f80(x86_fp80 %Val)
+declare fp128 @llvm.trunc.f128(fp128 %Val)
+declare ppc_fp128 @llvm.trunc.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id443">
+<h5><a class="toc-backref" href="#id1265">Overview:</a><a class="headerlink" href="#id443" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.trunc.*</span></tt>‘ intrinsics returns the operand rounded to the
+nearest integer not larger in magnitude than the operand.</p>
+</div>
+<div class="section" id="id444">
+<h5><a class="toc-backref" href="#id1266">Arguments:</a><a class="headerlink" href="#id444" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id445">
+<h5><a class="toc-backref" href="#id1267">Semantics:</a><a class="headerlink" href="#id445" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">trunc</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-rint-intrinsic">
+<h4><a class="toc-backref" href="#id1268">‘<tt class="docutils literal"><span class="pre">llvm.rint.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-rint-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id446">
+<h5><a class="toc-backref" href="#id1269">Syntax:</a><a class="headerlink" href="#id446" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.rint</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.rint.f32(float %Val)
+declare double @llvm.rint.f64(double %Val)
+declare x86_fp80 @llvm.rint.f80(x86_fp80 %Val)
+declare fp128 @llvm.rint.f128(fp128 %Val)
+declare ppc_fp128 @llvm.rint.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id447">
+<h5><a class="toc-backref" href="#id1270">Overview:</a><a class="headerlink" href="#id447" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.rint.*</span></tt>‘ intrinsics returns the operand rounded to the
+nearest integer. It may raise an inexact floating-point exception if the
+operand isn’t an integer.</p>
+</div>
+<div class="section" id="id448">
+<h5><a class="toc-backref" href="#id1271">Arguments:</a><a class="headerlink" href="#id448" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id449">
+<h5><a class="toc-backref" href="#id1272">Semantics:</a><a class="headerlink" href="#id449" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">rint</span></tt> functions
+would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-nearbyint-intrinsic">
+<h4><a class="toc-backref" href="#id1273">‘<tt class="docutils literal"><span class="pre">llvm.nearbyint.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-nearbyint-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id450">
+<h5><a class="toc-backref" href="#id1274">Syntax:</a><a class="headerlink" href="#id450" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.nearbyint</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.nearbyint.f32(float %Val)
+declare double @llvm.nearbyint.f64(double %Val)
+declare x86_fp80 @llvm.nearbyint.f80(x86_fp80 %Val)
+declare fp128 @llvm.nearbyint.f128(fp128 %Val)
+declare ppc_fp128 @llvm.nearbyint.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id451">
+<h5><a class="toc-backref" href="#id1275">Overview:</a><a class="headerlink" href="#id451" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.nearbyint.*</span></tt>‘ intrinsics returns the operand rounded to the
+nearest integer.</p>
+</div>
+<div class="section" id="id452">
+<h5><a class="toc-backref" href="#id1276">Arguments:</a><a class="headerlink" href="#id452" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id453">
+<h5><a class="toc-backref" href="#id1277">Semantics:</a><a class="headerlink" href="#id453" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">nearbyint</span></tt>
+functions would, and handles error conditions in the same way.</p>
+</div>
+</div>
+<div class="section" id="llvm-round-intrinsic">
+<h4><a class="toc-backref" href="#id1278">‘<tt class="docutils literal"><span class="pre">llvm.round.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-round-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id454">
+<h5><a class="toc-backref" href="#id1279">Syntax:</a><a class="headerlink" href="#id454" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.round</span></tt> on any
+floating point or vector of floating point type. Not all targets support
+all types however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.round.f32(float %Val)
+declare double @llvm.round.f64(double %Val)
+declare x86_fp80 @llvm.round.f80(x86_fp80 %Val)
+declare fp128 @llvm.round.f128(fp128 %Val)
+declare ppc_fp128 @llvm.round.ppcf128(ppc_fp128 %Val)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id455">
+<h5><a class="toc-backref" href="#id1280">Overview:</a><a class="headerlink" href="#id455" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.round.*</span></tt>‘ intrinsics returns the operand rounded to the
+nearest integer.</p>
+</div>
+<div class="section" id="id456">
+<h5><a class="toc-backref" href="#id1281">Arguments:</a><a class="headerlink" href="#id456" title="Permalink to this headline">¶</a></h5>
+<p>The argument and return value are floating point numbers of the same
+type.</p>
+</div>
+<div class="section" id="id457">
+<h5><a class="toc-backref" href="#id1282">Semantics:</a><a class="headerlink" href="#id457" title="Permalink to this headline">¶</a></h5>
+<p>This function returns the same values as the libm <tt class="docutils literal"><span class="pre">round</span></tt>
+functions would, and handles error conditions in the same way.</p>
+</div>
+</div>
+</div>
+<div class="section" id="bit-manipulation-intrinsics">
+<h3><a class="toc-backref" href="#id1283">Bit Manipulation Intrinsics</a><a class="headerlink" href="#bit-manipulation-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>LLVM provides intrinsics for a few important bit manipulation
+operations. These allow efficient code generation for some algorithms.</p>
+<div class="section" id="llvm-bswap-intrinsics">
+<h4><a class="toc-backref" href="#id1284">‘<tt class="docutils literal"><span class="pre">llvm.bswap.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-bswap-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id458">
+<h5><a class="toc-backref" href="#id1285">Syntax:</a><a class="headerlink" href="#id458" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic function. You can use bswap on any
+integer type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
+<div class="highlight-python"><div class="highlight"><pre>declare i16 @llvm.bswap.i16(i16 <id>)
+declare i32 @llvm.bswap.i32(i32 <id>)
+declare i64 @llvm.bswap.i64(i64 <id>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id459">
+<h5><a class="toc-backref" href="#id1286">Overview:</a><a class="headerlink" href="#id459" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.bswap</span></tt>‘ family of intrinsics is used to byte swap integer
+values with an even number of bytes (positive multiple of 16 bits).
+These are useful for performing operations on data that is not in the
+target’s native byte order.</p>
+</div>
+<div class="section" id="id460">
+<h5><a class="toc-backref" href="#id1287">Semantics:</a><a class="headerlink" href="#id460" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.bswap.i16</span></tt> intrinsic returns an i16 value that has the high
+and low byte of the input i16 swapped. Similarly, the <tt class="docutils literal"><span class="pre">llvm.bswap.i32</span></tt>
+intrinsic returns an i32 value that has the four bytes of the input i32
+swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the
+returned i32 will have its bytes in 3, 2, 1, 0 order. The
+<tt class="docutils literal"><span class="pre">llvm.bswap.i48</span></tt>, <tt class="docutils literal"><span class="pre">llvm.bswap.i64</span></tt> and other intrinsics extend this
+concept to additional even-byte lengths (6 bytes, 8 bytes and more,
+respectively).</p>
+</div>
+</div>
+<div class="section" id="llvm-ctpop-intrinsic">
+<h4><a class="toc-backref" href="#id1288">‘<tt class="docutils literal"><span class="pre">llvm.ctpop.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-ctpop-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id461">
+<h5><a class="toc-backref" href="#id1289">Syntax:</a><a class="headerlink" href="#id461" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer
+bit width, or on any vector with integer elements. Not all targets
+support all bit widths or vector types, however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare i8 @llvm.ctpop.i8(i8 <src>)
+declare i16 @llvm.ctpop.i16(i16 <src>)
+declare i32 @llvm.ctpop.i32(i32 <src>)
+declare i64 @llvm.ctpop.i64(i64 <src>)
+declare i256 @llvm.ctpop.i256(i256 <src>)
+declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id462">
+<h5><a class="toc-backref" href="#id1290">Overview:</a><a class="headerlink" href="#id462" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.ctpop</span></tt>‘ family of intrinsics counts the number of bits set
+in a value.</p>
+</div>
+<div class="section" id="id463">
+<h5><a class="toc-backref" href="#id1291">Arguments:</a><a class="headerlink" href="#id463" title="Permalink to this headline">¶</a></h5>
+<p>The only argument is the value to be counted. The argument may be of any
+integer type, or a vector with integer elements. The return type must
+match the argument type.</p>
+</div>
+<div class="section" id="id464">
+<h5><a class="toc-backref" href="#id1292">Semantics:</a><a class="headerlink" href="#id464" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.ctpop</span></tt>‘ intrinsic counts the 1’s in a variable, or within
+each element of a vector.</p>
+</div>
+</div>
+<div class="section" id="llvm-ctlz-intrinsic">
+<h4><a class="toc-backref" href="#id1293">‘<tt class="docutils literal"><span class="pre">llvm.ctlz.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-ctlz-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id465">
+<h5><a class="toc-backref" href="#id1294">Syntax:</a><a class="headerlink" href="#id465" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.ctlz</span></tt> on any
+integer bit width, or any vector whose elements are integers. Not all
+targets support all bit widths or vector types, however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare i8 @llvm.ctlz.i8 (i8 <src>, i1 <is_zero_undef>)
+declare i16 @llvm.ctlz.i16 (i16 <src>, i1 <is_zero_undef>)
+declare i32 @llvm.ctlz.i32 (i32 <src>, i1 <is_zero_undef>)
+declare i64 @llvm.ctlz.i64 (i64 <src>, i1 <is_zero_undef>)
+declare i256 @llvm.ctlz.i256(i256 <src>, i1 <is_zero_undef>)
+declase <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id466">
+<h5><a class="toc-backref" href="#id1295">Overview:</a><a class="headerlink" href="#id466" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.ctlz</span></tt>‘ family of intrinsic functions counts the number of
+leading zeros in a variable.</p>
+</div>
+<div class="section" id="id467">
+<h5><a class="toc-backref" href="#id1296">Arguments:</a><a class="headerlink" href="#id467" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is the value to be counted. This argument may be of
+any integer type, or a vector with integer element type. The return
+type must match the first argument type.</p>
+<p>The second argument must be a constant and is a flag to indicate whether
+the intrinsic should ensure that a zero as the first argument produces a
+defined result. Historically some architectures did not provide a
+defined result for zero values as efficiently, and many algorithms are
+now predicated on avoiding zero-value inputs.</p>
+</div>
+<div class="section" id="id468">
+<h5><a class="toc-backref" href="#id1297">Semantics:</a><a class="headerlink" href="#id468" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.ctlz</span></tt>‘ intrinsic counts the leading (most significant)
+zeros in a variable, or within each element of the vector. If
+<tt class="docutils literal"><span class="pre">src</span> <span class="pre">==</span> <span class="pre">0</span></tt> then the result is the size in bits of the type of <tt class="docutils literal"><span class="pre">src</span></tt>
+if <tt class="docutils literal"><span class="pre">is_zero_undef</span> <span class="pre">==</span> <span class="pre">0</span></tt> and <tt class="docutils literal"><span class="pre">undef</span></tt> otherwise. For example,
+<tt class="docutils literal"><span class="pre">llvm.ctlz(i32</span> <span class="pre">2)</span> <span class="pre">=</span> <span class="pre">30</span></tt>.</p>
+</div>
+</div>
+<div class="section" id="llvm-cttz-intrinsic">
+<h4><a class="toc-backref" href="#id1298">‘<tt class="docutils literal"><span class="pre">llvm.cttz.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-cttz-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id469">
+<h5><a class="toc-backref" href="#id1299">Syntax:</a><a class="headerlink" href="#id469" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.cttz</span></tt> on any
+integer bit width, or any vector of integer elements. Not all targets
+support all bit widths or vector types, however.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare i8 @llvm.cttz.i8 (i8 <src>, i1 <is_zero_undef>)
+declare i16 @llvm.cttz.i16 (i16 <src>, i1 <is_zero_undef>)
+declare i32 @llvm.cttz.i32 (i32 <src>, i1 <is_zero_undef>)
+declare i64 @llvm.cttz.i64 (i64 <src>, i1 <is_zero_undef>)
+declare i256 @llvm.cttz.i256(i256 <src>, i1 <is_zero_undef>)
+declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>, i1 <is_zero_undef>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id470">
+<h5><a class="toc-backref" href="#id1300">Overview:</a><a class="headerlink" href="#id470" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.cttz</span></tt>‘ family of intrinsic functions counts the number of
+trailing zeros.</p>
+</div>
+<div class="section" id="id471">
+<h5><a class="toc-backref" href="#id1301">Arguments:</a><a class="headerlink" href="#id471" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is the value to be counted. This argument may be of
+any integer type, or a vector with integer element type. The return
+type must match the first argument type.</p>
+<p>The second argument must be a constant and is a flag to indicate whether
+the intrinsic should ensure that a zero as the first argument produces a
+defined result. Historically some architectures did not provide a
+defined result for zero values as efficiently, and many algorithms are
+now predicated on avoiding zero-value inputs.</p>
+</div>
+<div class="section" id="id472">
+<h5><a class="toc-backref" href="#id1302">Semantics:</a><a class="headerlink" href="#id472" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.cttz</span></tt>‘ intrinsic counts the trailing (least significant)
+zeros in a variable, or within each element of a vector. If <tt class="docutils literal"><span class="pre">src</span> <span class="pre">==</span> <span class="pre">0</span></tt>
+then the result is the size in bits of the type of <tt class="docutils literal"><span class="pre">src</span></tt> if
+<tt class="docutils literal"><span class="pre">is_zero_undef</span> <span class="pre">==</span> <span class="pre">0</span></tt> and <tt class="docutils literal"><span class="pre">undef</span></tt> otherwise. For example,
+<tt class="docutils literal"><span class="pre">llvm.cttz(2)</span> <span class="pre">=</span> <span class="pre">1</span></tt>.</p>
+</div>
+</div>
+</div>
+<div class="section" id="arithmetic-with-overflow-intrinsics">
+<span id="int-overflow"></span><h3><a class="toc-backref" href="#id1303">Arithmetic with Overflow Intrinsics</a><a class="headerlink" href="#arithmetic-with-overflow-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
+<div class="section" id="llvm-sadd-with-overflow-intrinsics">
+<h4><a class="toc-backref" href="#id1304">‘<tt class="docutils literal"><span class="pre">llvm.sadd.with.overflow.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-sadd-with-overflow-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id473">
+<h5><a class="toc-backref" href="#id1305">Syntax:</a><a class="headerlink" href="#id473" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.sadd.with.overflow</span></tt>
+on any integer bit width.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
+declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
+declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id474">
+<h5><a class="toc-backref" href="#id1306">Overview:</a><a class="headerlink" href="#id474" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.sadd.with.overflow</span></tt>‘ family of intrinsic functions perform
+a signed addition of the two arguments, and indicate whether an overflow
+occurred during the signed summation.</p>
+</div>
+<div class="section" id="id475">
+<h5><a class="toc-backref" href="#id1307">Arguments:</a><a class="headerlink" href="#id475" title="Permalink to this headline">¶</a></h5>
+<p>The arguments (%a and %b) and the first element of the result structure
+may be of integer types of any bit width, but they must have the same
+bit width. The second element of the result structure must be of type
+<tt class="docutils literal"><span class="pre">i1</span></tt>. <tt class="docutils literal"><span class="pre">%a</span></tt> and <tt class="docutils literal"><span class="pre">%b</span></tt> are the two values that will undergo signed
+addition.</p>
+</div>
+<div class="section" id="id476">
+<h5><a class="toc-backref" href="#id1308">Semantics:</a><a class="headerlink" href="#id476" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.sadd.with.overflow</span></tt>‘ family of intrinsic functions perform
+a signed addition of the two variables. They return a structure — the
+first element of which is the signed summation, and the second element
+of which is a bit specifying if the signed summation resulted in an
+overflow.</p>
+</div>
+<div class="section" id="id477">
+<h5><a class="toc-backref" href="#id1309">Examples:</a><a class="headerlink" href="#id477" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%res</span> <span class="p">=</span> <span class="k">call</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="vg">@llvm.sadd.with.overflow.i32</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%a</span><span class="p">,</span> <span class="k">i32</span> <span class="nv">%b</span><span class="p">)</span>
+<span class="nv">%sum</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">0</span>
+<span class="nv">%obit</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">1</span>
+<span class="k">br</span> <span class="k">i1</span> <span class="nv">%obit</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%overflow</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%normal</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="llvm-uadd-with-overflow-intrinsics">
+<h4><a class="toc-backref" href="#id1310">‘<tt class="docutils literal"><span class="pre">llvm.uadd.with.overflow.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-uadd-with-overflow-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id478">
+<h5><a class="toc-backref" href="#id1311">Syntax:</a><a class="headerlink" href="#id478" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.uadd.with.overflow</span></tt>
+on any integer bit width.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
+declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
+declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id479">
+<h5><a class="toc-backref" href="#id1312">Overview:</a><a class="headerlink" href="#id479" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.uadd.with.overflow</span></tt>‘ family of intrinsic functions perform
+an unsigned addition of the two arguments, and indicate whether a carry
+occurred during the unsigned summation.</p>
+</div>
+<div class="section" id="id480">
+<h5><a class="toc-backref" href="#id1313">Arguments:</a><a class="headerlink" href="#id480" title="Permalink to this headline">¶</a></h5>
+<p>The arguments (%a and %b) and the first element of the result structure
+may be of integer types of any bit width, but they must have the same
+bit width. The second element of the result structure must be of type
+<tt class="docutils literal"><span class="pre">i1</span></tt>. <tt class="docutils literal"><span class="pre">%a</span></tt> and <tt class="docutils literal"><span class="pre">%b</span></tt> are the two values that will undergo unsigned
+addition.</p>
+</div>
+<div class="section" id="id481">
+<h5><a class="toc-backref" href="#id1314">Semantics:</a><a class="headerlink" href="#id481" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.uadd.with.overflow</span></tt>‘ family of intrinsic functions perform
+an unsigned addition of the two arguments. They return a structure — the
+first element of which is the sum, and the second element of which is a
+bit specifying if the unsigned summation resulted in a carry.</p>
+</div>
+<div class="section" id="id482">
+<h5><a class="toc-backref" href="#id1315">Examples:</a><a class="headerlink" href="#id482" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%res</span> <span class="p">=</span> <span class="k">call</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="vg">@llvm.uadd.with.overflow.i32</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%a</span><span class="p">,</span> <span class="k">i32</span> <span class="nv">%b</span><span class="p">)</span>
+<span class="nv">%sum</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">0</span>
+<span class="nv">%obit</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">1</span>
+<span class="k">br</span> <span class="k">i1</span> <span class="nv">%obit</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%carry</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%normal</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="llvm-ssub-with-overflow-intrinsics">
+<h4><a class="toc-backref" href="#id1316">‘<tt class="docutils literal"><span class="pre">llvm.ssub.with.overflow.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-ssub-with-overflow-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id483">
+<h5><a class="toc-backref" href="#id1317">Syntax:</a><a class="headerlink" href="#id483" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.ssub.with.overflow</span></tt>
+on any integer bit width.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
+declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
+declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id484">
+<h5><a class="toc-backref" href="#id1318">Overview:</a><a class="headerlink" href="#id484" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.ssub.with.overflow</span></tt>‘ family of intrinsic functions perform
+a signed subtraction of the two arguments, and indicate whether an
+overflow occurred during the signed subtraction.</p>
+</div>
+<div class="section" id="id485">
+<h5><a class="toc-backref" href="#id1319">Arguments:</a><a class="headerlink" href="#id485" title="Permalink to this headline">¶</a></h5>
+<p>The arguments (%a and %b) and the first element of the result structure
+may be of integer types of any bit width, but they must have the same
+bit width. The second element of the result structure must be of type
+<tt class="docutils literal"><span class="pre">i1</span></tt>. <tt class="docutils literal"><span class="pre">%a</span></tt> and <tt class="docutils literal"><span class="pre">%b</span></tt> are the two values that will undergo signed
+subtraction.</p>
+</div>
+<div class="section" id="id486">
+<h5><a class="toc-backref" href="#id1320">Semantics:</a><a class="headerlink" href="#id486" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.ssub.with.overflow</span></tt>‘ family of intrinsic functions perform
+a signed subtraction of the two arguments. They return a structure — the
+first element of which is the subtraction, and the second element of
+which is a bit specifying if the signed subtraction resulted in an
+overflow.</p>
+</div>
+<div class="section" id="id487">
+<h5><a class="toc-backref" href="#id1321">Examples:</a><a class="headerlink" href="#id487" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%res</span> <span class="p">=</span> <span class="k">call</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="vg">@llvm.ssub.with.overflow.i32</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%a</span><span class="p">,</span> <span class="k">i32</span> <span class="nv">%b</span><span class="p">)</span>
+<span class="nv">%sum</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">0</span>
+<span class="nv">%obit</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">1</span>
+<span class="k">br</span> <span class="k">i1</span> <span class="nv">%obit</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%overflow</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%normal</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="llvm-usub-with-overflow-intrinsics">
+<h4><a class="toc-backref" href="#id1322">‘<tt class="docutils literal"><span class="pre">llvm.usub.with.overflow.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-usub-with-overflow-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id488">
+<h5><a class="toc-backref" href="#id1323">Syntax:</a><a class="headerlink" href="#id488" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.usub.with.overflow</span></tt>
+on any integer bit width.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
+declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
+declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id489">
+<h5><a class="toc-backref" href="#id1324">Overview:</a><a class="headerlink" href="#id489" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.usub.with.overflow</span></tt>‘ family of intrinsic functions perform
+an unsigned subtraction of the two arguments, and indicate whether an
+overflow occurred during the unsigned subtraction.</p>
+</div>
+<div class="section" id="id490">
+<h5><a class="toc-backref" href="#id1325">Arguments:</a><a class="headerlink" href="#id490" title="Permalink to this headline">¶</a></h5>
+<p>The arguments (%a and %b) and the first element of the result structure
+may be of integer types of any bit width, but they must have the same
+bit width. The second element of the result structure must be of type
+<tt class="docutils literal"><span class="pre">i1</span></tt>. <tt class="docutils literal"><span class="pre">%a</span></tt> and <tt class="docutils literal"><span class="pre">%b</span></tt> are the two values that will undergo unsigned
+subtraction.</p>
+</div>
+<div class="section" id="id491">
+<h5><a class="toc-backref" href="#id1326">Semantics:</a><a class="headerlink" href="#id491" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.usub.with.overflow</span></tt>‘ family of intrinsic functions perform
+an unsigned subtraction of the two arguments. They return a structure —
+the first element of which is the subtraction, and the second element of
+which is a bit specifying if the unsigned subtraction resulted in an
+overflow.</p>
+</div>
+<div class="section" id="id492">
+<h5><a class="toc-backref" href="#id1327">Examples:</a><a class="headerlink" href="#id492" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%res</span> <span class="p">=</span> <span class="k">call</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="vg">@llvm.usub.with.overflow.i32</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%a</span><span class="p">,</span> <span class="k">i32</span> <span class="nv">%b</span><span class="p">)</span>
+<span class="nv">%sum</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">0</span>
+<span class="nv">%obit</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">1</span>
+<span class="k">br</span> <span class="k">i1</span> <span class="nv">%obit</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%overflow</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%normal</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="llvm-smul-with-overflow-intrinsics">
+<h4><a class="toc-backref" href="#id1328">‘<tt class="docutils literal"><span class="pre">llvm.smul.with.overflow.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-smul-with-overflow-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id493">
+<h5><a class="toc-backref" href="#id1329">Syntax:</a><a class="headerlink" href="#id493" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.smul.with.overflow</span></tt>
+on any integer bit width.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
+declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
+declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id494">
+<h5><a class="toc-backref" href="#id1330">Overview:</a><a class="headerlink" href="#id494" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.smul.with.overflow</span></tt>‘ family of intrinsic functions perform
+a signed multiplication of the two arguments, and indicate whether an
+overflow occurred during the signed multiplication.</p>
+</div>
+<div class="section" id="id495">
+<h5><a class="toc-backref" href="#id1331">Arguments:</a><a class="headerlink" href="#id495" title="Permalink to this headline">¶</a></h5>
+<p>The arguments (%a and %b) and the first element of the result structure
+may be of integer types of any bit width, but they must have the same
+bit width. The second element of the result structure must be of type
+<tt class="docutils literal"><span class="pre">i1</span></tt>. <tt class="docutils literal"><span class="pre">%a</span></tt> and <tt class="docutils literal"><span class="pre">%b</span></tt> are the two values that will undergo signed
+multiplication.</p>
+</div>
+<div class="section" id="id496">
+<h5><a class="toc-backref" href="#id1332">Semantics:</a><a class="headerlink" href="#id496" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.smul.with.overflow</span></tt>‘ family of intrinsic functions perform
+a signed multiplication of the two arguments. They return a structure —
+the first element of which is the multiplication, and the second element
+of which is a bit specifying if the signed multiplication resulted in an
+overflow.</p>
+</div>
+<div class="section" id="id497">
+<h5><a class="toc-backref" href="#id1333">Examples:</a><a class="headerlink" href="#id497" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%res</span> <span class="p">=</span> <span class="k">call</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="vg">@llvm.smul.with.overflow.i32</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%a</span><span class="p">,</span> <span class="k">i32</span> <span class="nv">%b</span><span class="p">)</span>
+<span class="nv">%sum</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">0</span>
+<span class="nv">%obit</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">1</span>
+<span class="k">br</span> <span class="k">i1</span> <span class="nv">%obit</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%overflow</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%normal</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="llvm-umul-with-overflow-intrinsics">
+<h4><a class="toc-backref" href="#id1334">‘<tt class="docutils literal"><span class="pre">llvm.umul.with.overflow.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-umul-with-overflow-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id498">
+<h5><a class="toc-backref" href="#id1335">Syntax:</a><a class="headerlink" href="#id498" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.umul.with.overflow</span></tt>
+on any integer bit width.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
+declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
+declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id499">
+<h5><a class="toc-backref" href="#id1336">Overview:</a><a class="headerlink" href="#id499" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.umul.with.overflow</span></tt>‘ family of intrinsic functions perform
+a unsigned multiplication of the two arguments, and indicate whether an
+overflow occurred during the unsigned multiplication.</p>
+</div>
+<div class="section" id="id500">
+<h5><a class="toc-backref" href="#id1337">Arguments:</a><a class="headerlink" href="#id500" title="Permalink to this headline">¶</a></h5>
+<p>The arguments (%a and %b) and the first element of the result structure
+may be of integer types of any bit width, but they must have the same
+bit width. The second element of the result structure must be of type
+<tt class="docutils literal"><span class="pre">i1</span></tt>. <tt class="docutils literal"><span class="pre">%a</span></tt> and <tt class="docutils literal"><span class="pre">%b</span></tt> are the two values that will undergo unsigned
+multiplication.</p>
+</div>
+<div class="section" id="id501">
+<h5><a class="toc-backref" href="#id1338">Semantics:</a><a class="headerlink" href="#id501" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.umul.with.overflow</span></tt>‘ family of intrinsic functions perform
+an unsigned multiplication of the two arguments. They return a structure —
+the first element of which is the multiplication, and the second
+element of which is a bit specifying if the unsigned multiplication
+resulted in an overflow.</p>
+</div>
+<div class="section" id="id502">
+<h5><a class="toc-backref" href="#id1339">Examples:</a><a class="headerlink" href="#id502" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%res</span> <span class="p">=</span> <span class="k">call</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="vg">@llvm.umul.with.overflow.i32</span><span class="p">(</span><span class="k">i32</span> <span class="nv">%a</span><span class="p">,</span> <span class="k">i32</span> <span class="nv">%b</span><span class="p">)</span>
+<span class="nv">%sum</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">0</span>
+<span class="nv">%obit</span> <span class="p">=</span> <span class="k">extractvalue</span> <span class="p">{</span><span class="k">i32</span><span class="p">,</span> <span class="k">i1</span><span class="p">}</span> <span class="nv">%res</span><span class="p">,</span> <span class="m">1</span>
+<span class="k">br</span> <span class="k">i1</span> <span class="nv">%obit</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%overflow</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%normal</span>
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="specialised-arithmetic-intrinsics">
+<h3><a class="toc-backref" href="#id1340">Specialised Arithmetic Intrinsics</a><a class="headerlink" href="#specialised-arithmetic-intrinsics" title="Permalink to this headline">¶</a></h3>
+<div class="section" id="llvm-canonicalize-intrinsic">
+<h4><a class="toc-backref" href="#id1341">‘<tt class="docutils literal"><span class="pre">llvm.canonicalize.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-canonicalize-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id503">
+<h5><a class="toc-backref" href="#id1342">Syntax:</a><a class="headerlink" href="#id503" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.canonicalize.f32(float %a)
+declare double @llvm.canonicalize.f64(double %b)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id504">
+<h5><a class="toc-backref" href="#id1343">Overview:</a><a class="headerlink" href="#id504" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.canonicalize.*</span></tt>‘ intrinsic returns the platform specific canonical
+encoding of a floating point number. This canonicalization is useful for
+implementing certain numeric primitives such as frexp. The canonical encoding is
+defined by IEEE-754-2008 to be:</p>
+<div class="highlight-python"><div class="highlight"><pre>2.1.8 canonical encoding: The preferred encoding of a floating-point
+representation in a format. Applied to declets, significands of finite
+numbers, infinities, and NaNs, especially in decimal formats.
+</pre></div>
+</div>
+<p>This operation can also be considered equivalent to the IEEE-754-2008
+conversion of a floating-point value to the same format. NaNs are handled
+according to section 6.2.</p>
+<p>Examples of non-canonical encodings:</p>
+<ul class="simple">
+<li>x87 pseudo denormals, pseudo NaNs, pseudo Infinity, Unnormals. These are
+converted to a canonical representation per hardware-specific protocol.</li>
+<li>Many normal decimal floating point numbers have non-canonical alternative
+encodings.</li>
+<li>Some machines, like GPUs or ARMv7 NEON, do not support subnormal values.
+These are treated as non-canonical encodings of zero and with be flushed to
+a zero of the same sign by this operation.</li>
+</ul>
+<p>Note that per IEEE-754-2008 6.2, systems that support signaling NaNs with
+default exception handling must signal an invalid exception, and produce a
+quiet NaN result.</p>
+<p>This function should always be implementable as multiplication by 1.0, provided
+that the compiler does not constant fold the operation. Likewise, division by
+1.0 and <tt class="docutils literal"><span class="pre">llvm.minnum(x,</span> <span class="pre">x)</span></tt> are possible implementations. Addition with
+-0.0 is also sufficient provided that the rounding mode is not -Infinity.</p>
+<p><tt class="docutils literal"><span class="pre">@llvm.canonicalize</span></tt> must preserve the equality relation. That is:</p>
+<ul class="simple">
+<li><tt class="docutils literal"><span class="pre">(@llvm.canonicalize(x)</span> <span class="pre">==</span> <span class="pre">x)</span></tt> is equivalent to <tt class="docutils literal"><span class="pre">(x</span> <span class="pre">==</span> <span class="pre">x)</span></tt></li>
+<li><tt class="docutils literal"><span class="pre">(@llvm.canonicalize(x)</span> <span class="pre">==</span> <span class="pre">@llvm.canonicalize(y))</span></tt> is equivalent to
+to <tt class="docutils literal"><span class="pre">(x</span> <span class="pre">==</span> <span class="pre">y)</span></tt></li>
+</ul>
+<p>Additionally, the sign of zero must be conserved:
+<tt class="docutils literal"><span class="pre">@llvm.canonicalize(-0.0)</span> <span class="pre">=</span> <span class="pre">-0.0</span></tt> and <tt class="docutils literal"><span class="pre">@llvm.canonicalize(+0.0)</span> <span class="pre">=</span> <span class="pre">+0.0</span></tt></p>
+<p>The payload bits of a NaN must be conserved, with two exceptions.
+First, environments which use only a single canonical representation of NaN
+must perform said canonicalization. Second, SNaNs must be quieted per the
+usual methods.</p>
+<p>The canonicalization operation may be optimized away if:</p>
+<ul class="simple">
+<li>The input is known to be canonical. For example, it was produced by a
+floating-point operation that is required by the standard to be canonical.</li>
+<li>The result is consumed only by (or fused with) other floating-point
+operations. That is, the bits of the floating point value are not examined.</li>
+</ul>
+</div>
+</div>
+<div class="section" id="llvm-fmuladd-intrinsic">
+<h4><a class="toc-backref" href="#id1344">‘<tt class="docutils literal"><span class="pre">llvm.fmuladd.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-fmuladd-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id505">
+<h5><a class="toc-backref" href="#id1345">Syntax:</a><a class="headerlink" href="#id505" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.fmuladd.f32(float %a, float %b, float %c)
+declare double @llvm.fmuladd.f64(double %a, double %b, double %c)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id506">
+<h5><a class="toc-backref" href="#id1346">Overview:</a><a class="headerlink" href="#id506" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.fmuladd.*</span></tt>‘ intrinsic functions represent multiply-add
+expressions that can be fused if the code generator determines that (a) the
+target instruction set has support for a fused operation, and (b) that the
+fused operation is more efficient than the equivalent, separate pair of mul
+and add instructions.</p>
+</div>
+<div class="section" id="id507">
+<h5><a class="toc-backref" href="#id1347">Arguments:</a><a class="headerlink" href="#id507" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.fmuladd.*</span></tt>‘ intrinsics each take three arguments: two
+multiplicands, a and b, and an addend c.</p>
+</div>
+<div class="section" id="id508">
+<h5><a class="toc-backref" href="#id1348">Semantics:</a><a class="headerlink" href="#id508" title="Permalink to this headline">¶</a></h5>
+<p>The expression:</p>
+<div class="highlight-python"><div class="highlight"><pre>%0 = call float @llvm.fmuladd.f32(%a, %b, %c)
+</pre></div>
+</div>
+<p>is equivalent to the expression a * b + c, except that rounding will
+not be performed between the multiplication and addition steps if the
+code generator fuses the operations. Fusion is not guaranteed, even if
+the target platform supports it. If a fused multiply-add is required the
+corresponding llvm.fma.* intrinsic function should be used
+instead. This never sets errno, just as ‘<tt class="docutils literal"><span class="pre">llvm.fma.*</span></tt>‘.</p>
+</div>
+<div class="section" id="id509">
+<h5><a class="toc-backref" href="#id1349">Examples:</a><a class="headerlink" href="#id509" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%r2</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">float</span> <span class="vg">@llvm.fmuladd.f32</span><span class="p">(</span><span class="kt">float</span> <span class="nv">%a</span><span class="p">,</span> <span class="kt">float</span> <span class="nv">%b</span><span class="p">,</span> <span class="kt">float</span> <span class="nv">%c</span><span class="p">)</span> <span class="c">; yields float:r2 = (a * b) + c</span>
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="half-precision-floating-point-intrinsics">
+<h3><a class="toc-backref" href="#id1350">Half Precision Floating Point Intrinsics</a><a class="headerlink" href="#half-precision-floating-point-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>For most target platforms, half precision floating point is a
+storage-only format. This means that it is a dense encoding (in memory)
+but does not support computation in the format.</p>
+<p>This means that code must first load the half-precision floating point
+value as an i16, then convert it to float with
+<a class="reference internal" href="#int-convert-from-fp16"><em>llvm.convert.from.fp16</em></a>. Computation can
+then be performed on the float value (including extending to double
+etc). To store the value back to memory, it is first converted to float
+if needed, then converted to i16 with
+<a class="reference internal" href="#int-convert-to-fp16"><em>llvm.convert.to.fp16</em></a>, then storing as an
+i16 value.</p>
+<div class="section" id="llvm-convert-to-fp16-intrinsic">
+<span id="int-convert-to-fp16"></span><h4><a class="toc-backref" href="#id1351">‘<tt class="docutils literal"><span class="pre">llvm.convert.to.fp16</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-convert-to-fp16-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id510">
+<h5><a class="toc-backref" href="#id1352">Syntax:</a><a class="headerlink" href="#id510" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i16 @llvm.convert.to.fp16.f32(float %a)
+declare i16 @llvm.convert.to.fp16.f64(double %a)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id511">
+<h5><a class="toc-backref" href="#id1353">Overview:</a><a class="headerlink" href="#id511" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.convert.to.fp16</span></tt>‘ intrinsic function performs a conversion from a
+conventional floating point type to half precision floating point format.</p>
+</div>
+<div class="section" id="id512">
+<h5><a class="toc-backref" href="#id1354">Arguments:</a><a class="headerlink" href="#id512" title="Permalink to this headline">¶</a></h5>
+<p>The intrinsic function contains single argument - the value to be
+converted.</p>
+</div>
+<div class="section" id="id513">
+<h5><a class="toc-backref" href="#id1355">Semantics:</a><a class="headerlink" href="#id513" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.convert.to.fp16</span></tt>‘ intrinsic function performs a conversion from a
+conventional floating point format to half precision floating point format. The
+return value is an <tt class="docutils literal"><span class="pre">i16</span></tt> which contains the converted number.</p>
+</div>
+<div class="section" id="id514">
+<h5><a class="toc-backref" href="#id1356">Examples:</a><a class="headerlink" href="#id514" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%res</span> <span class="p">=</span> <span class="k">call</span> <span class="k">i16</span> <span class="vg">@llvm.convert.to.fp16.f32</span><span class="p">(</span><span class="kt">float</span> <span class="nv">%a</span><span class="p">)</span>
+<span class="k">store</span> <span class="k">i16</span> <span class="nv">%res</span><span class="p">,</span> <span class="k">i16</span><span class="p">*</span> <span class="vg">@x</span><span class="p">,</span> <span class="k">align</span> <span class="m">2</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="llvm-convert-from-fp16-intrinsic">
+<span id="int-convert-from-fp16"></span><h4><a class="toc-backref" href="#id1357">‘<tt class="docutils literal"><span class="pre">llvm.convert.from.fp16</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-convert-from-fp16-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id515">
+<h5><a class="toc-backref" href="#id1358">Syntax:</a><a class="headerlink" href="#id515" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare float @llvm.convert.from.fp16.f32(i16 %a)
+declare double @llvm.convert.from.fp16.f64(i16 %a)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id516">
+<h5><a class="toc-backref" href="#id1359">Overview:</a><a class="headerlink" href="#id516" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.convert.from.fp16</span></tt>‘ intrinsic function performs a
+conversion from half precision floating point format to single precision
+floating point format.</p>
+</div>
+<div class="section" id="id517">
+<h5><a class="toc-backref" href="#id1360">Arguments:</a><a class="headerlink" href="#id517" title="Permalink to this headline">¶</a></h5>
+<p>The intrinsic function contains single argument - the value to be
+converted.</p>
+</div>
+<div class="section" id="id518">
+<h5><a class="toc-backref" href="#id1361">Semantics:</a><a class="headerlink" href="#id518" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.convert.from.fp16</span></tt>‘ intrinsic function performs a
+conversion from half single precision floating point format to single
+precision floating point format. The input half-float value is
+represented by an <tt class="docutils literal"><span class="pre">i16</span></tt> value.</p>
+</div>
+<div class="section" id="id519">
+<h5><a class="toc-backref" href="#id1362">Examples:</a><a class="headerlink" href="#id519" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%a</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i16</span><span class="p">,</span> <span class="k">i16</span><span class="p">*</span> <span class="vg">@x</span><span class="p">,</span> <span class="k">align</span> <span class="m">2</span>
+<span class="nv">%res</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">float</span> <span class="vg">@llvm.convert.from.fp16</span><span class="p">(</span><span class="k">i16</span> <span class="nv">%a</span><span class="p">)</span>
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="debugger-intrinsics">
+<span id="dbg-intrinsics"></span><h3><a class="toc-backref" href="#id1363">Debugger Intrinsics</a><a class="headerlink" href="#debugger-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>The LLVM debugger intrinsics (which all start with <tt class="docutils literal"><span class="pre">llvm.dbg.</span></tt>
+prefix), are described in the <a class="reference external" href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
+Debugging</a>
+document.</p>
+</div>
+<div class="section" id="exception-handling-intrinsics">
+<h3><a class="toc-backref" href="#id1364">Exception Handling Intrinsics</a><a class="headerlink" href="#exception-handling-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>The LLVM exception handling intrinsics (which all start with
+<tt class="docutils literal"><span class="pre">llvm.eh.</span></tt> prefix), are described in the <a class="reference external" href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
+Handling</a> document.</p>
+</div>
+<div class="section" id="trampoline-intrinsics">
+<span id="int-trampoline"></span><h3><a class="toc-backref" href="#id1365">Trampoline Intrinsics</a><a class="headerlink" href="#trampoline-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>These intrinsics make it possible to excise one parameter, marked with
+the <a class="reference internal" href="#nest"><em>nest</em></a> attribute, from a function. The result is a
+callable function pointer lacking the nest parameter - the caller does
+not need to provide a value for it. Instead, the value to use is stored
+in advance in a “trampoline”, a block of memory usually allocated on the
+stack, which also contains code to splice the nest value into the
+argument list. This is used to implement the GCC nested function address
+extension.</p>
+<p>For example, if the function is <tt class="docutils literal"><span class="pre">i32</span> <span class="pre">f(i8*</span> <span class="pre">nest</span> <span class="pre">%c,</span> <span class="pre">i32</span> <span class="pre">%x,</span> <span class="pre">i32</span> <span class="pre">%y)</span></tt>
+then the resulting function pointer has signature <tt class="docutils literal"><span class="pre">i32</span> <span class="pre">(i32,</span> <span class="pre">i32)*</span></tt>.
+It can be created as follows:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="nv">%tramp</span> <span class="p">=</span> <span class="k">alloca</span> <span class="p">[</span><span class="m">10</span> <span class="k">x</span> <span class="k">i8</span><span class="p">],</span> <span class="k">align</span> <span class="m">4</span> <span class="c">; size and alignment only correct for X86</span>
+<span class="nv">%tramp1</span> <span class="p">=</span> <span class="k">getelementptr</span> <span class="p">[</span><span class="m">10</span> <span class="k">x</span> <span class="k">i8</span><span class="p">],</span> <span class="p">[</span><span class="m">10</span> <span class="k">x</span> <span class="k">i8</span><span class="p">]*</span> <span class="nv">%tramp</span><span class="p">,</span> <span class="k">i32</span> <span class="m">0</span><span class="p">,</span> <span class="k">i32</span> <span class="m">0</span>
+<span class="k">call</span> <span class="k">i8</span><span class="p">*</span> <span class="vg">@llvm.init.trampoline</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%tramp1</span><span class="p">,</span> <span class="k">i8</span><span class="p">*</span> <span class="k">bitcast</span> <span class="p">(</span><span class="k">i32</span> <span class="p">(</span><span class="k">i8</span><span class="p">*,</span> <span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">)*</span> <span class="vg">@f</span> <span class="k">to</span> <span class="k">i8</span><span class="p">*),</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%nval</span><span class="p">)</span>
+<span class="nv">%p</span> <span class="p">=</span> <span class="k">call</span> <span class="k">i8</span><span class="p">*</span> <span class="vg">@llvm.adjust.trampoline</span><span class="p">(</span><span class="k">i8</span><span class="p">*</span> <span class="nv">%tramp1</span><span class="p">)</span>
+<span class="nv">%fp</span> <span class="p">=</span> <span class="k">bitcast</span> <span class="k">i8</span><span class="p">*</span> <span class="nv">%p</span> <span class="k">to</span> <span class="k">i32</span> <span class="p">(</span><span class="k">i32</span><span class="p">,</span> <span class="k">i32</span><span class="p">)*</span>
+</pre></div>
+</div>
+<p>The call <tt class="docutils literal"><span class="pre">%val</span> <span class="pre">=</span> <span class="pre">call</span> <span class="pre">i32</span> <span class="pre">%fp(i32</span> <span class="pre">%x,</span> <span class="pre">i32</span> <span class="pre">%y)</span></tt> is then equivalent to
+<tt class="docutils literal"><span class="pre">%val</span> <span class="pre">=</span> <span class="pre">call</span> <span class="pre">i32</span> <span class="pre">%f(i8*</span> <span class="pre">%nval,</span> <span class="pre">i32</span> <span class="pre">%x,</span> <span class="pre">i32</span> <span class="pre">%y)</span></tt>.</p>
+<div class="section" id="llvm-init-trampoline-intrinsic">
+<span id="int-it"></span><h4><a class="toc-backref" href="#id1366">‘<tt class="docutils literal"><span class="pre">llvm.init.trampoline</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-init-trampoline-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id520">
+<h5><a class="toc-backref" href="#id1367">Syntax:</a><a class="headerlink" href="#id520" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id521">
+<h5><a class="toc-backref" href="#id1368">Overview:</a><a class="headerlink" href="#id521" title="Permalink to this headline">¶</a></h5>
+<p>This fills the memory pointed to by <tt class="docutils literal"><span class="pre">tramp</span></tt> with executable code,
+turning it into a trampoline.</p>
+</div>
+<div class="section" id="id522">
+<h5><a class="toc-backref" href="#id1369">Arguments:</a><a class="headerlink" href="#id522" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.init.trampoline</span></tt> intrinsic takes three arguments, all
+pointers. The <tt class="docutils literal"><span class="pre">tramp</span></tt> argument must point to a sufficiently large and
+sufficiently aligned block of memory; this memory is written to by the
+intrinsic. Note that the size and the alignment are target-specific -
+LLVM currently provides no portable way of determining them, so a
+front-end that generates this intrinsic needs to have some
+target-specific knowledge. The <tt class="docutils literal"><span class="pre">func</span></tt> argument must hold a function
+bitcast to an <tt class="docutils literal"><span class="pre">i8*</span></tt>.</p>
+</div>
+<div class="section" id="id523">
+<h5><a class="toc-backref" href="#id1370">Semantics:</a><a class="headerlink" href="#id523" title="Permalink to this headline">¶</a></h5>
+<p>The block of memory pointed to by <tt class="docutils literal"><span class="pre">tramp</span></tt> is filled with target
+dependent code, turning it into a function. Then <tt class="docutils literal"><span class="pre">tramp</span></tt> needs to be
+passed to <a class="reference internal" href="#int-at"><em>llvm.adjust.trampoline</em></a> to get a pointer which can
+be <a class="reference internal" href="#int-trampoline"><em>bitcast (to a new function) and called</em></a>. The new
+function’s signature is the same as that of <tt class="docutils literal"><span class="pre">func</span></tt> with any arguments
+marked with the <tt class="docutils literal"><span class="pre">nest</span></tt> attribute removed. At most one such <tt class="docutils literal"><span class="pre">nest</span></tt>
+argument is allowed, and it must be of pointer type. Calling the new
+function is equivalent to calling <tt class="docutils literal"><span class="pre">func</span></tt> with the same argument list,
+but with <tt class="docutils literal"><span class="pre">nval</span></tt> used for the missing <tt class="docutils literal"><span class="pre">nest</span></tt> argument. If, after
+calling <tt class="docutils literal"><span class="pre">llvm.init.trampoline</span></tt>, the memory pointed to by <tt class="docutils literal"><span class="pre">tramp</span></tt> is
+modified, then the effect of any later call to the returned function
+pointer is undefined.</p>
+</div>
+</div>
+<div class="section" id="llvm-adjust-trampoline-intrinsic">
+<span id="int-at"></span><h4><a class="toc-backref" href="#id1371">‘<tt class="docutils literal"><span class="pre">llvm.adjust.trampoline</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-adjust-trampoline-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id524">
+<h5><a class="toc-backref" href="#id1372">Syntax:</a><a class="headerlink" href="#id524" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i8* @llvm.adjust.trampoline(i8* <tramp>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id525">
+<h5><a class="toc-backref" href="#id1373">Overview:</a><a class="headerlink" href="#id525" title="Permalink to this headline">¶</a></h5>
+<p>This performs any required machine-specific adjustment to the address of
+a trampoline (passed as <tt class="docutils literal"><span class="pre">tramp</span></tt>).</p>
+</div>
+<div class="section" id="id526">
+<h5><a class="toc-backref" href="#id1374">Arguments:</a><a class="headerlink" href="#id526" title="Permalink to this headline">¶</a></h5>
+<p><tt class="docutils literal"><span class="pre">tramp</span></tt> must point to a block of memory which already has trampoline
+code filled in by a previous call to
+<a class="reference internal" href="#int-it"><em>llvm.init.trampoline</em></a>.</p>
+</div>
+<div class="section" id="id527">
+<h5><a class="toc-backref" href="#id1375">Semantics:</a><a class="headerlink" href="#id527" title="Permalink to this headline">¶</a></h5>
+<p>On some architectures the address of the code to be executed needs to be
+different than the address where the trampoline is actually stored. This
+intrinsic returns the executable address corresponding to <tt class="docutils literal"><span class="pre">tramp</span></tt>
+after performing the required machine specific adjustments. The pointer
+returned can then be <a class="reference internal" href="#int-trampoline"><em>bitcast and executed</em></a>.</p>
+</div>
+</div>
+</div>
+<div class="section" id="masked-vector-load-and-store-intrinsics">
+<span id="int-mload-mstore"></span><h3><a class="toc-backref" href="#id1376">Masked Vector Load and Store Intrinsics</a><a class="headerlink" href="#masked-vector-load-and-store-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>LLVM provides intrinsics for predicated vector load and store operations. The predicate is specified by a mask operand, which holds one bit per vector element, switching the associated vector lane on or off. The memory addresses corresponding to the “off” lanes are not accessed. When all bits of the mask are on, the intrinsic is identical to a regular vector load or store. When all bits are off, no memory is accessed.</p>
+<div class="section" id="llvm-masked-load-intrinsics">
+<span id="int-mload"></span><h4><a class="toc-backref" href="#id1377">‘<tt class="docutils literal"><span class="pre">llvm.masked.load.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-masked-load-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id528">
+<h5><a class="toc-backref" href="#id1378">Syntax:</a><a class="headerlink" href="#id528" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. The loaded data is a vector of any integer or floating point data type.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare <16 x float> @llvm.masked.load.v16f32 (<16 x float>* <ptr>, i32 <alignment>, <16 x i1> <mask>, <16 x float> <passthru>)
+declare <2 x double> @llvm.masked.load.v2f64 (<2 x double>* <ptr>, i32 <alignment>, <2 x i1> <mask>, <2 x double> <passthru>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id529">
+<h5><a class="toc-backref" href="#id1379">Overview:</a><a class="headerlink" href="#id529" title="Permalink to this headline">¶</a></h5>
+<p>Reads a vector from memory according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes. The masked-off lanes in the result vector are taken from the corresponding lanes of the ‘<tt class="docutils literal"><span class="pre">passthru</span></tt>‘ operand.</p>
+</div>
+<div class="section" id="id530">
+<h5><a class="toc-backref" href="#id1380">Arguments:</a><a class="headerlink" href="#id530" title="Permalink to this headline">¶</a></h5>
+<p>The first operand is the base pointer for the load. The second operand is the alignment of the source location. It must be a constant integer value. The third operand, mask, is a vector of boolean values with the same number of elements as the return type. The fourth is a pass-through value that is used to fill the masked-off lanes of the result. The return type, underlying type of the base pointer and the type of the ‘<tt class="docutils literal"><span class="pre">passthru</span></tt>‘ operand are the same vector types.</p>
+</div>
+<div class="section" id="id531">
+<h5><a class="toc-backref" href="#id1381">Semantics:</a><a class="headerlink" href="#id531" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.masked.load</span></tt>‘ intrinsic is designed for conditional reading of selected vector elements in a single IR operation. It is useful for targets that support vector masked loads and allows vectorizing predicated basic blocks on these targets. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar load operations.
+The result of this operation is equivalent to a regular vector load instruction followed by a ‘select’ between the loaded and the passthru values, predicated on the same mask. However, using this intrinsic prevents exceptions on memory access to masked-off lanes.</p>
+<div class="highlight-python"><div class="highlight"><pre>%res = call <16 x float> @llvm.masked.load.v16f32 (<16 x float>* %ptr, i32 4, <16 x i1>%mask, <16 x float> %passthru)
+
+;; The result of the two following instructions is identical aside from potential memory access exception
+%loadlal = load <16 x float>, <16 x float>* %ptr, align 4
+%res = select <16 x i1> %mask, <16 x float> %loadlal, <16 x float> %passthru
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="llvm-masked-store-intrinsics">
+<span id="int-mstore"></span><h4><a class="toc-backref" href="#id1382">‘<tt class="docutils literal"><span class="pre">llvm.masked.store.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-masked-store-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id532">
+<h5><a class="toc-backref" href="#id1383">Syntax:</a><a class="headerlink" href="#id532" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. The data stored in memory is a vector of any integer or floating point data type.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.masked.store.v8i32 (<8 x i32> <value>, <8 x i32> * <ptr>, i32 <alignment>, <8 x i1> <mask>)
+declare void @llvm.masked.store.v16f32(<16 x i32> <value>, <16 x i32>* <ptr>, i32 <alignment>, <16 x i1> <mask>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id533">
+<h5><a class="toc-backref" href="#id1384">Overview:</a><a class="headerlink" href="#id533" title="Permalink to this headline">¶</a></h5>
+<p>Writes a vector to memory according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes.</p>
+</div>
+<div class="section" id="id534">
+<h5><a class="toc-backref" href="#id1385">Arguments:</a><a class="headerlink" href="#id534" title="Permalink to this headline">¶</a></h5>
+<p>The first operand is the vector value to be written to memory. The second operand is the base pointer for the store, it has the same underlying type as the value operand. The third operand is the alignment of the destination location. The fourth operand, mask, is a vector of boolean values. The types of the mask and the value operand must have the same number of vector elements.</p>
+</div>
+<div class="section" id="id535">
+<h5><a class="toc-backref" href="#id1386">Semantics:</a><a class="headerlink" href="#id535" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.masked.store</span></tt>‘ intrinsics is designed for conditional writing of selected vector elements in a single IR operation. It is useful for targets that support vector masked store and allows vectorizing predicated basic blocks on these targets. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar store operations.
+The result of this operation is equivalent to a load-modify-store sequence. However, using this intrinsic prevents exceptions and data races on memory access to masked-off lanes.</p>
+<div class="highlight-python"><div class="highlight"><pre>call void @llvm.masked.store.v16f32(<16 x float> %value, <16 x float>* %ptr, i32 4, <16 x i1> %mask)
+
+;; The result of the following instructions is identical aside from potential data races and memory access exceptions
+%oldval = load <16 x float>, <16 x float>* %ptr, align 4
+%res = select <16 x i1> %mask, <16 x float> %value, <16 x float> %oldval
+store <16 x float> %res, <16 x float>* %ptr, align 4
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="masked-vector-gather-and-scatter-intrinsics">
+<h3><a class="toc-backref" href="#id1387">Masked Vector Gather and Scatter Intrinsics</a><a class="headerlink" href="#masked-vector-gather-and-scatter-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>LLVM provides intrinsics for vector gather and scatter operations. They are similar to <a class="reference internal" href="#int-mload-mstore"><em>Masked Vector Load and Store</em></a>, except they are designed for arbitrary memory accesses, rather than sequential memory accesses. Gather and scatter also employ a mask operand, which holds one bit per vector element, switching the associated vector lane on or off. The memory addresses corresponding to the “off” lanes are not accessed. When all bits are off, no memory is accessed.</p>
+<div class="section" id="llvm-masked-gather-intrinsics">
+<span id="int-mgather"></span><h4><a class="toc-backref" href="#id1388">‘<tt class="docutils literal"><span class="pre">llvm.masked.gather.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-masked-gather-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id536">
+<h5><a class="toc-backref" href="#id1389">Syntax:</a><a class="headerlink" href="#id536" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. The loaded data are multiple scalar values of any integer or floating point data type gathered together into one vector.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare <16 x float> @llvm.masked.gather.v16f32 (<16 x float*> <ptrs>, i32 <alignment>, <16 x i1> <mask>, <16 x float> <passthru>)
+declare <2 x double> @llvm.masked.gather.v2f64 (<2 x double*> <ptrs>, i32 <alignment>, <2 x i1> <mask>, <2 x double> <passthru>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id537">
+<h5><a class="toc-backref" href="#id1390">Overview:</a><a class="headerlink" href="#id537" title="Permalink to this headline">¶</a></h5>
+<p>Reads scalar values from arbitrary memory locations and gathers them into one vector. The memory locations are provided in the vector of pointers ‘<tt class="docutils literal"><span class="pre">ptrs</span></tt>‘. The memory is accessed according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes. The masked-off lanes in the result vector are taken from the corresponding lanes of the ‘<tt class="docutils literal"><span class="pre">passthru</span></tt>‘ operand.</p>
+</div>
+<div class="section" id="id538">
+<h5><a class="toc-backref" href="#id1391">Arguments:</a><a class="headerlink" href="#id538" title="Permalink to this headline">¶</a></h5>
+<p>The first operand is a vector of pointers which holds all memory addresses to read. The second operand is an alignment of the source addresses. It must be a constant integer value. The third operand, mask, is a vector of boolean values with the same number of elements as the return type. The fourth is a pass-through value that is used to fill the masked-off lanes of the result. The return type, underlying type of the vector of pointers and the type of the ‘<tt class="docutils literal"><span class="pre">passthru</span></tt>‘ operand are the same vector types.</p>
+</div>
+<div class="section" id="id539">
+<h5><a class="toc-backref" href="#id1392">Semantics:</a><a class="headerlink" href="#id539" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.masked.gather</span></tt>‘ intrinsic is designed for conditional reading of multiple scalar values from arbitrary memory locations in a single IR operation. It is useful for targets that support vector masked gathers and allows vectorizing basic blocks with data and control divergence. Other targets may support this intrinsic differently, for example by lowering it into a sequence of scalar load operations.
+The semantics of this operation are equivalent to a sequence of conditional scalar loads with subsequent gathering all loaded values into a single vector. The mask restricts memory access to certain lanes and facilitates vectorization of predicated basic blocks.</p>
+<div class="highlight-python"><div class="highlight"><pre>%res = call <4 x double> @llvm.masked.gather.v4f64 (<4 x double*> %ptrs, i32 8, <4 x i1>%mask, <4 x double> <true, true, true, true>)
+
+;; The gather with all-true mask is equivalent to the following instruction sequence
+%ptr0 = extractelement <4 x double*> %ptrs, i32 0
+%ptr1 = extractelement <4 x double*> %ptrs, i32 1
+%ptr2 = extractelement <4 x double*> %ptrs, i32 2
+%ptr3 = extractelement <4 x double*> %ptrs, i32 3
+
+%val0 = load double, double* %ptr0, align 8
+%val1 = load double, double* %ptr1, align 8
+%val2 = load double, double* %ptr2, align 8
+%val3 = load double, double* %ptr3, align 8
+
+%vec0 = insertelement <4 x double>undef, %val0, 0
+%vec01 = insertelement <4 x double>%vec0, %val1, 1
+%vec012 = insertelement <4 x double>%vec01, %val2, 2
+%vec0123 = insertelement <4 x double>%vec012, %val3, 3
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="llvm-masked-scatter-intrinsics">
+<span id="int-mscatter"></span><h4><a class="toc-backref" href="#id1393">‘<tt class="docutils literal"><span class="pre">llvm.masked.scatter.*</span></tt>‘ Intrinsics</a><a class="headerlink" href="#llvm-masked-scatter-intrinsics" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id540">
+<h5><a class="toc-backref" href="#id1394">Syntax:</a><a class="headerlink" href="#id540" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. The data stored in memory is a vector of any integer or floating point data type. Each vector element is stored in an arbitrary memory addresses. Scatter with overlapping addresses is guaranteed to be ordered from least-significant to most-significant element.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.masked.scatter.v8i32 (<8 x i32> <value>, <8 x i32*> <ptrs>, i32 <alignment>, <8 x i1> <mask>)
+declare void @llvm.masked.scatter.v16f32(<16 x i32> <value>, <16 x i32*> <ptrs>, i32 <alignment>, <16 x i1> <mask>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id541">
+<h5><a class="toc-backref" href="#id1395">Overview:</a><a class="headerlink" href="#id541" title="Permalink to this headline">¶</a></h5>
+<p>Writes each element from the value vector to the corresponding memory address. The memory addresses are represented as a vector of pointers. Writing is done according to the provided mask. The mask holds a bit for each vector lane, and is used to prevent memory accesses to the masked-off lanes.</p>
+</div>
+<div class="section" id="id542">
+<h5><a class="toc-backref" href="#id1396">Arguments:</a><a class="headerlink" href="#id542" title="Permalink to this headline">¶</a></h5>
+<p>The first operand is a vector value to be written to memory. The second operand is a vector of pointers, pointing to where the value elements should be stored. It has the same underlying type as the value operand. The third operand is an alignment of the destination addresses. The fourth operand, mask, is a vector of boolean values. The types of the mask and the value operand must have the same number of vector elements.</p>
+</div>
+<div class="section" id="id543">
+<h5><a class="toc-backref" href="#id1397">Semantics:</a><a class="headerlink" href="#id543" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.masked.scatter</span></tt>‘ intrinsics is designed for writing selected vector elements to arbitrary memory addresses in a single IR operation. The operation may be conditional, when not all bits in the mask are switched on. It is useful for targets that support vector masked scatter and allows vectorizing basic blocks with data and control divergency. Other targets may support this intrinsic differently, for example by lowering it into a sequence of branches that guard scalar store operations.</p>
+<div class="highlight-python"><div class="highlight"><pre>;; This instruction unconditionaly stores data vector in multiple addresses
+call @llvm.masked.scatter.v8i32 (<8 x i32> %value, <8 x i32*> %ptrs, i32 4, <8 x i1> <true, true, .. true>)
+
+;; It is equivalent to a list of scalar stores
+%val0 = extractelement <8 x i32> %value, i32 0
+%val1 = extractelement <8 x i32> %value, i32 1
+..
+%val7 = extractelement <8 x i32> %value, i32 7
+%ptr0 = extractelement <8 x i32*> %ptrs, i32 0
+%ptr1 = extractelement <8 x i32*> %ptrs, i32 1
+..
+%ptr7 = extractelement <8 x i32*> %ptrs, i32 7
+;; Note: the order of the following stores is important when they overlap:
+store i32 %val0, i32* %ptr0, align 4
+store i32 %val1, i32* %ptr1, align 4
+..
+store i32 %val7, i32* %ptr7, align 4
+</pre></div>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="memory-use-markers">
+<h3><a class="toc-backref" href="#id1398">Memory Use Markers</a><a class="headerlink" href="#memory-use-markers" title="Permalink to this headline">¶</a></h3>
+<p>This class of intrinsics provides information about the lifetime of
+memory objects and ranges where variables are immutable.</p>
+<div class="section" id="llvm-lifetime-start-intrinsic">
+<span id="int-lifestart"></span><h4><a class="toc-backref" href="#id1399">‘<tt class="docutils literal"><span class="pre">llvm.lifetime.start</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-lifetime-start-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id544">
+<h5><a class="toc-backref" href="#id1400">Syntax:</a><a class="headerlink" href="#id544" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id545">
+<h5><a class="toc-backref" href="#id1401">Overview:</a><a class="headerlink" href="#id545" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.lifetime.start</span></tt>‘ intrinsic specifies the start of a memory
+object’s lifetime.</p>
+</div>
+<div class="section" id="id546">
+<h5><a class="toc-backref" href="#id1402">Arguments:</a><a class="headerlink" href="#id546" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a constant integer representing the size of the
+object, or -1 if it is variable sized. The second argument is a pointer
+to the object.</p>
+</div>
+<div class="section" id="id547">
+<h5><a class="toc-backref" href="#id1403">Semantics:</a><a class="headerlink" href="#id547" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic indicates that before this point in the code, the value
+of the memory pointed to by <tt class="docutils literal"><span class="pre">ptr</span></tt> is dead. This means that it is known
+to never be used and has an undefined value. A load from the pointer
+that precedes this intrinsic can be replaced with <tt class="docutils literal"><span class="pre">'undef'</span></tt>.</p>
+</div>
+</div>
+<div class="section" id="llvm-lifetime-end-intrinsic">
+<span id="int-lifeend"></span><h4><a class="toc-backref" href="#id1404">‘<tt class="docutils literal"><span class="pre">llvm.lifetime.end</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-lifetime-end-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id548">
+<h5><a class="toc-backref" href="#id1405">Syntax:</a><a class="headerlink" href="#id548" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id549">
+<h5><a class="toc-backref" href="#id1406">Overview:</a><a class="headerlink" href="#id549" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.lifetime.end</span></tt>‘ intrinsic specifies the end of a memory
+object’s lifetime.</p>
+</div>
+<div class="section" id="id550">
+<h5><a class="toc-backref" href="#id1407">Arguments:</a><a class="headerlink" href="#id550" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a constant integer representing the size of the
+object, or -1 if it is variable sized. The second argument is a pointer
+to the object.</p>
+</div>
+<div class="section" id="id551">
+<h5><a class="toc-backref" href="#id1408">Semantics:</a><a class="headerlink" href="#id551" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic indicates that after this point in the code, the value of
+the memory pointed to by <tt class="docutils literal"><span class="pre">ptr</span></tt> is dead. This means that it is known to
+never be used and has an undefined value. Any stores into the memory
+object following this intrinsic may be removed as dead.</p>
+</div>
+</div>
+<div class="section" id="llvm-invariant-start-intrinsic">
+<h4><a class="toc-backref" href="#id1409">‘<tt class="docutils literal"><span class="pre">llvm.invariant.start</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-invariant-start-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id552">
+<h5><a class="toc-backref" href="#id1410">Syntax:</a><a class="headerlink" href="#id552" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id553">
+<h5><a class="toc-backref" href="#id1411">Overview:</a><a class="headerlink" href="#id553" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.invariant.start</span></tt>‘ intrinsic specifies that the contents of
+a memory object will not change.</p>
+</div>
+<div class="section" id="id554">
+<h5><a class="toc-backref" href="#id1412">Arguments:</a><a class="headerlink" href="#id554" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a constant integer representing the size of the
+object, or -1 if it is variable sized. The second argument is a pointer
+to the object.</p>
+</div>
+<div class="section" id="id555">
+<h5><a class="toc-backref" href="#id1413">Semantics:</a><a class="headerlink" href="#id555" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic indicates that until an <tt class="docutils literal"><span class="pre">llvm.invariant.end</span></tt> that uses
+the return value, the referenced memory location is constant and
+unchanging.</p>
+</div>
+</div>
+<div class="section" id="llvm-invariant-end-intrinsic">
+<h4><a class="toc-backref" href="#id1414">‘<tt class="docutils literal"><span class="pre">llvm.invariant.end</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-invariant-end-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id556">
+<h5><a class="toc-backref" href="#id1415">Syntax:</a><a class="headerlink" href="#id556" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id557">
+<h5><a class="toc-backref" href="#id1416">Overview:</a><a class="headerlink" href="#id557" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.invariant.end</span></tt>‘ intrinsic specifies that the contents of a
+memory object are mutable.</p>
+</div>
+<div class="section" id="id558">
+<h5><a class="toc-backref" href="#id1417">Arguments:</a><a class="headerlink" href="#id558" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is the matching <tt class="docutils literal"><span class="pre">llvm.invariant.start</span></tt> intrinsic.
+The second argument is a constant integer representing the size of the
+object, or -1 if it is variable sized and the third argument is a
+pointer to the object.</p>
+</div>
+<div class="section" id="id559">
+<h5><a class="toc-backref" href="#id1418">Semantics:</a><a class="headerlink" href="#id559" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic indicates that the memory is mutable again.</p>
+</div>
+</div>
+</div>
+<div class="section" id="general-intrinsics">
+<h3><a class="toc-backref" href="#id1419">General Intrinsics</a><a class="headerlink" href="#general-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>This class of intrinsics is designed to be generic and has no specific
+purpose.</p>
+<div class="section" id="llvm-var-annotation-intrinsic">
+<h4><a class="toc-backref" href="#id1420">‘<tt class="docutils literal"><span class="pre">llvm.var.annotation</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-var-annotation-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id560">
+<h5><a class="toc-backref" href="#id1421">Syntax:</a><a class="headerlink" href="#id560" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id561">
+<h5><a class="toc-backref" href="#id1422">Overview:</a><a class="headerlink" href="#id561" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.var.annotation</span></tt>‘ intrinsic.</p>
+</div>
+<div class="section" id="id562">
+<h5><a class="toc-backref" href="#id1423">Arguments:</a><a class="headerlink" href="#id562" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a pointer to a value, the second is a pointer to a
+global string, the third is a pointer to a global string which is the
+source file name, and the last argument is the line number.</p>
+</div>
+<div class="section" id="id563">
+<h5><a class="toc-backref" href="#id1424">Semantics:</a><a class="headerlink" href="#id563" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic allows annotation of local variables with arbitrary
+strings. This can be useful for special purpose optimizations that want
+to look for these annotations. These have no other defined use; they are
+ignored by code generation and optimization.</p>
+</div>
+</div>
+<div class="section" id="llvm-ptr-annotation-intrinsic">
+<h4><a class="toc-backref" href="#id1425">‘<tt class="docutils literal"><span class="pre">llvm.ptr.annotation.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-ptr-annotation-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id564">
+<h5><a class="toc-backref" href="#id1426">Syntax:</a><a class="headerlink" href="#id564" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use ‘<tt class="docutils literal"><span class="pre">llvm.ptr.annotation</span></tt>‘ on a
+pointer to an integer of any width. <em>NOTE</em> you must specify an address space for
+the pointer. The identifier for the default address space is the integer
+‘<tt class="docutils literal"><span class="pre">0</span></tt>‘.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare i8* @llvm.ptr.annotation.p<address space>i8(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
+declare i16* @llvm.ptr.annotation.p<address space>i16(i16* <val>, i8* <str>, i8* <str>, i32 <int>)
+declare i32* @llvm.ptr.annotation.p<address space>i32(i32* <val>, i8* <str>, i8* <str>, i32 <int>)
+declare i64* @llvm.ptr.annotation.p<address space>i64(i64* <val>, i8* <str>, i8* <str>, i32 <int>)
+declare i256* @llvm.ptr.annotation.p<address space>i256(i256* <val>, i8* <str>, i8* <str>, i32 <int>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id565">
+<h5><a class="toc-backref" href="#id1427">Overview:</a><a class="headerlink" href="#id565" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.ptr.annotation</span></tt>‘ intrinsic.</p>
+</div>
+<div class="section" id="id566">
+<h5><a class="toc-backref" href="#id1428">Arguments:</a><a class="headerlink" href="#id566" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a pointer to an integer value of arbitrary bitwidth
+(result of some expression), the second is a pointer to a global string, the
+third is a pointer to a global string which is the source file name, and the
+last argument is the line number. It returns the value of the first argument.</p>
+</div>
+<div class="section" id="id567">
+<h5><a class="toc-backref" href="#id1429">Semantics:</a><a class="headerlink" href="#id567" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic allows annotation of a pointer to an integer with arbitrary
+strings. This can be useful for special purpose optimizations that want to look
+for these annotations. These have no other defined use; they are ignored by code
+generation and optimization.</p>
+</div>
+</div>
+<div class="section" id="llvm-annotation-intrinsic">
+<h4><a class="toc-backref" href="#id1430">‘<tt class="docutils literal"><span class="pre">llvm.annotation.*</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-annotation-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id568">
+<h5><a class="toc-backref" href="#id1431">Syntax:</a><a class="headerlink" href="#id568" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use ‘<tt class="docutils literal"><span class="pre">llvm.annotation</span></tt>‘ on
+any integer bit width.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
+declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
+declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
+declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
+declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id569">
+<h5><a class="toc-backref" href="#id1432">Overview:</a><a class="headerlink" href="#id569" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.annotation</span></tt>‘ intrinsic.</p>
+</div>
+<div class="section" id="id570">
+<h5><a class="toc-backref" href="#id1433">Arguments:</a><a class="headerlink" href="#id570" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is an integer value (result of some expression), the
+second is a pointer to a global string, the third is a pointer to a
+global string which is the source file name, and the last argument is
+the line number. It returns the value of the first argument.</p>
+</div>
+<div class="section" id="id571">
+<h5><a class="toc-backref" href="#id1434">Semantics:</a><a class="headerlink" href="#id571" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic allows annotations to be put on arbitrary expressions
+with arbitrary strings. This can be useful for special purpose
+optimizations that want to look for these annotations. These have no
+other defined use; they are ignored by code generation and optimization.</p>
+</div>
+</div>
+<div class="section" id="llvm-trap-intrinsic">
+<h4><a class="toc-backref" href="#id1435">‘<tt class="docutils literal"><span class="pre">llvm.trap</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-trap-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id572">
+<h5><a class="toc-backref" href="#id1436">Syntax:</a><a class="headerlink" href="#id572" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.trap() noreturn nounwind
+</pre></div>
+</div>
+</div>
+<div class="section" id="id573">
+<h5><a class="toc-backref" href="#id1437">Overview:</a><a class="headerlink" href="#id573" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.trap</span></tt>‘ intrinsic.</p>
+</div>
+<div class="section" id="id574">
+<h5><a class="toc-backref" href="#id1438">Arguments:</a><a class="headerlink" href="#id574" title="Permalink to this headline">¶</a></h5>
+<p>None.</p>
+</div>
+<div class="section" id="id575">
+<h5><a class="toc-backref" href="#id1439">Semantics:</a><a class="headerlink" href="#id575" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic is lowered to the target dependent trap instruction. If
+the target does not have a trap instruction, this intrinsic will be
+lowered to a call of the <tt class="docutils literal"><span class="pre">abort()</span></tt> function.</p>
+</div>
+</div>
+<div class="section" id="llvm-debugtrap-intrinsic">
+<h4><a class="toc-backref" href="#id1440">‘<tt class="docutils literal"><span class="pre">llvm.debugtrap</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-debugtrap-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id576">
+<h5><a class="toc-backref" href="#id1441">Syntax:</a><a class="headerlink" href="#id576" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.debugtrap() nounwind
+</pre></div>
+</div>
+</div>
+<div class="section" id="id577">
+<h5><a class="toc-backref" href="#id1442">Overview:</a><a class="headerlink" href="#id577" title="Permalink to this headline">¶</a></h5>
+<p>The ‘<tt class="docutils literal"><span class="pre">llvm.debugtrap</span></tt>‘ intrinsic.</p>
+</div>
+<div class="section" id="id578">
+<h5><a class="toc-backref" href="#id1443">Arguments:</a><a class="headerlink" href="#id578" title="Permalink to this headline">¶</a></h5>
+<p>None.</p>
+</div>
+<div class="section" id="id579">
+<h5><a class="toc-backref" href="#id1444">Semantics:</a><a class="headerlink" href="#id579" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic is lowered to code which is intended to cause an
+execution trap with the intention of requesting the attention of a
+debugger.</p>
+</div>
+</div>
+<div class="section" id="llvm-stackprotector-intrinsic">
+<h4><a class="toc-backref" href="#id1445">‘<tt class="docutils literal"><span class="pre">llvm.stackprotector</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-stackprotector-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id580">
+<h5><a class="toc-backref" href="#id1446">Syntax:</a><a class="headerlink" href="#id580" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id581">
+<h5><a class="toc-backref" href="#id1447">Overview:</a><a class="headerlink" href="#id581" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.stackprotector</span></tt> intrinsic takes the <tt class="docutils literal"><span class="pre">guard</span></tt> and stores it
+onto the stack at <tt class="docutils literal"><span class="pre">slot</span></tt>. The stack slot is adjusted to ensure that it
+is placed on the stack before local variables.</p>
+</div>
+<div class="section" id="id582">
+<h5><a class="toc-backref" href="#id1448">Arguments:</a><a class="headerlink" href="#id582" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.stackprotector</span></tt> intrinsic requires two pointer arguments.
+The first argument is the value loaded from the stack guard
+<tt class="docutils literal"><span class="pre">@__stack_chk_guard</span></tt>. The second variable is an <tt class="docutils literal"><span class="pre">alloca</span></tt> that has
+enough space to hold the value of the guard.</p>
+</div>
+<div class="section" id="id583">
+<h5><a class="toc-backref" href="#id1449">Semantics:</a><a class="headerlink" href="#id583" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic causes the prologue/epilogue inserter to force the position of
+the <tt class="docutils literal"><span class="pre">AllocaInst</span></tt> stack slot to be before local variables on the stack. This is
+to ensure that if a local variable on the stack is overwritten, it will destroy
+the value of the guard. When the function exits, the guard on the stack is
+checked against the original guard by <tt class="docutils literal"><span class="pre">llvm.stackprotectorcheck</span></tt>. If they are
+different, then <tt class="docutils literal"><span class="pre">llvm.stackprotectorcheck</span></tt> causes the program to abort by
+calling the <tt class="docutils literal"><span class="pre">__stack_chk_fail()</span></tt> function.</p>
+</div>
+</div>
+<div class="section" id="llvm-stackprotectorcheck-intrinsic">
+<h4><a class="toc-backref" href="#id1450">‘<tt class="docutils literal"><span class="pre">llvm.stackprotectorcheck</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-stackprotectorcheck-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id584">
+<h5><a class="toc-backref" href="#id1451">Syntax:</a><a class="headerlink" href="#id584" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.stackprotectorcheck(i8** <guard>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id585">
+<h5><a class="toc-backref" href="#id1452">Overview:</a><a class="headerlink" href="#id585" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.stackprotectorcheck</span></tt> intrinsic compares <tt class="docutils literal"><span class="pre">guard</span></tt> against an already
+created stack protector and if they are not equal calls the
+<tt class="docutils literal"><span class="pre">__stack_chk_fail()</span></tt> function.</p>
+</div>
+<div class="section" id="id586">
+<h5><a class="toc-backref" href="#id1453">Arguments:</a><a class="headerlink" href="#id586" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.stackprotectorcheck</span></tt> intrinsic requires one pointer argument, the
+the variable <tt class="docutils literal"><span class="pre">@__stack_chk_guard</span></tt>.</p>
+</div>
+<div class="section" id="id587">
+<h5><a class="toc-backref" href="#id1454">Semantics:</a><a class="headerlink" href="#id587" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic is provided to perform the stack protector check by comparing
+<tt class="docutils literal"><span class="pre">guard</span></tt> with the stack slot created by <tt class="docutils literal"><span class="pre">llvm.stackprotector</span></tt> and if the
+values do not match call the <tt class="docutils literal"><span class="pre">__stack_chk_fail()</span></tt> function.</p>
+<p>The reason to provide this as an IR level intrinsic instead of implementing it
+via other IR operations is that in order to perform this operation at the IR
+level without an intrinsic, one would need to create additional basic blocks to
+handle the success/failure cases. This makes it difficult to stop the stack
+protector check from disrupting sibling tail calls in Codegen. With this
+intrinsic, we are able to generate the stack protector basic blocks late in
+codegen after the tail call decision has occurred.</p>
+</div>
+</div>
+<div class="section" id="llvm-objectsize-intrinsic">
+<h4><a class="toc-backref" href="#id1455">‘<tt class="docutils literal"><span class="pre">llvm.objectsize</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-objectsize-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id588">
+<h5><a class="toc-backref" href="#id1456">Syntax:</a><a class="headerlink" href="#id588" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i32 @llvm.objectsize.i32(i8* <object>, i1 <min>)
+declare i64 @llvm.objectsize.i64(i8* <object>, i1 <min>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id589">
+<h5><a class="toc-backref" href="#id1457">Overview:</a><a class="headerlink" href="#id589" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.objectsize</span></tt> intrinsic is designed to provide information to
+the optimizers to determine at compile time whether a) an operation
+(like memcpy) will overflow a buffer that corresponds to an object, or
+b) that a runtime check for overflow isn’t necessary. An object in this
+context means an allocation of a specific class, structure, array, or
+other object.</p>
+</div>
+<div class="section" id="id590">
+<h5><a class="toc-backref" href="#id1458">Arguments:</a><a class="headerlink" href="#id590" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.objectsize</span></tt> intrinsic takes two arguments. The first
+argument is a pointer to or into the <tt class="docutils literal"><span class="pre">object</span></tt>. The second argument is
+a boolean and determines whether <tt class="docutils literal"><span class="pre">llvm.objectsize</span></tt> returns 0 (if true)
+or -1 (if false) when the object size is unknown. The second argument
+only accepts constants.</p>
+</div>
+<div class="section" id="id591">
+<h5><a class="toc-backref" href="#id1459">Semantics:</a><a class="headerlink" href="#id591" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.objectsize</span></tt> intrinsic is lowered to a constant representing
+the size of the object concerned. If the size cannot be determined at
+compile time, <tt class="docutils literal"><span class="pre">llvm.objectsize</span></tt> returns <tt class="docutils literal"><span class="pre">i32/i64</span> <span class="pre">-1</span> <span class="pre">or</span> <span class="pre">0</span></tt> (depending
+on the <tt class="docutils literal"><span class="pre">min</span></tt> argument).</p>
+</div>
+</div>
+<div class="section" id="llvm-expect-intrinsic">
+<h4><a class="toc-backref" href="#id1460">‘<tt class="docutils literal"><span class="pre">llvm.expect</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-expect-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id592">
+<h5><a class="toc-backref" href="#id1461">Syntax:</a><a class="headerlink" href="#id592" title="Permalink to this headline">¶</a></h5>
+<p>This is an overloaded intrinsic. You can use <tt class="docutils literal"><span class="pre">llvm.expect</span></tt> on any
+integer bit width.</p>
+<div class="highlight-python"><div class="highlight"><pre>declare i1 @llvm.expect.i1(i1 <val>, i1 <expected_val>)
+declare i32 @llvm.expect.i32(i32 <val>, i32 <expected_val>)
+declare i64 @llvm.expect.i64(i64 <val>, i64 <expected_val>)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id593">
+<h5><a class="toc-backref" href="#id1462">Overview:</a><a class="headerlink" href="#id593" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.expect</span></tt> intrinsic provides information about expected (the
+most probable) value of <tt class="docutils literal"><span class="pre">val</span></tt>, which can be used by optimizers.</p>
+</div>
+<div class="section" id="id594">
+<h5><a class="toc-backref" href="#id1463">Arguments:</a><a class="headerlink" href="#id594" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.expect</span></tt> intrinsic takes two arguments. The first argument is
+a value. The second argument is an expected value, this needs to be a
+constant value, variables are not allowed.</p>
+</div>
+<div class="section" id="id595">
+<h5><a class="toc-backref" href="#id1464">Semantics:</a><a class="headerlink" href="#id595" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic is lowered to the <tt class="docutils literal"><span class="pre">val</span></tt>.</p>
+</div>
+</div>
+<div class="section" id="llvm-assume-intrinsic">
+<span id="int-assume"></span><h4><a class="toc-backref" href="#id1465">‘<tt class="docutils literal"><span class="pre">llvm.assume</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-assume-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id596">
+<h5><a class="toc-backref" href="#id1466">Syntax:</a><a class="headerlink" href="#id596" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.assume(i1 %cond)
+</pre></div>
+</div>
+</div>
+<div class="section" id="id597">
+<h5><a class="toc-backref" href="#id1467">Overview:</a><a class="headerlink" href="#id597" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.assume</span></tt> allows the optimizer to assume that the provided
+condition is true. This information can then be used in simplifying other parts
+of the code.</p>
+</div>
+<div class="section" id="id598">
+<h5><a class="toc-backref" href="#id1468">Arguments:</a><a class="headerlink" href="#id598" title="Permalink to this headline">¶</a></h5>
+<p>The condition which the optimizer may assume is always true.</p>
+</div>
+<div class="section" id="id599">
+<h5><a class="toc-backref" href="#id1469">Semantics:</a><a class="headerlink" href="#id599" title="Permalink to this headline">¶</a></h5>
+<p>The intrinsic allows the optimizer to assume that the provided condition is
+always true whenever the control flow reaches the intrinsic call. No code is
+generated for this intrinsic, and instructions that contribute only to the
+provided condition are not used for code generation. If the condition is
+violated during execution, the behavior is undefined.</p>
+<p>Note that the optimizer might limit the transformations performed on values
+used by the <tt class="docutils literal"><span class="pre">llvm.assume</span></tt> intrinsic in order to preserve the instructions
+only used to form the intrinsic’s input argument. This might prove undesirable
+if the extra information provided by the <tt class="docutils literal"><span class="pre">llvm.assume</span></tt> intrinsic does not cause
+sufficient overall improvement in code quality. For this reason,
+<tt class="docutils literal"><span class="pre">llvm.assume</span></tt> should not be used to document basic mathematical invariants
+that the optimizer can otherwise deduce or facts that are of little use to the
+optimizer.</p>
+</div>
+</div>
+<div class="section" id="llvm-bitset-test-intrinsic">
+<span id="bitset-test"></span><h4><a class="toc-backref" href="#id1470">‘<tt class="docutils literal"><span class="pre">llvm.bitset.test</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-bitset-test-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id600">
+<h5><a class="toc-backref" href="#id1471">Syntax:</a><a class="headerlink" href="#id600" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare i1 @llvm.bitset.test(i8* %ptr, metadata %bitset) nounwind readnone
+</pre></div>
+</div>
+</div>
+<div class="section" id="id601">
+<h5><a class="toc-backref" href="#id1472">Arguments:</a><a class="headerlink" href="#id601" title="Permalink to this headline">¶</a></h5>
+<p>The first argument is a pointer to be tested. The second argument is a
+metadata string containing the name of a <a class="reference internal" href="BitSets.html"><em>bitset</em></a>.</p>
+</div>
+<div class="section" id="id602">
+<h5><a class="toc-backref" href="#id1473">Overview:</a><a class="headerlink" href="#id602" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.bitset.test</span></tt> intrinsic tests whether the given pointer is a
+member of the given bitset.</p>
+</div>
+</div>
+<div class="section" id="llvm-donothing-intrinsic">
+<h4><a class="toc-backref" href="#id1474">‘<tt class="docutils literal"><span class="pre">llvm.donothing</span></tt>‘ Intrinsic</a><a class="headerlink" href="#llvm-donothing-intrinsic" title="Permalink to this headline">¶</a></h4>
+<div class="section" id="id603">
+<h5><a class="toc-backref" href="#id1475">Syntax:</a><a class="headerlink" href="#id603" title="Permalink to this headline">¶</a></h5>
+<div class="highlight-python"><div class="highlight"><pre>declare void @llvm.donothing() nounwind readnone
+</pre></div>
+</div>
+</div>
+<div class="section" id="id604">
+<h5><a class="toc-backref" href="#id1476">Overview:</a><a class="headerlink" href="#id604" title="Permalink to this headline">¶</a></h5>
+<p>The <tt class="docutils literal"><span class="pre">llvm.donothing</span></tt> intrinsic doesn’t perform any operation. It’s one of only
+two intrinsics (besides <tt class="docutils literal"><span class="pre">llvm.experimental.patchpoint</span></tt>) that can be called
+with an invoke instruction.</p>
+</div>
+<div class="section" id="id605">
+<h5><a class="toc-backref" href="#id1477">Arguments:</a><a class="headerlink" href="#id605" title="Permalink to this headline">¶</a></h5>
+<p>None.</p>
+</div>
+<div class="section" id="id606">
+<h5><a class="toc-backref" href="#id1478">Semantics:</a><a class="headerlink" href="#id606" title="Permalink to this headline">¶</a></h5>
+<p>This intrinsic does nothing, and it’s removed by optimizers and ignored
+by codegen.</p>
+</div>
+</div>
+</div>
+<div class="section" id="stack-map-intrinsics">
+<h3><a class="toc-backref" href="#id1479">Stack Map Intrinsics</a><a class="headerlink" href="#stack-map-intrinsics" title="Permalink to this headline">¶</a></h3>
+<p>LLVM provides experimental intrinsics to support runtime patching
+mechanisms commonly desired in dynamic language JITs. These intrinsics
+are described in <a class="reference internal" href="StackMaps.html"><em>Stack maps and patch points in LLVM</em></a>.</p>
+</div>
+</div>
+</div>
+
+
+ </div>
+ </div>
+ <div class="clearer"></div>
+ </div>
+ <div class="related">
+ <h3>Navigation</h3>
+ <ul>
+ <li class="right" style="margin-right: 10px">
+ <a href="genindex.html" title="General Index"
+ >index</a></li>
+ <li class="right" >
+ <a href="CMake.html" title="Building LLVM with CMake"
+ >next</a> |</li>
+ <li class="right" >
+ <a href="index.html" title="Overview"
+ >previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="index.html">Documentation</a>»</li>
+
+ </ul>
+ </div>
+ <div class="footer">
+ © Copyright 2003-2016, LLVM Project.
+ Last updated on 2016-01-04.
+ Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.2.3.
+ </div>
+ </body>
+</html>
\ No newline at end of file
Added: www-releases/trunk/3.7.1/docs/Lexicon.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/3.7.1/docs/Lexicon.html?rev=257950&view=auto
==============================================================================
--- www-releases/trunk/3.7.1/docs/Lexicon.html (added)
+++ www-releases/trunk/3.7.1/docs/Lexicon.html Fri Jan 15 17:13:16 2016
@@ -0,0 +1,328 @@
+
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
+ "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+
+
+<html xmlns="http://www.w3.org/1999/xhtml">
+ <head>
+ <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+
+ <title>The LLVM Lexicon — LLVM 3.7 documentation</title>
+
+ <link rel="stylesheet" href="_static/llvm-theme.css" type="text/css" />
+ <link rel="stylesheet" href="_static/pygments.css" type="text/css" />
+
+ <script type="text/javascript">
+ var DOCUMENTATION_OPTIONS = {
+ URL_ROOT: './',
+ VERSION: '3.7',
+ COLLAPSE_INDEX: false,
+ FILE_SUFFIX: '.html',
+ HAS_SOURCE: true
+ };
+ </script>
+ <script type="text/javascript" src="_static/jquery.js"></script>
+ <script type="text/javascript" src="_static/underscore.js"></script>
+ <script type="text/javascript" src="_static/doctools.js"></script>
+ <link rel="top" title="LLVM 3.7 documentation" href="index.html" />
+ <link rel="next" title="How To Add Your Build Configuration To LLVM Buildbot Infrastructure" href="HowToAddABuilder.html" />
+ <link rel="prev" title="Frequently Asked Questions (FAQ)" href="FAQ.html" />
+<style type="text/css">
+ table.right { float: right; margin-left: 20px; }
+ table.right td { border: 1px solid #ccc; }
+</style>
+
+ </head>
+ <body>
+<div class="logo">
+ <a href="index.html">
+ <img src="_static/logo.png"
+ alt="LLVM Logo" width="250" height="88"/></a>
+</div>
+
+ <div class="related">
+ <h3>Navigation</h3>
+ <ul>
+ <li class="right" style="margin-right: 10px">
+ <a href="genindex.html" title="General Index"
+ accesskey="I">index</a></li>
+ <li class="right" >
+ <a href="HowToAddABuilder.html" title="How To Add Your Build Configuration To LLVM Buildbot Infrastructure"
+ accesskey="N">next</a> |</li>
+ <li class="right" >
+ <a href="FAQ.html" title="Frequently Asked Questions (FAQ)"
+ accesskey="P">previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="index.html">Documentation</a>»</li>
+
+ </ul>
+ </div>
+
+
+ <div class="document">
+ <div class="documentwrapper">
+ <div class="body">
+
+ <div class="section" id="the-llvm-lexicon">
+<h1>The LLVM Lexicon<a class="headerlink" href="#the-llvm-lexicon" title="Permalink to this headline">¶</a></h1>
+<div class="admonition note">
+<p class="first admonition-title">Note</p>
+<p class="last">This document is a work in progress!</p>
+</div>
+<div class="section" id="definitions">
+<h2>Definitions<a class="headerlink" href="#definitions" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="a">
+<h3>A<a class="headerlink" href="#a" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>ADCE</strong></dt>
+<dd>Aggressive Dead Code Elimination</dd>
+<dt><strong>AST</strong></dt>
+<dd><p class="first">Abstract Syntax Tree.</p>
+<p>Due to Clang’s influence (mostly the fact that parsing and semantic
+analysis are so intertwined for C and especially C++), the typical
+working definition of AST in the LLVM community is roughly “the
+compiler’s first complete symbolic (as opposed to textual)
+representation of an input program”.
+As such, an “AST” might be a more general graph instead of a “tree”
+(consider the symbolic representation for the type of a typical “linked
+list node”). This working definition is closer to what some authors
+call an “annotated abstract syntax tree”.</p>
+<p class="last">Consult your favorite compiler book or search engine for more details.</p>
+</dd>
+</dl>
+</div>
+<div class="section" id="b">
+<h3>B<a class="headerlink" href="#b" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils" id="lexicon-bb-vectorization">
+<dt><strong>BB Vectorization</strong></dt>
+<dd>Basic-Block Vectorization</dd>
+<dt><strong>BURS</strong></dt>
+<dd>Bottom Up Rewriting System — A method of instruction selection for code
+generation. An example is the <a class="reference external" href="http://www.program-transformation.org/Transform/BURG">BURG</a> tool.</dd>
+</dl>
+</div>
+<div class="section" id="c">
+<h3>C<a class="headerlink" href="#c" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>CFI</strong></dt>
+<dd>Call Frame Information. Used in DWARF debug info and in C++ unwind info
+to show how the function prolog lays out the stack frame.</dd>
+<dt><strong>CIE</strong></dt>
+<dd>Common Information Entry. A kind of CFI used to reduce the size of FDEs.
+The compiler creates a CIE which contains the information common across all
+the FDEs. Each FDE then points to its CIE.</dd>
+<dt><strong>CSE</strong></dt>
+<dd>Common Subexpression Elimination. An optimization that removes common
+subexpression compuation. For example <tt class="docutils literal"><span class="pre">(a+b)*(a+b)</span></tt> has two subexpressions
+that are the same: <tt class="docutils literal"><span class="pre">(a+b)</span></tt>. This optimization would perform the addition
+only once and then perform the multiply (but only if it’s computationally
+correct/safe).</dd>
+</dl>
+</div>
+<div class="section" id="d">
+<h3>D<a class="headerlink" href="#d" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>DAG</strong></dt>
+<dd>Directed Acyclic Graph</dd>
+</dl>
+<dl class="docutils" id="derived-pointers">
+<span id="derived-pointer"></span><dt><strong>Derived Pointer</strong></dt>
+<dd>A pointer to the interior of an object, such that a garbage collector is
+unable to use the pointer for reachability analysis. While a derived pointer
+is live, the corresponding object pointer must be kept in a root, otherwise
+the collector might free the referenced object. With copying collectors,
+derived pointers pose an additional hazard that they may be invalidated at
+any <a class="reference internal" href="#safe-point">safe point</a>. This term is used in opposition to <a class="reference internal" href="#object-pointer">object pointer</a>.</dd>
+<dt><strong>DSA</strong></dt>
+<dd>Data Structure Analysis</dd>
+<dt><strong>DSE</strong></dt>
+<dd>Dead Store Elimination</dd>
+</dl>
+</div>
+<div class="section" id="f">
+<h3>F<a class="headerlink" href="#f" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>FCA</strong></dt>
+<dd>First Class Aggregate</dd>
+<dt><strong>FDE</strong></dt>
+<dd>Frame Description Entry. A kind of CFI used to describe the stack frame of
+one function.</dd>
+</dl>
+</div>
+<div class="section" id="g">
+<h3>G<a class="headerlink" href="#g" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>GC</strong></dt>
+<dd>Garbage Collection. The practice of using reachability analysis instead of
+explicit memory management to reclaim unused memory.</dd>
+</dl>
+</div>
+<div class="section" id="h">
+<h3>H<a class="headerlink" href="#h" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils" id="heap">
+<dt><strong>Heap</strong></dt>
+<dd>In garbage collection, the region of memory which is managed using
+reachability analysis.</dd>
+</dl>
+</div>
+<div class="section" id="i">
+<h3>I<a class="headerlink" href="#i" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>IPA</strong></dt>
+<dd>Inter-Procedural Analysis. Refers to any variety of code analysis that
+occurs between procedures, functions or compilation units (modules).</dd>
+<dt><strong>IPO</strong></dt>
+<dd>Inter-Procedural Optimization. Refers to any variety of code optimization
+that occurs between procedures, functions or compilation units (modules).</dd>
+<dt><strong>ISel</strong></dt>
+<dd>Instruction Selection</dd>
+</dl>
+</div>
+<div class="section" id="l">
+<h3>L<a class="headerlink" href="#l" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>LCSSA</strong></dt>
+<dd>Loop-Closed Static Single Assignment Form</dd>
+<dt><strong>LGTM</strong></dt>
+<dd>“Looks Good To Me”. In a review thread, this indicates that the
+reviewer thinks that the patch is okay to commit.</dd>
+<dt><strong>LICM</strong></dt>
+<dd>Loop Invariant Code Motion</dd>
+<dt><strong>LSDA</strong></dt>
+<dd>Language Specific Data Area. C++ “zero cost” unwinding is built on top a
+generic unwinding mechanism. As the unwinder walks each frame, it calls
+a “personality” function to do language specific analysis. Each function’s
+FDE points to an optional LSDA which is passed to the personality function.
+For C++, the LSDA contain info about the type and location of catch
+statements in that function.</dd>
+<dt><strong>Load-VN</strong></dt>
+<dd>Load Value Numbering</dd>
+<dt><strong>LTO</strong></dt>
+<dd>Link-Time Optimization</dd>
+</dl>
+</div>
+<div class="section" id="m">
+<h3>M<a class="headerlink" href="#m" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>MC</strong></dt>
+<dd>Machine Code</dd>
+</dl>
+</div>
+<div class="section" id="n">
+<h3>N<a class="headerlink" href="#n" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>NFC</strong></dt>
+<dd>“No functional change”. Used in a commit message to indicate that a patch
+is a pure refactoring/cleanup.
+Usually used in the first line, so it is visible without opening the
+actual commit email.</dd>
+</dl>
+</div>
+<div class="section" id="o">
+<h3>O<a class="headerlink" href="#o" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils" id="object-pointers">
+<span id="object-pointer"></span><dt><strong>Object Pointer</strong></dt>
+<dd>A pointer to an object such that the garbage collector is able to trace
+references contained within the object. This term is used in opposition to
+<a class="reference internal" href="#derived-pointer">derived pointer</a>.</dd>
+</dl>
+</div>
+<div class="section" id="p">
+<h3>P<a class="headerlink" href="#p" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>PRE</strong></dt>
+<dd>Partial Redundancy Elimination</dd>
+</dl>
+</div>
+<div class="section" id="r">
+<h3>R<a class="headerlink" href="#r" title="Permalink to this headline">¶</a></h3>
+<p><strong>RAUW</strong></p>
+<blockquote>
+<div>Replace All Uses With. The functions <tt class="docutils literal"><span class="pre">User::replaceUsesOfWith()</span></tt>,
+<tt class="docutils literal"><span class="pre">Value::replaceAllUsesWith()</span></tt>, and
+<tt class="docutils literal"><span class="pre">Constant::replaceUsesOfWithOnConstant()</span></tt> implement the replacement of one
+Value with another by iterating over its def/use chain and fixing up all of
+the pointers to point to the new value. See
+also <a class="reference external" href="ProgrammersManual.html#iterating-over-def-use-use-def-chains">def/use chains</a>.</div></blockquote>
+<dl class="docutils">
+<dt><strong>Reassociation</strong></dt>
+<dd>Rearranging associative expressions to promote better redundancy elimination
+and other optimization. For example, changing <tt class="docutils literal"><span class="pre">(A+B-A)</span></tt> into <tt class="docutils literal"><span class="pre">(B+A-A)</span></tt>,
+permitting it to be optimized into <tt class="docutils literal"><span class="pre">(B+0)</span></tt> then <tt class="docutils literal"><span class="pre">(B)</span></tt>.</dd>
+</dl>
+<dl class="docutils" id="stack-roots">
+<span id="roots"></span><dt><strong>Root</strong></dt>
+<dd>In garbage collection, a pointer variable lying outside of the <a class="reference internal" href="#heap">heap</a> from
+which the collector begins its reachability analysis. In the context of code
+generation, “root” almost always refers to a “stack root” — a local or
+temporary variable within an executing function.</dd>
+<dt><strong>RPO</strong></dt>
+<dd>Reverse postorder</dd>
+</dl>
+</div>
+<div class="section" id="s">
+<h3>S<a class="headerlink" href="#s" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils" id="safe-point">
+<dt><strong>Safe Point</strong></dt>
+<dd>In garbage collection, it is necessary to identify <a class="reference internal" href="#stack-roots">stack roots</a> so that
+reachability analysis may proceed. It may be infeasible to provide this
+information for every instruction, so instead the information may is
+calculated only at designated safe points. With a copying collector,
+<a class="reference internal" href="#derived-pointers">derived pointers</a> must not be retained across safe points and <a class="reference internal" href="#object-pointers">object
+pointers</a> must be reloaded from stack roots.</dd>
+<dt><strong>SDISel</strong></dt>
+<dd>Selection DAG Instruction Selection.</dd>
+<dt><strong>SCC</strong></dt>
+<dd>Strongly Connected Component</dd>
+<dt><strong>SCCP</strong></dt>
+<dd>Sparse Conditional Constant Propagation</dd>
+<dt><strong>SLP</strong></dt>
+<dd>Superword-Level Parallelism, same as <a class="reference internal" href="#lexicon-bb-vectorization"><em>Basic-Block Vectorization</em></a>.</dd>
+<dt><strong>SRoA</strong></dt>
+<dd>Scalar Replacement of Aggregates</dd>
+<dt><strong>SSA</strong></dt>
+<dd>Static Single Assignment</dd>
+<dt><strong>Stack Map</strong></dt>
+<dd>In garbage collection, metadata emitted by the code generator which
+identifies <a class="reference internal" href="#roots">roots</a> within the stack frame of an executing function.</dd>
+</dl>
+</div>
+<div class="section" id="t">
+<h3>T<a class="headerlink" href="#t" title="Permalink to this headline">¶</a></h3>
+<dl class="docutils">
+<dt><strong>TBAA</strong></dt>
+<dd>Type-Based Alias Analysis</dd>
+</dl>
+</div>
+</div>
+</div>
+
+
+ </div>
+ </div>
+ <div class="clearer"></div>
+ </div>
+ <div class="related">
+ <h3>Navigation</h3>
+ <ul>
+ <li class="right" style="margin-right: 10px">
+ <a href="genindex.html" title="General Index"
+ >index</a></li>
+ <li class="right" >
+ <a href="HowToAddABuilder.html" title="How To Add Your Build Configuration To LLVM Buildbot Infrastructure"
+ >next</a> |</li>
+ <li class="right" >
+ <a href="FAQ.html" title="Frequently Asked Questions (FAQ)"
+ >previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="index.html">Documentation</a>»</li>
+
+ </ul>
+ </div>
+ <div class="footer">
+ © Copyright 2003-2016, LLVM Project.
+ Last updated on 2016-01-04.
+ Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.2.3.
+ </div>
+ </body>
+</html>
\ No newline at end of file
Added: www-releases/trunk/3.7.1/docs/LibFuzzer.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/3.7.1/docs/LibFuzzer.html?rev=257950&view=auto
==============================================================================
--- www-releases/trunk/3.7.1/docs/LibFuzzer.html (added)
+++ www-releases/trunk/3.7.1/docs/LibFuzzer.html Fri Jan 15 17:13:16 2016
@@ -0,0 +1,513 @@
+
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
+ "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
+
+
+<html xmlns="http://www.w3.org/1999/xhtml">
+ <head>
+ <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+
+ <title>LibFuzzer â a library for coverage-guided fuzz testing. — LLVM 3.7 documentation</title>
+
+ <link rel="stylesheet" href="_static/llvm-theme.css" type="text/css" />
+ <link rel="stylesheet" href="_static/pygments.css" type="text/css" />
+
+ <script type="text/javascript">
+ var DOCUMENTATION_OPTIONS = {
+ URL_ROOT: './',
+ VERSION: '3.7',
+ COLLAPSE_INDEX: false,
+ FILE_SUFFIX: '.html',
+ HAS_SOURCE: true
+ };
+ </script>
+ <script type="text/javascript" src="_static/jquery.js"></script>
+ <script type="text/javascript" src="_static/underscore.js"></script>
+ <script type="text/javascript" src="_static/doctools.js"></script>
+ <link rel="top" title="LLVM 3.7 documentation" href="index.html" />
+ <link rel="next" title="LLVM Alias Analysis Infrastructure" href="AliasAnalysis.html" />
+ <link rel="prev" title="LLVM Extensions" href="Extensions.html" />
+<style type="text/css">
+ table.right { float: right; margin-left: 20px; }
+ table.right td { border: 1px solid #ccc; }
+</style>
+
+ </head>
+ <body>
+<div class="logo">
+ <a href="index.html">
+ <img src="_static/logo.png"
+ alt="LLVM Logo" width="250" height="88"/></a>
+</div>
+
+ <div class="related">
+ <h3>Navigation</h3>
+ <ul>
+ <li class="right" style="margin-right: 10px">
+ <a href="genindex.html" title="General Index"
+ accesskey="I">index</a></li>
+ <li class="right" >
+ <a href="AliasAnalysis.html" title="LLVM Alias Analysis Infrastructure"
+ accesskey="N">next</a> |</li>
+ <li class="right" >
+ <a href="Extensions.html" title="LLVM Extensions"
+ accesskey="P">previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="index.html">Documentation</a>»</li>
+
+ </ul>
+ </div>
+
+
+ <div class="document">
+ <div class="documentwrapper">
+ <div class="body">
+
+ <div class="section" id="libfuzzer-a-library-for-coverage-guided-fuzz-testing">
+<h1>LibFuzzer – a library for coverage-guided fuzz testing.<a class="headerlink" href="#libfuzzer-a-library-for-coverage-guided-fuzz-testing" title="Permalink to this headline">¶</a></h1>
+<div class="contents local topic" id="contents">
+<ul class="simple">
+<li><a class="reference internal" href="#introduction" id="id3">Introduction</a></li>
+<li><a class="reference internal" href="#flags" id="id4">Flags</a></li>
+<li><a class="reference internal" href="#usage-examples" id="id5">Usage examples</a><ul>
+<li><a class="reference internal" href="#toy-example" id="id6">Toy example</a></li>
+<li><a class="reference internal" href="#pcre2" id="id7">PCRE2</a></li>
+<li><a class="reference internal" href="#heartbleed" id="id8">Heartbleed</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#advanced-features" id="id9">Advanced features</a><ul>
+<li><a class="reference internal" href="#tokens" id="id10">Tokens</a></li>
+<li><a class="reference internal" href="#afl-compatibility" id="id11">AFL compatibility</a></li>
+<li><a class="reference internal" href="#how-good-is-my-fuzzer" id="id12">How good is my fuzzer?</a></li>
+<li><a class="reference internal" href="#user-supplied-mutators" id="id13">User-supplied mutators</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#fuzzing-components-of-llvm" id="id14">Fuzzing components of LLVM</a><ul>
+<li><a class="reference internal" href="#clang-format-fuzzer" id="id15">clang-format-fuzzer</a></li>
+<li><a class="reference internal" href="#clang-fuzzer" id="id16">clang-fuzzer</a></li>
+<li><a class="reference internal" href="#buildbot" id="id17">Buildbot</a></li>
+<li><a class="reference internal" href="#pre-fuzzed-test-inputs-in-git" id="id18">Pre-fuzzed test inputs in git</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#faq" id="id19">FAQ</a><ul>
+<li><a class="reference internal" href="#q-why-fuzzer-does-not-use-any-of-the-llvm-support" id="id20">Q. Why Fuzzer does not use any of the LLVM support?</a></li>
+<li><a class="reference internal" href="#q-what-about-windows-then-the-fuzzer-contains-code-that-does-not-build-on-windows" id="id21">Q. What about Windows then? The Fuzzer contains code that does not build on Windows.</a></li>
+<li><a class="reference internal" href="#q-when-this-fuzzer-is-not-a-good-solution-for-a-problem" id="id22">Q. When this Fuzzer is not a good solution for a problem?</a></li>
+<li><a class="reference internal" href="#q-so-what-exactly-this-fuzzer-is-good-for" id="id23">Q. So, what exactly this Fuzzer is good for?</a></li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="section" id="introduction">
+<h2><a class="toc-backref" href="#id3">Introduction</a><a class="headerlink" href="#introduction" title="Permalink to this headline">¶</a></h2>
+<p>This library is intended primarily for in-process coverage-guided fuzz testing
+(fuzzing) of other libraries. The typical workflow looks like this:</p>
+<ul class="simple">
+<li>Build the Fuzzer library as a static archive (or just a set of .o files).
+Note that the Fuzzer contains the main() function.
+Preferably do <em>not</em> use sanitizers while building the Fuzzer.</li>
+<li>Build the library you are going to test with
+<cite>-fsanitize-coverage={bb,edge}[,indirect-calls,8bit-counters]</cite>
+and one of the sanitizers. We recommend to build the library in several
+different modes (e.g. asan, msan, lsan, ubsan, etc) and even using different
+optimizations options (e.g. -O0, -O1, -O2) to diversify testing.</li>
+<li>Build a test driver using the same options as the library.
+The test driver is a C/C++ file containing interesting calls to the library
+inside a single function <tt class="docutils literal"><span class="pre">extern</span> <span class="pre">"C"</span> <span class="pre">void</span> <span class="pre">LLVMFuzzerTestOneInput(const</span> <span class="pre">uint8_t</span> <span class="pre">*Data,</span> <span class="pre">size_t</span> <span class="pre">Size);</span></tt></li>
+<li>Link the Fuzzer, the library and the driver together into an executable
+using the same sanitizer options as for the library.</li>
+<li>Collect the initial corpus of inputs for the
+fuzzer (a directory with test inputs, one file per input).
+The better your inputs are the faster you will find something interesting.
+Also try to keep your inputs small, otherwise the Fuzzer will run too slow.
+By default, the Fuzzer limits the size of every input to 64 bytes
+(use <tt class="docutils literal"><span class="pre">-max_len=N</span></tt> to override).</li>
+<li>Run the fuzzer with the test corpus. As new interesting test cases are
+discovered they will be added to the corpus. If a bug is discovered by
+the sanitizer (asan, etc) it will be reported as usual and the reproducer
+will be written to disk.
+Each Fuzzer process is single-threaded (unless the library starts its own
+threads). You can run the Fuzzer on the same corpus in multiple processes
+in parallel.</li>
+</ul>
+<p>The Fuzzer is similar in concept to <a class="reference external" href="http://lcamtuf.coredump.cx/afl/">AFL</a>,
+but uses in-process Fuzzing, which is more fragile, more restrictive, but
+potentially much faster as it has no overhead for process start-up.
+It uses LLVM’s <a class="reference external" href="http://clang.llvm.org/docs/SanitizerCoverage.html">SanitizerCoverage</a> instrumentation to get in-process
+coverage-feedback</p>
+<p>The code resides in the LLVM repository, requires the fresh Clang compiler to build
+and is used to fuzz various parts of LLVM,
+but the Fuzzer itself does not (and should not) depend on any
+part of LLVM and can be used for other projects w/o requiring the rest of LLVM.</p>
+</div>
+<div class="section" id="flags">
+<h2><a class="toc-backref" href="#id4">Flags</a><a class="headerlink" href="#flags" title="Permalink to this headline">¶</a></h2>
+<p>The most important flags are:</p>
+<div class="highlight-python"><div class="highlight"><pre>seed 0 Random seed. If 0, seed is generated.
+runs -1 Number of individual test runs (-1 for infinite runs).
+max_len 64 Maximum length of the test input.
+cross_over 1 If 1, cross over inputs.
+mutate_depth 5 Apply this number of consecutive mutations to each input.
+timeout 1200 Timeout in seconds (if positive). If one unit runs more than this number of seconds the process will abort.
+help 0 Print help.
+save_minimized_corpus 0 If 1, the minimized corpus is saved into the first input directory
+jobs 0 Number of jobs to run. If jobs >= 1 we spawn this number of jobs in separate worker processes with stdout/stderr redirected to fuzz-JOB.log.
+workers 0 Number of simultaneous worker processes to run the jobs. If zero, "min(jobs,NumberOfCpuCores()/2)" is used.
+tokens 0 Use the file with tokens (one token per line) to fuzz a token based input language.
+apply_tokens 0 Read the given input file, substitute bytes with tokens and write the result to stdout.
+sync_command 0 Execute an external command "<sync_command> <test_corpus>" to synchronize the test corpus.
+sync_timeout 600 Minimum timeout between syncs.
+</pre></div>
+</div>
+<p>For the full list of flags run the fuzzer binary with <tt class="docutils literal"><span class="pre">-help=1</span></tt>.</p>
+</div>
+<div class="section" id="usage-examples">
+<h2><a class="toc-backref" href="#id5">Usage examples</a><a class="headerlink" href="#usage-examples" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="toy-example">
+<h3><a class="toc-backref" href="#id6">Toy example</a><a class="headerlink" href="#toy-example" title="Permalink to this headline">¶</a></h3>
+<p>A simple function that does something interesting if it receives the input “HI!”:</p>
+<div class="highlight-python"><div class="highlight"><pre>cat << EOF >> test_fuzzer.cc
+extern "C" void LLVMFuzzerTestOneInput(const unsigned char *data, unsigned long size) {
+ if (size > 0 && data[0] == 'H')
+ if (size > 1 && data[1] == 'I')
+ if (size > 2 && data[2] == '!')
+ __builtin_trap();
+}
+EOF
+# Get lib/Fuzzer. Assuming that you already have fresh clang in PATH.
+svn co http://llvm.org/svn/llvm-project/llvm/trunk/lib/Fuzzer
+# Build lib/Fuzzer files.
+clang -c -g -O2 -std=c++11 Fuzzer/*.cpp -IFuzzer
+# Build test_fuzzer.cc with asan and link against lib/Fuzzer.
+clang++ -fsanitize=address -fsanitize-coverage=edge test_fuzzer.cc Fuzzer*.o
+# Run the fuzzer with no corpus.
+./a.out
+</pre></div>
+</div>
+<p>You should get <tt class="docutils literal"><span class="pre">Illegal</span> <span class="pre">instruction</span> <span class="pre">(core</span> <span class="pre">dumped)</span></tt> pretty quickly.</p>
+</div>
+<div class="section" id="pcre2">
+<h3><a class="toc-backref" href="#id7">PCRE2</a><a class="headerlink" href="#pcre2" title="Permalink to this headline">¶</a></h3>
+<p>Here we show how to use lib/Fuzzer on something real, yet simple: <a class="reference external" href="http://www.pcre.org/">pcre2</a>:</p>
+<div class="highlight-python"><div class="highlight"><pre>COV_FLAGS=" -fsanitize-coverage=edge,indirect-calls,8bit-counters"
+# Get PCRE2
+svn co svn://vcs.exim.org/pcre2/code/trunk pcre
+# Get lib/Fuzzer. Assuming that you already have fresh clang in PATH.
+svn co http://llvm.org/svn/llvm-project/llvm/trunk/lib/Fuzzer
+# Build PCRE2 with AddressSanitizer and coverage.
+(cd pcre; ./autogen.sh; CC="clang -fsanitize=address $COV_FLAGS" ./configure --prefix=`pwd`/../inst && make -j && make install)
+# Build lib/Fuzzer files.
+clang -c -g -O2 -std=c++11 Fuzzer/*.cpp -IFuzzer
+# Build the actual function that does something interesting with PCRE2.
+cat << EOF > pcre_fuzzer.cc
+#include <string.h>
+#include "pcre2posix.h"
+extern "C" void LLVMFuzzerTestOneInput(const unsigned char *data, size_t size) {
+ if (size < 1) return;
+ char *str = new char[size+1];
+ memcpy(str, data, size);
+ str[size] = 0;
+ regex_t preg;
+ if (0 == regcomp(&preg, str, 0)) {
+ regexec(&preg, str, 0, 0, 0);
+ regfree(&preg);
+ }
+ delete [] str;
+}
+EOF
+clang++ -g -fsanitize=address $COV_FLAGS -c -std=c++11 -I inst/include/ pcre_fuzzer.cc
+# Link.
+clang++ -g -fsanitize=address -Wl,--whole-archive inst/lib/*.a -Wl,-no-whole-archive Fuzzer*.o pcre_fuzzer.o -o pcre_fuzzer
+</pre></div>
+</div>
+<p>This will give you a binary of the fuzzer, called <tt class="docutils literal"><span class="pre">pcre_fuzzer</span></tt>.
+Now, create a directory that will hold the test corpus:</p>
+<div class="highlight-python"><div class="highlight"><pre>mkdir -p CORPUS
+</pre></div>
+</div>
+<p>For simple input languages like regular expressions this is all you need.
+For more complicated inputs populate the directory with some input samples.
+Now run the fuzzer with the corpus dir as the only parameter:</p>
+<div class="highlight-python"><div class="highlight"><pre>./pcre_fuzzer ./CORPUS
+</pre></div>
+</div>
+<p>You will see output like this:</p>
+<div class="highlight-python"><div class="highlight"><pre>Seed: 1876794929
+#0 READ cov 0 bits 0 units 1 exec/s 0
+#1 pulse cov 3 bits 0 units 1 exec/s 0
+#1 INITED cov 3 bits 0 units 1 exec/s 0
+#2 pulse cov 208 bits 0 units 1 exec/s 0
+#2 NEW cov 208 bits 0 units 2 exec/s 0 L: 64
+#3 NEW cov 217 bits 0 units 3 exec/s 0 L: 63
+#4 pulse cov 217 bits 0 units 3 exec/s 0
+</pre></div>
+</div>
+<ul class="simple">
+<li>The <tt class="docutils literal"><span class="pre">Seed:</span></tt> line shows you the current random seed (you can change it with <tt class="docutils literal"><span class="pre">-seed=N</span></tt> flag).</li>
+<li>The <tt class="docutils literal"><span class="pre">READ</span></tt> line shows you how many input files were read (since you passed an empty dir there were inputs, but one dummy input was synthesised).</li>
+<li>The <tt class="docutils literal"><span class="pre">INITED</span></tt> line shows you that how many inputs will be fuzzed.</li>
+<li>The <tt class="docutils literal"><span class="pre">NEW</span></tt> lines appear with the fuzzer finds a new interesting input, which is saved to the CORPUS dir. If multiple corpus dirs are given, the first one is used.</li>
+<li>The <tt class="docutils literal"><span class="pre">pulse</span></tt> lines appear periodically to show the current status.</li>
+</ul>
+<p>Now, interrupt the fuzzer and run it again the same way. You will see:</p>
+<div class="highlight-python"><div class="highlight"><pre>Seed: 1879995378
+#0 READ cov 0 bits 0 units 564 exec/s 0
+#1 pulse cov 502 bits 0 units 564 exec/s 0
+...
+#512 pulse cov 2933 bits 0 units 564 exec/s 512
+#564 INITED cov 2991 bits 0 units 344 exec/s 564
+#1024 pulse cov 2991 bits 0 units 344 exec/s 1024
+#1455 NEW cov 2995 bits 0 units 345 exec/s 1455 L: 49
+</pre></div>
+</div>
+<p>This time you were running the fuzzer with a non-empty input corpus (564 items).
+As the first step, the fuzzer minimized the set to produce 344 interesting items (the <tt class="docutils literal"><span class="pre">INITED</span></tt> line)</p>
+<p>It is quite convenient to store test corpuses in git.
+As an example, here is a git repository with test inputs for the above PCRE2 fuzzer:</p>
+<div class="highlight-python"><div class="highlight"><pre>git clone https://github.com/kcc/fuzzing-with-sanitizers.git
+./pcre_fuzzer ./fuzzing-with-sanitizers/pcre2/C1/
+</pre></div>
+</div>
+<p>You may run <tt class="docutils literal"><span class="pre">N</span></tt> independent fuzzer jobs in parallel on <tt class="docutils literal"><span class="pre">M</span></tt> CPUs:</p>
+<div class="highlight-python"><div class="highlight"><pre>N=100; M=4; ./pcre_fuzzer ./CORPUS -jobs=$N -workers=$M
+</pre></div>
+</div>
+<p>By default (<tt class="docutils literal"><span class="pre">-reload=1</span></tt>) the fuzzer processes will periodically scan the CORPUS directory
+and reload any new tests. This way the test inputs found by one process will be picked up
+by all others.</p>
+<p>If <tt class="docutils literal"><span class="pre">-workers=$M</span></tt> is not supplied, <tt class="docutils literal"><span class="pre">min($N,NumberOfCpuCore/2)</span></tt> will be used.</p>
+</div>
+<div class="section" id="heartbleed">
+<h3><a class="toc-backref" href="#id8">Heartbleed</a><a class="headerlink" href="#heartbleed" title="Permalink to this headline">¶</a></h3>
+<p>Remember <a class="reference external" href="http://en.wikipedia.org/wiki/Heartbleed">Heartbleed</a>?
+As it was recently <a class="reference external" href="https://blog.hboeck.de/archives/868-How-Heartbleed-couldve-been-found.html">shown</a>,
+fuzzing with AddressSanitizer can find Heartbleed. Indeed, here are the step-by-step instructions
+to find Heartbleed with LibFuzzer:</p>
+<div class="highlight-python"><div class="highlight"><pre>wget https://www.openssl.org/source/openssl-1.0.1f.tar.gz
+tar xf openssl-1.0.1f.tar.gz
+COV_FLAGS="-fsanitize-coverage=edge,indirect-calls" # -fsanitize-coverage=8bit-counters
+(cd openssl-1.0.1f/ && ./config &&
+ make -j 32 CC="clang -g -fsanitize=address $COV_FLAGS")
+# Get and build LibFuzzer
+svn co http://llvm.org/svn/llvm-project/llvm/trunk/lib/Fuzzer
+clang -c -g -O2 -std=c++11 Fuzzer/*.cpp -IFuzzer
+# Get examples of key/pem files.
+git clone https://github.com/hannob/selftls
+cp selftls/server* . -v
+cat << EOF > handshake-fuzz.cc
+#include <openssl/ssl.h>
+#include <openssl/err.h>
+#include <assert.h>
+SSL_CTX *sctx;
+int Init() {
+ SSL_library_init();
+ SSL_load_error_strings();
+ ERR_load_BIO_strings();
+ OpenSSL_add_all_algorithms();
+ assert (sctx = SSL_CTX_new(TLSv1_method()));
+ assert (SSL_CTX_use_certificate_file(sctx, "server.pem", SSL_FILETYPE_PEM));
+ assert (SSL_CTX_use_PrivateKey_file(sctx, "server.key", SSL_FILETYPE_PEM));
+ return 0;
+}
+extern "C" void LLVMFuzzerTestOneInput(unsigned char *Data, size_t Size) {
+ static int unused = Init();
+ SSL *server = SSL_new(sctx);
+ BIO *sinbio = BIO_new(BIO_s_mem());
+ BIO *soutbio = BIO_new(BIO_s_mem());
+ SSL_set_bio(server, sinbio, soutbio);
+ SSL_set_accept_state(server);
+ BIO_write(sinbio, Data, Size);
+ SSL_do_handshake(server);
+ SSL_free(server);
+}
+EOF
+# Build the fuzzer.
+clang++ -g handshake-fuzz.cc -fsanitize=address \
+ openssl-1.0.1f/libssl.a openssl-1.0.1f/libcrypto.a Fuzzer*.o
+# Run 20 independent fuzzer jobs.
+./a.out -jobs=20 -workers=20
+</pre></div>
+</div>
+<p>Voila:</p>
+<div class="highlight-python"><div class="highlight"><pre>#1048576 pulse cov 3424 bits 0 units 9 exec/s 24385
+=================================================================
+==17488==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x629000004748 at pc 0x00000048c979 bp 0x7fffe3e864f0 sp 0x7fffe3e85ca8
+READ of size 60731 at 0x629000004748 thread T0
+ #0 0x48c978 in __asan_memcpy
+ #1 0x4db504 in tls1_process_heartbeat openssl-1.0.1f/ssl/t1_lib.c:2586:3
+ #2 0x580be3 in ssl3_read_bytes openssl-1.0.1f/ssl/s3_pkt.c:1092:4
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="advanced-features">
+<h2><a class="toc-backref" href="#id9">Advanced features</a><a class="headerlink" href="#advanced-features" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="tokens">
+<h3><a class="toc-backref" href="#id10">Tokens</a><a class="headerlink" href="#tokens" title="Permalink to this headline">¶</a></h3>
+<p>By default, the fuzzer is not aware of complexities of the input language
+and when fuzzing e.g. a C++ parser it will mostly stress the lexer.
+It is very hard for the fuzzer to come up with something like <tt class="docutils literal"><span class="pre">reinterpret_cast<int></span></tt>
+from a test corpus that doesn’t have it.
+See a detailed discussion of this topic at
+<a class="reference external" href="http://lcamtuf.blogspot.com/2015/01/afl-fuzz-making-up-grammar-with.html">http://lcamtuf.blogspot.com/2015/01/afl-fuzz-making-up-grammar-with.html</a>.</p>
+<p>lib/Fuzzer implements a simple technique that allows to fuzz input languages with
+long tokens. All you need is to prepare a text file containing up to 253 tokens, one token per line,
+and pass it to the fuzzer as <tt class="docutils literal"><span class="pre">-tokens=TOKENS_FILE.txt</span></tt>.
+Three implicit tokens are added: <tt class="docutils literal"><span class="pre">"</span> <span class="pre">"</span></tt>, <tt class="docutils literal"><span class="pre">"\t"</span></tt>, and <tt class="docutils literal"><span class="pre">"\n"</span></tt>.
+The fuzzer itself will still be mutating a string of bytes
+but before passing this input to the target library it will replace every byte <tt class="docutils literal"><span class="pre">b</span></tt> with the <tt class="docutils literal"><span class="pre">b</span></tt>-th token.
+If there are less than <tt class="docutils literal"><span class="pre">b</span></tt> tokens, a space will be added instead.</p>
+</div>
+<div class="section" id="afl-compatibility">
+<h3><a class="toc-backref" href="#id11">AFL compatibility</a><a class="headerlink" href="#afl-compatibility" title="Permalink to this headline">¶</a></h3>
+<p>LibFuzzer can be used in parallel with <a class="reference external" href="http://lcamtuf.coredump.cx/afl/">AFL</a> on the same test corpus.
+Both fuzzers expect the test corpus to reside in a directory, one file per input.
+You can run both fuzzers on the same corpus in parallel:</p>
+<div class="highlight-python"><div class="highlight"><pre>./afl-fuzz -i testcase_dir -o findings_dir /path/to/program -r @@
+./llvm-fuzz testcase_dir findings_dir # Will write new tests to testcase_dir
+</pre></div>
+</div>
+<p>Periodically restart both fuzzers so that they can use each other’s findings.</p>
+</div>
+<div class="section" id="how-good-is-my-fuzzer">
+<h3><a class="toc-backref" href="#id12">How good is my fuzzer?</a><a class="headerlink" href="#how-good-is-my-fuzzer" title="Permalink to this headline">¶</a></h3>
+<p>Once you implement your target function <tt class="docutils literal"><span class="pre">LLVMFuzzerTestOneInput</span></tt> and fuzz it to death,
+you will want to know whether the function or the corpus can be improved further.
+One easy to use metric is, of course, code coverage.
+You can get the coverage for your corpus like this:</p>
+<div class="highlight-python"><div class="highlight"><pre>ASAN_OPTIONS=coverage_pcs=1 ./fuzzer CORPUS_DIR -runs=0
+</pre></div>
+</div>
+<p>This will run all the tests in the CORPUS_DIR but will not generate any new tests
+and dump covered PCs to disk before exiting.
+Then you can subtract the set of covered PCs from the set of all instrumented PCs in the binary,
+see <a class="reference external" href="http://clang.llvm.org/docs/SanitizerCoverage.html">SanitizerCoverage</a> for details.</p>
+</div>
+<div class="section" id="user-supplied-mutators">
+<h3><a class="toc-backref" href="#id13">User-supplied mutators</a><a class="headerlink" href="#user-supplied-mutators" title="Permalink to this headline">¶</a></h3>
+<p>LibFuzzer allows to use custom (user-supplied) mutators,
+see <a class="reference external" href="https://github.com/llvm-mirror/llvm/blob/master/lib/Fuzzer/FuzzerInterface.h">FuzzerInterface.h</a></p>
+</div>
+</div>
+<div class="section" id="fuzzing-components-of-llvm">
+<h2><a class="toc-backref" href="#id14">Fuzzing components of LLVM</a><a class="headerlink" href="#fuzzing-components-of-llvm" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="clang-format-fuzzer">
+<h3><a class="toc-backref" href="#id15">clang-format-fuzzer</a><a class="headerlink" href="#clang-format-fuzzer" title="Permalink to this headline">¶</a></h3>
+<p>The inputs are random pieces of C++-like text.</p>
+<p>Build (make sure to use fresh clang as the host compiler):</p>
+<div class="highlight-python"><div class="highlight"><pre>cmake -GNinja -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DLLVM_USE_SANITIZER=Address -DLLVM_USE_SANITIZE_COVERAGE=YES -DCMAKE_BUILD_TYPE=Release /path/to/llvm
+ninja clang-format-fuzzer
+mkdir CORPUS_DIR
+./bin/clang-format-fuzzer CORPUS_DIR
+</pre></div>
+</div>
+<p>Optionally build other kinds of binaries (asan+Debug, msan, ubsan, etc).</p>
+<p>TODO: commit the pre-fuzzed corpus to svn (?).</p>
+<p>Tracking bug: <a class="reference external" href="https://llvm.org/bugs/show_bug.cgi?id=23052">https://llvm.org/bugs/show_bug.cgi?id=23052</a></p>
+</div>
+<div class="section" id="clang-fuzzer">
+<h3><a class="toc-backref" href="#id16">clang-fuzzer</a><a class="headerlink" href="#clang-fuzzer" title="Permalink to this headline">¶</a></h3>
+<p>The default behavior is very similar to <tt class="docutils literal"><span class="pre">clang-format-fuzzer</span></tt>.
+Clang can also be fuzzed with <a class="reference internal" href="#tokens">Tokens</a> using <tt class="docutils literal"><span class="pre">-tokens=$LLVM/lib/Fuzzer/cxx_fuzzer_tokens.txt</span></tt> option.</p>
+<p>Tracking bug: <a class="reference external" href="https://llvm.org/bugs/show_bug.cgi?id=23057">https://llvm.org/bugs/show_bug.cgi?id=23057</a></p>
+</div>
+<div class="section" id="buildbot">
+<h3><a class="toc-backref" href="#id17">Buildbot</a><a class="headerlink" href="#buildbot" title="Permalink to this headline">¶</a></h3>
+<p>We have a buildbot that runs the above fuzzers for LLVM components
+24/7/365 at <a class="reference external" href="http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fuzzer">http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fuzzer</a> .</p>
+</div>
+<div class="section" id="pre-fuzzed-test-inputs-in-git">
+<h3><a class="toc-backref" href="#id18">Pre-fuzzed test inputs in git</a><a class="headerlink" href="#pre-fuzzed-test-inputs-in-git" title="Permalink to this headline">¶</a></h3>
+<p>The buildbot occumulates large test corpuses over time.
+The corpuses are stored in git on github and can be used like this:</p>
+<div class="highlight-python"><div class="highlight"><pre>git clone https://github.com/kcc/fuzzing-with-sanitizers.git
+bin/clang-format-fuzzer fuzzing-with-sanitizers/llvm/clang-format/C1
+bin/clang-fuzzer fuzzing-with-sanitizers/llvm/clang/C1/
+bin/clang-fuzzer fuzzing-with-sanitizers/llvm/clang/TOK1 -tokens=$LLVM/llvm/lib/Fuzzer/cxx_fuzzer_tokens.txt
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="faq">
+<h2><a class="toc-backref" href="#id19">FAQ</a><a class="headerlink" href="#faq" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="q-why-fuzzer-does-not-use-any-of-the-llvm-support">
+<h3><a class="toc-backref" href="#id20">Q. Why Fuzzer does not use any of the LLVM support?</a><a class="headerlink" href="#q-why-fuzzer-does-not-use-any-of-the-llvm-support" title="Permalink to this headline">¶</a></h3>
+<p>There are two reasons.</p>
+<p>First, we want this library to be used outside of the LLVM w/o users having to
+build the rest of LLVM. This may sound unconvincing for many LLVM folks,
+but in practice the need for building the whole LLVM frightens many potential
+users – and we want more users to use this code.</p>
+<p>Second, there is a subtle technical reason not to rely on the rest of LLVM, or
+any other large body of code (maybe not even STL). When coverage instrumentation
+is enabled, it will also instrument the LLVM support code which will blow up the
+coverage set of the process (since the fuzzer is in-process). In other words, by
+using more external dependencies we will slow down the fuzzer while the main
+reason for it to exist is extreme speed.</p>
+</div>
+<div class="section" id="q-what-about-windows-then-the-fuzzer-contains-code-that-does-not-build-on-windows">
+<h3><a class="toc-backref" href="#id21">Q. What about Windows then? The Fuzzer contains code that does not build on Windows.</a><a class="headerlink" href="#q-what-about-windows-then-the-fuzzer-contains-code-that-does-not-build-on-windows" title="Permalink to this headline">¶</a></h3>
+<p>The sanitizer coverage support does not work on Windows either as of 01/2015.
+Once it’s there, we’ll need to re-implement OS-specific parts (I/O, signals).</p>
+</div>
+<div class="section" id="q-when-this-fuzzer-is-not-a-good-solution-for-a-problem">
+<h3><a class="toc-backref" href="#id22">Q. When this Fuzzer is not a good solution for a problem?</a><a class="headerlink" href="#q-when-this-fuzzer-is-not-a-good-solution-for-a-problem" title="Permalink to this headline">¶</a></h3>
+<ul class="simple">
+<li>If the test inputs are validated by the target library and the validator
+asserts/crashes on invalid inputs, the in-process fuzzer is not applicable
+(we could use fork() w/o exec, but it comes with extra overhead).</li>
+<li>Bugs in the target library may accumulate w/o being detected. E.g. a memory
+corruption that goes undetected at first and then leads to a crash while
+testing another input. This is why it is highly recommended to run this
+in-process fuzzer with all sanitizers to detect most bugs on the spot.</li>
+<li>It is harder to protect the in-process fuzzer from excessive memory
+consumption and infinite loops in the target library (still possible).</li>
+<li>The target library should not have significant global state that is not
+reset between the runs.</li>
+<li>Many interesting target libs are not designed in a way that supports
+the in-process fuzzer interface (e.g. require a file path instead of a
+byte array).</li>
+<li>If a single test run takes a considerable fraction of a second (or
+more) the speed benefit from the in-process fuzzer is negligible.</li>
+<li>If the target library runs persistent threads (that outlive
+execution of one test) the fuzzing results will be unreliable.</li>
+</ul>
+</div>
+<div class="section" id="q-so-what-exactly-this-fuzzer-is-good-for">
+<h3><a class="toc-backref" href="#id23">Q. So, what exactly this Fuzzer is good for?</a><a class="headerlink" href="#q-so-what-exactly-this-fuzzer-is-good-for" title="Permalink to this headline">¶</a></h3>
+<p>This Fuzzer might be a good choice for testing libraries that have relatively
+small inputs, each input takes < 1ms to run, and the library code is not expected
+to crash on invalid inputs.
+Examples: regular expression matchers, text or binary format parsers.</p>
+</div>
+</div>
+</div>
+
+
+ </div>
+ </div>
+ <div class="clearer"></div>
+ </div>
+ <div class="related">
+ <h3>Navigation</h3>
+ <ul>
+ <li class="right" style="margin-right: 10px">
+ <a href="genindex.html" title="General Index"
+ >index</a></li>
+ <li class="right" >
+ <a href="AliasAnalysis.html" title="LLVM Alias Analysis Infrastructure"
+ >next</a> |</li>
+ <li class="right" >
+ <a href="Extensions.html" title="LLVM Extensions"
+ >previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="index.html">Documentation</a>»</li>
+
+ </ul>
+ </div>
+ <div class="footer">
+ © Copyright 2003-2016, LLVM Project.
+ Last updated on 2016-01-04.
+ Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.2.3.
+ </div>
+ </body>
+</html>
\ No newline at end of file
More information about the llvm-commits
mailing list