[www-releases] r349965 - Add 7.0.1 docs for clang, llvm, and lld
Tom Stellard via llvm-commits
llvm-commits at lists.llvm.org
Fri Dec 21 13:53:04 PST 2018
Added: www-releases/trunk/7.0.1/docs/tutorial/OCamlLangImpl7.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/7.0.1/docs/tutorial/OCamlLangImpl7.html?rev=349965&view=auto
==============================================================================
--- www-releases/trunk/7.0.1/docs/tutorial/OCamlLangImpl7.html (added)
+++ www-releases/trunk/7.0.1/docs/tutorial/OCamlLangImpl7.html Fri Dec 21 13:53:02 2018
@@ -0,0 +1,1749 @@
+
+
+<!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>7. Kaleidoscope: Extending the Language: Mutable Variables — LLVM 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: '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 7 documentation" href="../index.html" />
+ <link rel="up" title="LLVM Tutorial: Table of Contents" href="index.html" />
+ <link rel="next" title="8. Kaleidoscope: Conclusion and other useful LLVM tidbits" href="OCamlLangImpl8.html" />
+ <link rel="prev" title="6. Kaleidoscope: Extending the Language: User-defined Operators" href="OCamlLangImpl6.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="OCamlLangImpl8.html" title="8. Kaleidoscope: Conclusion and other useful LLVM tidbits"
+ accesskey="N">next</a> |</li>
+ <li class="right" >
+ <a href="OCamlLangImpl6.html" title="6. Kaleidoscope: Extending the Language: User-defined Operators"
+ accesskey="P">previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="../index.html">Documentation</a>»</li>
+
+ <li><a href="index.html" accesskey="U">LLVM Tutorial: Table of Contents</a> »</li>
+ </ul>
+ </div>
+
+
+ <div class="document">
+ <div class="documentwrapper">
+ <div class="body">
+
+ <div class="section" id="kaleidoscope-extending-the-language-mutable-variables">
+<h1>7. Kaleidoscope: Extending the Language: Mutable Variables<a class="headerlink" href="#kaleidoscope-extending-the-language-mutable-variables" title="Permalink to this headline">¶</a></h1>
+<div class="contents local topic" id="contents">
+<ul class="simple">
+<li><a class="reference internal" href="#chapter-7-introduction" id="id2">Chapter 7 Introduction</a></li>
+<li><a class="reference internal" href="#why-is-this-a-hard-problem" id="id3">Why is this a hard problem?</a></li>
+<li><a class="reference internal" href="#memory-in-llvm" id="id4">Memory in LLVM</a></li>
+<li><a class="reference internal" href="#mutable-variables-in-kaleidoscope" id="id5">Mutable Variables in Kaleidoscope</a></li>
+<li><a class="reference internal" href="#adjusting-existing-variables-for-mutation" id="id6">Adjusting Existing Variables for Mutation</a></li>
+<li><a class="reference internal" href="#new-assignment-operator" id="id7">New Assignment Operator</a></li>
+<li><a class="reference internal" href="#user-defined-local-variables" id="id8">User-defined Local Variables</a></li>
+<li><a class="reference internal" href="#id1" id="id9">Full Code Listing</a></li>
+</ul>
+</div>
+<div class="section" id="chapter-7-introduction">
+<h2><a class="toc-backref" href="#id2">7.1. Chapter 7 Introduction</a><a class="headerlink" href="#chapter-7-introduction" title="Permalink to this headline">¶</a></h2>
+<p>Welcome to Chapter 7 of the “<a class="reference external" href="index.html">Implementing a language with
+LLVM</a>” tutorial. In chapters 1 through 6, we’ve built a
+very respectable, albeit simple, <a class="reference external" href="http://en.wikipedia.org/wiki/Functional_programming">functional programming
+language</a>. In our
+journey, we learned some parsing techniques, how to build and represent
+an AST, how to build LLVM IR, and how to optimize the resultant code as
+well as JIT compile it.</p>
+<p>While Kaleidoscope is interesting as a functional language, the fact
+that it is functional makes it “too easy” to generate LLVM IR for it. In
+particular, a functional language makes it very easy to build LLVM IR
+directly in <a class="reference external" href="http://en.wikipedia.org/wiki/Static_single_assignment_form">SSA
+form</a>.
+Since LLVM requires that the input code be in SSA form, this is a very
+nice property and it is often unclear to newcomers how to generate code
+for an imperative language with mutable variables.</p>
+<p>The short (and happy) summary of this chapter is that there is no need
+for your front-end to build SSA form: LLVM provides highly tuned and
+well tested support for this, though the way it works is a bit
+unexpected for some.</p>
+</div>
+<div class="section" id="why-is-this-a-hard-problem">
+<h2><a class="toc-backref" href="#id3">7.2. Why is this a hard problem?</a><a class="headerlink" href="#why-is-this-a-hard-problem" title="Permalink to this headline">¶</a></h2>
+<p>To understand why mutable variables cause complexities in SSA
+construction, consider this extremely simple C example:</p>
+<div class="highlight-c"><div class="highlight"><pre><span class="kt">int</span> <span class="n">G</span><span class="p">,</span> <span class="n">H</span><span class="p">;</span>
+<span class="kt">int</span> <span class="nf">test</span><span class="p">(</span><span class="kt">_Bool</span> <span class="n">Condition</span><span class="p">)</span> <span class="p">{</span>
+ <span class="kt">int</span> <span class="n">X</span><span class="p">;</span>
+ <span class="k">if</span> <span class="p">(</span><span class="n">Condition</span><span class="p">)</span>
+ <span class="n">X</span> <span class="o">=</span> <span class="n">G</span><span class="p">;</span>
+ <span class="k">else</span>
+ <span class="n">X</span> <span class="o">=</span> <span class="n">H</span><span class="p">;</span>
+ <span class="k">return</span> <span class="n">X</span><span class="p">;</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>In this case, we have the variable “X”, whose value depends on the path
+executed in the program. Because there are two different possible values
+for X before the return instruction, a PHI node is inserted to merge the
+two values. The LLVM IR that we want for this example looks like this:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="vg">@G</span> <span class="p">=</span> <span class="k">weak</span> <span class="k">global</span> <span class="k">i32</span> <span class="m">0</span> <span class="c">; type of @G is i32*</span>
+<span class="vg">@H</span> <span class="p">=</span> <span class="k">weak</span> <span class="k">global</span> <span class="k">i32</span> <span class="m">0</span> <span class="c">; type of @H is i32*</span>
+
+<span class="k">define</span> <span class="k">i32</span> <span class="vg">@test</span><span class="p">(</span><span class="k">i1</span> <span class="nv">%Condition</span><span class="p">)</span> <span class="p">{</span>
+<span class="nl">entry:</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%Condition</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%cond_true</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%cond_false</span>
+
+<span class="nl">cond_true:</span>
+ <span class="nv">%X.0</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@G</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%cond_next</span>
+
+<span class="nl">cond_false:</span>
+ <span class="nv">%X.1</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@H</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%cond_next</span>
+
+<span class="nl">cond_next:</span>
+ <span class="nv">%X.2</span> <span class="p">=</span> <span class="k">phi</span> <span class="k">i32</span> <span class="p">[</span> <span class="nv">%X.1</span><span class="p">,</span> <span class="nv">%cond_false</span> <span class="p">],</span> <span class="p">[</span> <span class="nv">%X.0</span><span class="p">,</span> <span class="nv">%cond_true</span> <span class="p">]</span>
+ <span class="k">ret</span> <span class="k">i32</span> <span class="nv">%X.2</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>In this example, the loads from the G and H global variables are
+explicit in the LLVM IR, and they live in the then/else branches of the
+if statement (cond_true/cond_false). In order to merge the incoming
+values, the X.2 phi node in the cond_next block selects the right value
+to use based on where control flow is coming from: if control flow comes
+from the cond_false block, X.2 gets the value of X.1. Alternatively, if
+control flow comes from cond_true, it gets the value of X.0. The intent
+of this chapter is not to explain the details of SSA form. For more
+information, see one of the many <a class="reference external" href="http://en.wikipedia.org/wiki/Static_single_assignment_form">online
+references</a>.</p>
+<p>The question for this article is “who places the phi nodes when lowering
+assignments to mutable variables?”. The issue here is that LLVM
+<em>requires</em> that its IR be in SSA form: there is no “non-ssa” mode for
+it. However, SSA construction requires non-trivial algorithms and data
+structures, so it is inconvenient and wasteful for every front-end to
+have to reproduce this logic.</p>
+</div>
+<div class="section" id="memory-in-llvm">
+<h2><a class="toc-backref" href="#id4">7.3. Memory in LLVM</a><a class="headerlink" href="#memory-in-llvm" title="Permalink to this headline">¶</a></h2>
+<p>The ‘trick’ here is that while LLVM does require all register values to
+be in SSA form, it does not require (or permit) memory objects to be in
+SSA form. In the example above, note that the loads from G and H are
+direct accesses to G and H: they are not renamed or versioned. This
+differs from some other compiler systems, which do try to version memory
+objects. In LLVM, instead of encoding dataflow analysis of memory into
+the LLVM IR, it is handled with <a class="reference external" href="../WritingAnLLVMPass.html">Analysis
+Passes</a> which are computed on demand.</p>
+<p>With this in mind, the high-level idea is that we want to make a stack
+variable (which lives in memory, because it is on the stack) for each
+mutable object in a function. To take advantage of this trick, we need
+to talk about how LLVM represents stack variables.</p>
+<p>In LLVM, all memory accesses are explicit with load/store instructions,
+and it is carefully designed not to have (or need) an “address-of”
+operator. Notice how the type of the @G/@H global variables is actually
+“i32*” even though the variable is defined as “i32”. What this means is
+that @G defines <em>space</em> for an i32 in the global data area, but its
+<em>name</em> actually refers to the address for that space. Stack variables
+work the same way, except that instead of being declared with global
+variable definitions, they are declared with the <a class="reference external" href="../LangRef.html#alloca-instruction">LLVM alloca
+instruction</a>:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">define</span> <span class="k">i32</span> <span class="vg">@example</span><span class="p">()</span> <span class="p">{</span>
+<span class="nl">entry:</span>
+ <span class="nv">%X</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i32</span> <span class="c">; type of %X is i32*.</span>
+ <span class="p">...</span>
+ <span class="nv">%tmp</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%X</span> <span class="c">; load the stack value %X from the stack.</span>
+ <span class="nv">%tmp2</span> <span class="p">=</span> <span class="k">add</span> <span class="k">i32</span> <span class="nv">%tmp</span><span class="p">,</span> <span class="m">1</span> <span class="c">; increment it</span>
+ <span class="k">store</span> <span class="k">i32</span> <span class="nv">%tmp2</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%X</span> <span class="c">; store it back</span>
+ <span class="p">...</span>
+</pre></div>
+</div>
+<p>This code shows an example of how you can declare and manipulate a stack
+variable in the LLVM IR. Stack memory allocated with the alloca
+instruction is fully general: you can pass the address of the stack slot
+to functions, you can store it in other variables, etc. In our example
+above, we could rewrite the example to use the alloca technique to avoid
+using a PHI node:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="vg">@G</span> <span class="p">=</span> <span class="k">weak</span> <span class="k">global</span> <span class="k">i32</span> <span class="m">0</span> <span class="c">; type of @G is i32*</span>
+<span class="vg">@H</span> <span class="p">=</span> <span class="k">weak</span> <span class="k">global</span> <span class="k">i32</span> <span class="m">0</span> <span class="c">; type of @H is i32*</span>
+
+<span class="k">define</span> <span class="k">i32</span> <span class="vg">@test</span><span class="p">(</span><span class="k">i1</span> <span class="nv">%Condition</span><span class="p">)</span> <span class="p">{</span>
+<span class="nl">entry:</span>
+ <span class="nv">%X</span> <span class="p">=</span> <span class="k">alloca</span> <span class="k">i32</span> <span class="c">; type of %X is i32*.</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%Condition</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%cond_true</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%cond_false</span>
+
+<span class="nl">cond_true:</span>
+ <span class="nv">%X.0</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@G</span>
+ <span class="k">store</span> <span class="k">i32</span> <span class="nv">%X.0</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%X</span> <span class="c">; Update X</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%cond_next</span>
+
+<span class="nl">cond_false:</span>
+ <span class="nv">%X.1</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">*</span> <span class="vg">@H</span>
+ <span class="k">store</span> <span class="k">i32</span> <span class="nv">%X.1</span><span class="p">,</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%X</span> <span class="c">; Update X</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%cond_next</span>
+
+<span class="nl">cond_next:</span>
+ <span class="nv">%X.2</span> <span class="p">=</span> <span class="k">load</span> <span class="k">i32</span><span class="p">*</span> <span class="nv">%X</span> <span class="c">; Read X</span>
+ <span class="k">ret</span> <span class="k">i32</span> <span class="nv">%X.2</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>With this, we have discovered a way to handle arbitrary mutable
+variables without the need to create Phi nodes at all:</p>
+<ol class="arabic simple">
+<li>Each mutable variable becomes a stack allocation.</li>
+<li>Each read of the variable becomes a load from the stack.</li>
+<li>Each update of the variable becomes a store to the stack.</li>
+<li>Taking the address of a variable just uses the stack address
+directly.</li>
+</ol>
+<p>While this solution has solved our immediate problem, it introduced
+another one: we have now apparently introduced a lot of stack traffic
+for very simple and common operations, a major performance problem.
+Fortunately for us, the LLVM optimizer has a highly-tuned optimization
+pass named “mem2reg” that handles this case, promoting allocas like this
+into SSA registers, inserting Phi nodes as appropriate. If you run this
+example through the pass, for example, you’ll get:</p>
+<div class="highlight-bash"><div class="highlight"><pre><span class="nv">$ </span>llvm-as < example.ll | opt -mem2reg | llvm-dis
+ at G <span class="o">=</span> weak global i32 0
+ at H <span class="o">=</span> weak global i32 0
+
+define i32 @test<span class="o">(</span>i1 %Condition<span class="o">)</span> <span class="o">{</span>
+entry:
+ br i1 %Condition, label %cond_true, label %cond_false
+
+cond_true:
+ %X.0 <span class="o">=</span> load i32* @G
+ br label %cond_next
+
+cond_false:
+ %X.1 <span class="o">=</span> load i32* @H
+ br label %cond_next
+
+cond_next:
+ %X.01 <span class="o">=</span> phi i32 <span class="o">[</span> %X.1, %cond_false <span class="o">]</span>, <span class="o">[</span> %X.0, %cond_true <span class="o">]</span>
+ ret i32 %X.01
+<span class="o">}</span>
+</pre></div>
+</div>
+<p>The mem2reg pass implements the standard “iterated dominance frontier”
+algorithm for constructing SSA form and has a number of optimizations
+that speed up (very common) degenerate cases. The mem2reg optimization
+pass is the answer to dealing with mutable variables, and we highly
+recommend that you depend on it. Note that mem2reg only works on
+variables in certain circumstances:</p>
+<ol class="arabic simple">
+<li>mem2reg is alloca-driven: it looks for allocas and if it can handle
+them, it promotes them. It does not apply to global variables or heap
+allocations.</li>
+<li>mem2reg only looks for alloca instructions in the entry block of the
+function. Being in the entry block guarantees that the alloca is only
+executed once, which makes analysis simpler.</li>
+<li>mem2reg only promotes allocas whose uses are direct loads and stores.
+If the address of the stack object is passed to a function, or if any
+funny pointer arithmetic is involved, the alloca will not be
+promoted.</li>
+<li>mem2reg only works on allocas of <a class="reference external" href="../LangRef.html#first-class-types">first
+class</a> values (such as pointers,
+scalars and vectors), and only if the array size of the allocation is
+1 (or missing in the .ll file). mem2reg is not capable of promoting
+structs or arrays to registers. Note that the “sroa” pass is
+more powerful and can promote structs, “unions”, and arrays in many
+cases.</li>
+</ol>
+<p>All of these properties are easy to satisfy for most imperative
+languages, and we’ll illustrate it below with Kaleidoscope. The final
+question you may be asking is: should I bother with this nonsense for my
+front-end? Wouldn’t it be better if I just did SSA construction
+directly, avoiding use of the mem2reg optimization pass? In short, we
+strongly recommend that you use this technique for building SSA form,
+unless there is an extremely good reason not to. Using this technique
+is:</p>
+<ul class="simple">
+<li>Proven and well tested: clang uses this technique
+for local mutable variables. As such, the most common clients of LLVM
+are using this to handle a bulk of their variables. You can be sure
+that bugs are found fast and fixed early.</li>
+<li>Extremely Fast: mem2reg has a number of special cases that make it
+fast in common cases as well as fully general. For example, it has
+fast-paths for variables that are only used in a single block,
+variables that only have one assignment point, good heuristics to
+avoid insertion of unneeded phi nodes, etc.</li>
+<li>Needed for debug info generation: <a class="reference external" href="../SourceLevelDebugging.html">Debug information in
+LLVM</a> relies on having the address of
+the variable exposed so that debug info can be attached to it. This
+technique dovetails very naturally with this style of debug info.</li>
+</ul>
+<p>If nothing else, this makes it much easier to get your front-end up and
+running, and is very simple to implement. Lets extend Kaleidoscope with
+mutable variables now!</p>
+</div>
+<div class="section" id="mutable-variables-in-kaleidoscope">
+<h2><a class="toc-backref" href="#id5">7.4. Mutable Variables in Kaleidoscope</a><a class="headerlink" href="#mutable-variables-in-kaleidoscope" title="Permalink to this headline">¶</a></h2>
+<p>Now that we know the sort of problem we want to tackle, lets see what
+this looks like in the context of our little Kaleidoscope language.
+We’re going to add two features:</p>
+<ol class="arabic simple">
+<li>The ability to mutate variables with the ‘=’ operator.</li>
+<li>The ability to define new variables.</li>
+</ol>
+<p>While the first item is really what this is about, we only have
+variables for incoming arguments as well as for induction variables, and
+redefining those only goes so far :). Also, the ability to define new
+variables is a useful thing regardless of whether you will be mutating
+them. Here’s a motivating example that shows how we could use these:</p>
+<div class="highlight-python"><pre># Define ':' for sequencing: as a low-precedence operator that ignores operands
+# and just returns the RHS.
+def binary : 1 (x y) y;
+
+# Recursive fib, we could do this before.
+def fib(x)
+ if (x < 3) then
+ 1
+ else
+ fib(x-1)+fib(x-2);
+
+# Iterative fib.
+def fibi(x)
+ var a = 1, b = 1, c in
+ (for i = 3, i < x in
+ c = a + b :
+ a = b :
+ b = c) :
+ b;
+
+# Call it.
+fibi(10);</pre>
+</div>
+<p>In order to mutate variables, we have to change our existing variables
+to use the “alloca trick”. Once we have that, we’ll add our new
+operator, then extend Kaleidoscope to support new variable definitions.</p>
+</div>
+<div class="section" id="adjusting-existing-variables-for-mutation">
+<h2><a class="toc-backref" href="#id6">7.5. Adjusting Existing Variables for Mutation</a><a class="headerlink" href="#adjusting-existing-variables-for-mutation" title="Permalink to this headline">¶</a></h2>
+<p>The symbol table in Kaleidoscope is managed at code generation time by
+the ‘<tt class="docutils literal"><span class="pre">named_values</span></tt>‘ map. This map currently keeps track of the LLVM
+“Value*” that holds the double value for the named variable. In order
+to support mutation, we need to change this slightly, so that it
+<tt class="docutils literal"><span class="pre">named_values</span></tt> holds the <em>memory location</em> of the variable in
+question. Note that this change is a refactoring: it changes the
+structure of the code, but does not (by itself) change the behavior of
+the compiler. All of these changes are isolated in the Kaleidoscope code
+generator.</p>
+<p>At this point in Kaleidoscope’s development, it only supports variables
+for two things: incoming arguments to functions and the induction
+variable of ‘for’ loops. For consistency, we’ll allow mutation of these
+variables in addition to other user-defined variables. This means that
+these will both need memory locations.</p>
+<p>To start our transformation of Kaleidoscope, we’ll change the
+<tt class="docutils literal"><span class="pre">named_values</span></tt> map so that it maps to AllocaInst* instead of Value*.
+Once we do this, the C++ compiler will tell us what parts of the code we
+need to update:</p>
+<p><strong>Note:</strong> the ocaml bindings currently model both <tt class="docutils literal"><span class="pre">Value*</span></tt>‘s and
+<tt class="docutils literal"><span class="pre">AllocInst*</span></tt>‘s as <tt class="docutils literal"><span class="pre">Llvm.llvalue</span></tt>‘s, but this may change in the future
+to be more type safe.</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="k">let</span> <span class="n">named_values</span><span class="o">:(</span><span class="kt">string</span><span class="o">,</span> <span class="n">llvalue</span><span class="o">)</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">t</span> <span class="o">=</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">create</span> <span class="mi">10</span>
+</pre></div>
+</div>
+<p>Also, since we will need to create these alloca’s, we’ll use a helper
+function that ensures that the allocas are created in the entry block of
+the function:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="c">(* Create an alloca instruction in the entry block of the function. This</span>
+<span class="c"> * is used for mutable variables etc. *)</span>
+<span class="k">let</span> <span class="n">create_entry_block_alloca</span> <span class="n">the_function</span> <span class="n">var_name</span> <span class="o">=</span>
+ <span class="k">let</span> <span class="n">builder</span> <span class="o">=</span> <span class="n">builder_at</span> <span class="o">(</span><span class="n">instr_begin</span> <span class="o">(</span><span class="n">entry_block</span> <span class="n">the_function</span><span class="o">))</span> <span class="k">in</span>
+ <span class="n">build_alloca</span> <span class="n">double_type</span> <span class="n">var_name</span> <span class="n">builder</span>
+</pre></div>
+</div>
+<p>This funny looking code creates an <tt class="docutils literal"><span class="pre">Llvm.llbuilder</span></tt> object that is
+pointing at the first instruction of the entry block. It then creates an
+alloca with the expected name and returns it. Because all values in
+Kaleidoscope are doubles, there is no need to pass in a type to use.</p>
+<p>With this in place, the first functionality change we want to make is to
+variable references. In our new scheme, variables live on the stack, so
+code generating a reference to them actually needs to produce a load
+from the stack slot:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="k">let</span> <span class="k">rec</span> <span class="n">codegen_expr</span> <span class="o">=</span> <span class="k">function</span>
+ <span class="o">...</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Variable</span> <span class="n">name</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">v</span> <span class="o">=</span> <span class="k">try</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">named_values</span> <span class="n">name</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Not_found</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"unknown variable name"</span><span class="o">)</span>
+ <span class="k">in</span>
+ <span class="c">(* Load the value. *)</span>
+ <span class="n">build_load</span> <span class="n">v</span> <span class="n">name</span> <span class="n">builder</span>
+</pre></div>
+</div>
+<p>As you can see, this is pretty straightforward. Now we need to update
+the things that define the variables to set up the alloca. We’ll start
+with <tt class="docutils literal"><span class="pre">codegen_expr</span> <span class="pre">Ast.For</span> <span class="pre">...</span></tt> (see the <a class="reference external" href="#id1">full code listing</a>
+for the unabridged code):</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">For</span> <span class="o">(</span><span class="n">var_name</span><span class="o">,</span> <span class="n">start</span><span class="o">,</span> <span class="n">end_</span><span class="o">,</span> <span class="n">step</span><span class="o">,</span> <span class="n">body</span><span class="o">)</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">the_function</span> <span class="o">=</span> <span class="n">block_parent</span> <span class="o">(</span><span class="n">insertion_block</span> <span class="n">builder</span><span class="o">)</span> <span class="k">in</span>
+
+ <span class="c">(* Create an alloca for the variable in the entry block. *)</span>
+ <span class="k">let</span> <span class="n">alloca</span> <span class="o">=</span> <span class="n">create_entry_block_alloca</span> <span class="n">the_function</span> <span class="n">var_name</span> <span class="k">in</span>
+
+ <span class="c">(* Emit the start code first, without 'variable' in scope. *)</span>
+ <span class="k">let</span> <span class="n">start_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">start</span> <span class="k">in</span>
+
+ <span class="c">(* Store the value into the alloca. *)</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">start_val</span> <span class="n">alloca</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="o">...</span>
+
+ <span class="c">(* Within the loop, the variable is defined equal to the PHI node. If it</span>
+<span class="c"> * shadows an existing variable, we have to restore it, so save it</span>
+<span class="c"> * now. *)</span>
+ <span class="k">let</span> <span class="n">old_val</span> <span class="o">=</span>
+ <span class="k">try</span> <span class="nc">Some</span> <span class="o">(</span><span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">named_values</span> <span class="n">var_name</span><span class="o">)</span> <span class="k">with</span> <span class="nc">Not_found</span> <span class="o">-></span> <span class="nc">None</span>
+ <span class="k">in</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">alloca</span><span class="o">;</span>
+
+ <span class="o">...</span>
+
+ <span class="c">(* Compute the end condition. *)</span>
+ <span class="k">let</span> <span class="n">end_cond</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">end_</span> <span class="k">in</span>
+
+ <span class="c">(* Reload, increment, and restore the alloca. This handles the case where</span>
+<span class="c"> * the body of the loop mutates the variable. *)</span>
+ <span class="k">let</span> <span class="n">cur_var</span> <span class="o">=</span> <span class="n">build_load</span> <span class="n">alloca</span> <span class="n">var_name</span> <span class="n">builder</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">next_var</span> <span class="o">=</span> <span class="n">build_add</span> <span class="n">cur_var</span> <span class="n">step_val</span> <span class="s2">"nextvar"</span> <span class="n">builder</span> <span class="k">in</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">next_var</span> <span class="n">alloca</span> <span class="n">builder</span><span class="o">);</span>
+ <span class="o">...</span>
+</pre></div>
+</div>
+<p>This code is virtually identical to the code <a class="reference external" href="OCamlLangImpl5.html#code-generation-for-the-for-loop">before we allowed mutable
+variables</a>. The big difference is that
+we no longer have to construct a PHI node, and we use load/store to
+access the variable as needed.</p>
+<p>To support mutable argument variables, we need to also make allocas for
+them. The code for this is also pretty simple:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="c">(* Create an alloca for each argument and register the argument in the symbol</span>
+<span class="c"> * table so that references to it will succeed. *)</span>
+<span class="k">let</span> <span class="n">create_argument_allocas</span> <span class="n">the_function</span> <span class="n">proto</span> <span class="o">=</span>
+ <span class="k">let</span> <span class="n">args</span> <span class="o">=</span> <span class="k">match</span> <span class="n">proto</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Prototype</span> <span class="o">(_,</span> <span class="n">args</span><span class="o">)</span> <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">BinOpPrototype</span> <span class="o">(_,</span> <span class="n">args</span><span class="o">,</span> <span class="o">_)</span> <span class="o">-></span> <span class="n">args</span>
+ <span class="k">in</span>
+ <span class="nn">Array</span><span class="p">.</span><span class="n">iteri</span> <span class="o">(</span><span class="k">fun</span> <span class="n">i</span> <span class="n">ai</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">var_name</span> <span class="o">=</span> <span class="n">args</span><span class="o">.(</span><span class="n">i</span><span class="o">)</span> <span class="k">in</span>
+ <span class="c">(* Create an alloca for this variable. *)</span>
+ <span class="k">let</span> <span class="n">alloca</span> <span class="o">=</span> <span class="n">create_entry_block_alloca</span> <span class="n">the_function</span> <span class="n">var_name</span> <span class="k">in</span>
+
+ <span class="c">(* Store the initial value into the alloca. *)</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">ai</span> <span class="n">alloca</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Add arguments to variable symbol table. *)</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">alloca</span><span class="o">;</span>
+ <span class="o">)</span> <span class="o">(</span><span class="n">params</span> <span class="n">the_function</span><span class="o">)</span>
+</pre></div>
+</div>
+<p>For each argument, we make an alloca, store the input value to the
+function into the alloca, and register the alloca as the memory location
+for the argument. This method gets invoked by <tt class="docutils literal"><span class="pre">Codegen.codegen_func</span></tt>
+right after it sets up the entry block for the function.</p>
+<p>The final missing piece is adding the mem2reg pass, which allows us to
+get good codegen once again:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="k">let</span> <span class="n">main</span> <span class="bp">()</span> <span class="o">=</span>
+ <span class="o">...</span>
+ <span class="k">let</span> <span class="n">the_fpm</span> <span class="o">=</span> <span class="nn">PassManager</span><span class="p">.</span><span class="n">create_function</span> <span class="nn">Codegen</span><span class="p">.</span><span class="n">the_module</span> <span class="k">in</span>
+
+ <span class="c">(* Set up the optimizer pipeline. Start with registering info about how the</span>
+<span class="c"> * target lays out data structures. *)</span>
+ <span class="nn">DataLayout</span><span class="p">.</span><span class="n">add</span> <span class="o">(</span><span class="nn">ExecutionEngine</span><span class="p">.</span><span class="n">target_data</span> <span class="n">the_execution_engine</span><span class="o">)</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="c">(* Promote allocas to registers. *)</span>
+ <span class="n">add_memory_to_register_promotion</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="c">(* Do simple "peephole" optimizations and bit-twiddling optzn. *)</span>
+ <span class="n">add_instruction_combining</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="c">(* reassociate expressions. *)</span>
+ <span class="n">add_reassociation</span> <span class="n">the_fpm</span><span class="o">;</span>
+</pre></div>
+</div>
+<p>It is interesting to see what the code looks like before and after the
+mem2reg optimization runs. For example, this is the before/after code
+for our recursive fib function. Before the optimization:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">define</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%x</span><span class="p">)</span> <span class="p">{</span>
+<span class="nl">entry:</span>
+ <span class="nv">%x1</span> <span class="p">=</span> <span class="k">alloca</span> <span class="kt">double</span>
+ <span class="k">store</span> <span class="kt">double</span> <span class="nv">%x</span><span class="p">,</span> <span class="kt">double</span><span class="p">*</span> <span class="nv">%x1</span>
+ <span class="nv">%x2</span> <span class="p">=</span> <span class="k">load</span> <span class="kt">double</span><span class="p">*</span> <span class="nv">%x1</span>
+ <span class="nv">%cmptmp</span> <span class="p">=</span> <span class="k">fcmp</span> <span class="k">ult</span> <span class="kt">double</span> <span class="nv">%x2</span><span class="p">,</span> <span class="m">3.000000e+00</span>
+ <span class="nv">%booltmp</span> <span class="p">=</span> <span class="k">uitofp</span> <span class="k">i1</span> <span class="nv">%cmptmp</span> <span class="k">to</span> <span class="kt">double</span>
+ <span class="nv">%ifcond</span> <span class="p">=</span> <span class="k">fcmp</span> <span class="k">one</span> <span class="kt">double</span> <span class="nv">%booltmp</span><span class="p">,</span> <span class="m">0.000000e+00</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%ifcond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%then</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%else</span>
+
+<span class="nl">then:</span> <span class="c">; preds = %entry</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%ifcont</span>
+
+<span class="nl">else:</span> <span class="c">; preds = %entry</span>
+ <span class="nv">%x3</span> <span class="p">=</span> <span class="k">load</span> <span class="kt">double</span><span class="p">*</span> <span class="nv">%x1</span>
+ <span class="nv">%subtmp</span> <span class="p">=</span> <span class="k">fsub</span> <span class="kt">double</span> <span class="nv">%x3</span><span class="p">,</span> <span class="m">1.000000e+00</span>
+ <span class="nv">%calltmp</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%subtmp</span><span class="p">)</span>
+ <span class="nv">%x4</span> <span class="p">=</span> <span class="k">load</span> <span class="kt">double</span><span class="p">*</span> <span class="nv">%x1</span>
+ <span class="nv">%subtmp5</span> <span class="p">=</span> <span class="k">fsub</span> <span class="kt">double</span> <span class="nv">%x4</span><span class="p">,</span> <span class="m">2.000000e+00</span>
+ <span class="nv">%calltmp6</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%subtmp5</span><span class="p">)</span>
+ <span class="nv">%addtmp</span> <span class="p">=</span> <span class="k">fadd</span> <span class="kt">double</span> <span class="nv">%calltmp</span><span class="p">,</span> <span class="nv">%calltmp6</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%ifcont</span>
+
+<span class="nl">ifcont:</span> <span class="c">; preds = %else, %then</span>
+ <span class="nv">%iftmp</span> <span class="p">=</span> <span class="k">phi</span> <span class="kt">double</span> <span class="p">[</span> <span class="m">1.000000e+00</span><span class="p">,</span> <span class="nv">%then</span> <span class="p">],</span> <span class="p">[</span> <span class="nv">%addtmp</span><span class="p">,</span> <span class="nv">%else</span> <span class="p">]</span>
+ <span class="k">ret</span> <span class="kt">double</span> <span class="nv">%iftmp</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Here there is only one variable (x, the input argument) but you can
+still see the extremely simple-minded code generation strategy we are
+using. In the entry block, an alloca is created, and the initial input
+value is stored into it. Each reference to the variable does a reload
+from the stack. Also, note that we didn’t modify the if/then/else
+expression, so it still inserts a PHI node. While we could make an
+alloca for it, it is actually easier to create a PHI node for it, so we
+still just make the PHI.</p>
+<p>Here is the code after the mem2reg pass runs:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">define</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%x</span><span class="p">)</span> <span class="p">{</span>
+<span class="nl">entry:</span>
+ <span class="nv">%cmptmp</span> <span class="p">=</span> <span class="k">fcmp</span> <span class="k">ult</span> <span class="kt">double</span> <span class="nv">%x</span><span class="p">,</span> <span class="m">3.000000e+00</span>
+ <span class="nv">%booltmp</span> <span class="p">=</span> <span class="k">uitofp</span> <span class="k">i1</span> <span class="nv">%cmptmp</span> <span class="k">to</span> <span class="kt">double</span>
+ <span class="nv">%ifcond</span> <span class="p">=</span> <span class="k">fcmp</span> <span class="k">one</span> <span class="kt">double</span> <span class="nv">%booltmp</span><span class="p">,</span> <span class="m">0.000000e+00</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%ifcond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%then</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%else</span>
+
+<span class="nl">then:</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%ifcont</span>
+
+<span class="nl">else:</span>
+ <span class="nv">%subtmp</span> <span class="p">=</span> <span class="k">fsub</span> <span class="kt">double</span> <span class="nv">%x</span><span class="p">,</span> <span class="m">1.000000e+00</span>
+ <span class="nv">%calltmp</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%subtmp</span><span class="p">)</span>
+ <span class="nv">%subtmp5</span> <span class="p">=</span> <span class="k">fsub</span> <span class="kt">double</span> <span class="nv">%x</span><span class="p">,</span> <span class="m">2.000000e+00</span>
+ <span class="nv">%calltmp6</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%subtmp5</span><span class="p">)</span>
+ <span class="nv">%addtmp</span> <span class="p">=</span> <span class="k">fadd</span> <span class="kt">double</span> <span class="nv">%calltmp</span><span class="p">,</span> <span class="nv">%calltmp6</span>
+ <span class="k">br</span> <span class="kt">label</span> <span class="nv">%ifcont</span>
+
+<span class="nl">ifcont:</span> <span class="c">; preds = %else, %then</span>
+ <span class="nv">%iftmp</span> <span class="p">=</span> <span class="k">phi</span> <span class="kt">double</span> <span class="p">[</span> <span class="m">1.000000e+00</span><span class="p">,</span> <span class="nv">%then</span> <span class="p">],</span> <span class="p">[</span> <span class="nv">%addtmp</span><span class="p">,</span> <span class="nv">%else</span> <span class="p">]</span>
+ <span class="k">ret</span> <span class="kt">double</span> <span class="nv">%iftmp</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>This is a trivial case for mem2reg, since there are no redefinitions of
+the variable. The point of showing this is to calm your tension about
+inserting such blatent inefficiencies :).</p>
+<p>After the rest of the optimizers run, we get:</p>
+<div class="highlight-llvm"><div class="highlight"><pre><span class="k">define</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%x</span><span class="p">)</span> <span class="p">{</span>
+<span class="nl">entry:</span>
+ <span class="nv">%cmptmp</span> <span class="p">=</span> <span class="k">fcmp</span> <span class="k">ult</span> <span class="kt">double</span> <span class="nv">%x</span><span class="p">,</span> <span class="m">3.000000e+00</span>
+ <span class="nv">%booltmp</span> <span class="p">=</span> <span class="k">uitofp</span> <span class="k">i1</span> <span class="nv">%cmptmp</span> <span class="k">to</span> <span class="kt">double</span>
+ <span class="nv">%ifcond</span> <span class="p">=</span> <span class="k">fcmp</span> <span class="k">ueq</span> <span class="kt">double</span> <span class="nv">%booltmp</span><span class="p">,</span> <span class="m">0.000000e+00</span>
+ <span class="k">br</span> <span class="k">i1</span> <span class="nv">%ifcond</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%else</span><span class="p">,</span> <span class="kt">label</span> <span class="nv">%ifcont</span>
+
+<span class="nl">else:</span>
+ <span class="nv">%subtmp</span> <span class="p">=</span> <span class="k">fsub</span> <span class="kt">double</span> <span class="nv">%x</span><span class="p">,</span> <span class="m">1.000000e+00</span>
+ <span class="nv">%calltmp</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%subtmp</span><span class="p">)</span>
+ <span class="nv">%subtmp5</span> <span class="p">=</span> <span class="k">fsub</span> <span class="kt">double</span> <span class="nv">%x</span><span class="p">,</span> <span class="m">2.000000e+00</span>
+ <span class="nv">%calltmp6</span> <span class="p">=</span> <span class="k">call</span> <span class="kt">double</span> <span class="vg">@fib</span><span class="p">(</span><span class="kt">double</span> <span class="nv">%subtmp5</span><span class="p">)</span>
+ <span class="nv">%addtmp</span> <span class="p">=</span> <span class="k">fadd</span> <span class="kt">double</span> <span class="nv">%calltmp</span><span class="p">,</span> <span class="nv">%calltmp6</span>
+ <span class="k">ret</span> <span class="kt">double</span> <span class="nv">%addtmp</span>
+
+<span class="nl">ifcont:</span>
+ <span class="k">ret</span> <span class="kt">double</span> <span class="m">1.000000e+00</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Here we see that the simplifycfg pass decided to clone the return
+instruction into the end of the ‘else’ block. This allowed it to
+eliminate some branches and the PHI node.</p>
+<p>Now that all symbol table references are updated to use stack variables,
+we’ll add the assignment operator.</p>
+</div>
+<div class="section" id="new-assignment-operator">
+<h2><a class="toc-backref" href="#id7">7.6. New Assignment Operator</a><a class="headerlink" href="#new-assignment-operator" title="Permalink to this headline">¶</a></h2>
+<p>With our current framework, adding a new assignment operator is really
+simple. We will parse it just like any other binary operator, but handle
+it internally (instead of allowing the user to define it). The first
+step is to set a precedence:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="k">let</span> <span class="n">main</span> <span class="bp">()</span> <span class="o">=</span>
+ <span class="c">(* Install standard binary operators.</span>
+<span class="c"> * 1 is the lowest precedence. *)</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'='</span> <span class="mi">2</span><span class="o">;</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'<'</span> <span class="mi">10</span><span class="o">;</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'+'</span> <span class="mi">20</span><span class="o">;</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'-'</span> <span class="mi">20</span><span class="o">;</span>
+ <span class="o">...</span>
+</pre></div>
+</div>
+<p>Now that the parser knows the precedence of the binary operator, it
+takes care of all the parsing and AST generation. We just need to
+implement codegen for the assignment operator. This looks like:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="k">let</span> <span class="k">rec</span> <span class="n">codegen_expr</span> <span class="o">=</span> <span class="k">function</span>
+ <span class="k">begin</span> <span class="k">match</span> <span class="n">op</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="sc">'='</span> <span class="o">-></span>
+ <span class="c">(* Special case '=' because we don't want to emit the LHS as an</span>
+<span class="c"> * expression. *)</span>
+ <span class="k">let</span> <span class="n">name</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">lhs</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Variable</span> <span class="n">name</span> <span class="o">-></span> <span class="n">name</span>
+ <span class="o">|</span> <span class="o">_</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"destination of '=' must be a variable"</span><span class="o">)</span>
+ <span class="k">in</span>
+</pre></div>
+</div>
+<p>Unlike the rest of the binary operators, our assignment operator doesn’t
+follow the “emit LHS, emit RHS, do computation” model. As such, it is
+handled as a special case before the other binary operators are handled.
+The other strange thing is that it requires the LHS to be a variable. It
+is invalid to have “(x+1) = expr” - only things like “x = expr” are
+allowed.</p>
+<div class="highlight-ocaml"><div class="highlight"><pre> <span class="c">(* Codegen the rhs. *)</span>
+ <span class="k">let</span> <span class="n">val_</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">rhs</span> <span class="k">in</span>
+
+ <span class="c">(* Lookup the name. *)</span>
+ <span class="k">let</span> <span class="n">variable</span> <span class="o">=</span> <span class="k">try</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">named_values</span> <span class="n">name</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Not_found</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"unknown variable name"</span><span class="o">)</span>
+ <span class="k">in</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">val_</span> <span class="n">variable</span> <span class="n">builder</span><span class="o">);</span>
+ <span class="n">val_</span>
+<span class="o">|</span> <span class="o">_</span> <span class="o">-></span>
+ <span class="o">...</span>
+</pre></div>
+</div>
+<p>Once we have the variable, codegen’ing the assignment is
+straightforward: we emit the RHS of the assignment, create a store, and
+return the computed value. Returning a value allows for chained
+assignments like “X = (Y = Z)”.</p>
+<p>Now that we have an assignment operator, we can mutate loop variables
+and arguments. For example, we can now run code like this:</p>
+<div class="highlight-python"><pre># Function to print a double.
+extern printd(x);
+
+# Define ':' for sequencing: as a low-precedence operator that ignores operands
+# and just returns the RHS.
+def binary : 1 (x y) y;
+
+def test(x)
+ printd(x) :
+ x = 4 :
+ printd(x);
+
+test(123);</pre>
+</div>
+<p>When run, this example prints “123” and then “4”, showing that we did
+actually mutate the value! Okay, we have now officially implemented our
+goal: getting this to work requires SSA construction in the general
+case. However, to be really useful, we want the ability to define our
+own local variables, lets add this next!</p>
+</div>
+<div class="section" id="user-defined-local-variables">
+<h2><a class="toc-backref" href="#id8">7.7. User-defined Local Variables</a><a class="headerlink" href="#user-defined-local-variables" title="Permalink to this headline">¶</a></h2>
+<p>Adding var/in is just like any other other extensions we made to
+Kaleidoscope: we extend the lexer, the parser, the AST and the code
+generator. The first step for adding our new ‘var/in’ construct is to
+extend the lexer. As before, this is pretty trivial, the code looks like
+this:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="k">type</span> <span class="n">token</span> <span class="o">=</span>
+ <span class="o">...</span>
+ <span class="c">(* var definition *)</span>
+ <span class="o">|</span> <span class="nn">Var</span>
+
+<span class="p">...</span>
+
+<span class="n">and</span> <span class="n">lex_ident</span> <span class="n">buffer</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">...</span>
+ <span class="o">|</span> <span class="s2">"in"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">In</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"binary"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Binary</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"unary"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Unary</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"var"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Var</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">...</span>
+</pre></div>
+</div>
+<p>The next step is to define the AST node that we will construct. For
+var/in, it looks like this:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="k">type</span> <span class="n">expr</span> <span class="o">=</span>
+ <span class="o">...</span>
+ <span class="c">(* variant for var/in. *)</span>
+ <span class="o">|</span> <span class="nc">Var</span> <span class="k">of</span> <span class="o">(</span><span class="kt">string</span> <span class="o">*</span> <span class="n">expr</span> <span class="n">option</span><span class="o">)</span> <span class="kt">array</span> <span class="o">*</span> <span class="n">expr</span>
+ <span class="o">...</span>
+</pre></div>
+</div>
+<p>var/in allows a list of names to be defined all at once, and each name
+can optionally have an initializer value. As such, we capture this
+information in the VarNames vector. Also, var/in has a body, this body
+is allowed to access the variables defined by the var/in.</p>
+<p>With this in place, we can define the parser pieces. The first thing we
+do is add it as a primary expression:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="c">(* primary</span>
+<span class="c"> * ::= identifier</span>
+<span class="c"> * ::= numberexpr</span>
+<span class="c"> * ::= parenexpr</span>
+<span class="c"> * ::= ifexpr</span>
+<span class="c"> * ::= forexpr</span>
+<span class="c"> * ::= varexpr *)</span>
+<span class="k">let</span> <span class="k">rec</span> <span class="n">parse_primary</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">...</span>
+ <span class="c">(* varexpr</span>
+<span class="c"> * ::= 'var' identifier ('=' expression?</span>
+<span class="c"> * (',' identifier ('=' expression)?)* 'in' expression *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Var</span><span class="o">;</span>
+ <span class="c">(* At least one variable name is required. *)</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span> <span class="o">??</span> <span class="s2">"expected identifier after var"</span><span class="o">;</span>
+ <span class="n">init</span><span class="o">=</span><span class="n">parse_var_init</span><span class="o">;</span>
+ <span class="n">var_names</span><span class="o">=</span><span class="n">parse_var_names</span> <span class="o">[(</span><span class="n">id</span><span class="o">,</span> <span class="n">init</span><span class="o">)];</span>
+ <span class="c">(* At this point, we have to have 'in'. *)</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">In</span> <span class="o">??</span> <span class="s2">"expected 'in' keyword after 'var'"</span><span class="o">;</span>
+ <span class="n">body</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">Var</span> <span class="o">(</span><span class="nn">Array</span><span class="p">.</span><span class="n">of_list</span> <span class="o">(</span><span class="nn">List</span><span class="p">.</span><span class="n">rev</span> <span class="n">var_names</span><span class="o">),</span> <span class="n">body</span><span class="o">)</span>
+
+<span class="o">...</span>
+
+<span class="ow">and</span> <span class="n">parse_var_init</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="c">(* read in the optional initializer. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">'='</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span> <span class="nc">Some</span> <span class="n">e</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="nc">None</span>
+
+<span class="ow">and</span> <span class="n">parse_var_names</span> <span class="n">accumulator</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">','</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span> <span class="o">??</span> <span class="s2">"expected identifier list after var"</span><span class="o">;</span>
+ <span class="n">init</span><span class="o">=</span><span class="n">parse_var_init</span><span class="o">;</span>
+ <span class="n">e</span><span class="o">=</span><span class="n">parse_var_names</span> <span class="o">((</span><span class="n">id</span><span class="o">,</span> <span class="n">init</span><span class="o">)</span> <span class="o">::</span> <span class="n">accumulator</span><span class="o">)</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">e</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">accumulator</span>
+</pre></div>
+</div>
+<p>Now that we can parse and represent the code, we need to support
+emission of LLVM IR for it. This code starts out with:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="k">let</span> <span class="k">rec</span> <span class="n">codegen_expr</span> <span class="o">=</span> <span class="k">function</span>
+ <span class="o">...</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Var</span> <span class="o">(</span><span class="n">var_names</span><span class="o">,</span> <span class="n">body</span><span class="o">)</span>
+ <span class="k">let</span> <span class="n">old_bindings</span> <span class="o">=</span> <span class="n">ref</span> <span class="bp">[]</span> <span class="k">in</span>
+
+ <span class="k">let</span> <span class="n">the_function</span> <span class="o">=</span> <span class="n">block_parent</span> <span class="o">(</span><span class="n">insertion_block</span> <span class="n">builder</span><span class="o">)</span> <span class="k">in</span>
+
+ <span class="c">(* Register all variables and emit their initializer. *)</span>
+ <span class="nn">Array</span><span class="p">.</span><span class="n">iter</span> <span class="o">(</span><span class="k">fun</span> <span class="o">(</span><span class="n">var_name</span><span class="o">,</span> <span class="n">init</span><span class="o">)</span> <span class="o">-></span>
+</pre></div>
+</div>
+<p>Basically it loops over all the variables, installing them one at a
+time. For each variable we put into the symbol table, we remember the
+previous value that we replace in OldBindings.</p>
+<div class="highlight-ocaml"><div class="highlight"><pre> <span class="c">(* Emit the initializer before adding the variable to scope, this</span>
+<span class="c"> * prevents the initializer from referencing the variable itself, and</span>
+<span class="c"> * permits stuff like this:</span>
+<span class="c"> * var a = 1 in</span>
+<span class="c"> * var a = a in ... # refers to outer 'a'. *)</span>
+ <span class="k">let</span> <span class="n">init_val</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">init</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">init</span> <span class="o">-></span> <span class="n">codegen_expr</span> <span class="n">init</span>
+ <span class="c">(* If not specified, use 0.0. *)</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="n">const_float</span> <span class="n">double_type</span> <span class="mi">0</span><span class="o">.</span><span class="mi">0</span>
+ <span class="k">in</span>
+
+ <span class="k">let</span> <span class="n">alloca</span> <span class="o">=</span> <span class="n">create_entry_block_alloca</span> <span class="n">the_function</span> <span class="n">var_name</span> <span class="k">in</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">init_val</span> <span class="n">alloca</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Remember the old variable binding so that we can restore the binding</span>
+<span class="c"> * when we unrecurse. *)</span>
+
+ <span class="k">begin</span>
+ <span class="k">try</span>
+ <span class="k">let</span> <span class="n">old_value</span> <span class="o">=</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="k">in</span>
+ <span class="n">old_bindings</span> <span class="o">:=</span> <span class="o">(</span><span class="n">var_name</span><span class="o">,</span> <span class="n">old_value</span><span class="o">)</span> <span class="o">::</span> <span class="o">!</span><span class="n">old_bindings</span><span class="o">;</span>
+ <span class="k">with</span> <span class="nc">Not_found</span> <span class="o">></span> <span class="bp">()</span>
+ <span class="k">end</span><span class="o">;</span>
+
+ <span class="c">(* Remember this binding. *)</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">alloca</span><span class="o">;</span>
+<span class="o">)</span> <span class="n">var_names</span><span class="o">;</span>
+</pre></div>
+</div>
+<p>There are more comments here than code. The basic idea is that we emit
+the initializer, create the alloca, then update the symbol table to
+point to it. Once all the variables are installed in the symbol table,
+we evaluate the body of the var/in expression:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="c">(* Codegen the body, now that all vars are in scope. *)</span>
+<span class="k">let</span> <span class="n">body_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">body</span> <span class="k">in</span>
+</pre></div>
+</div>
+<p>Finally, before returning, we restore the previous variable bindings:</p>
+<div class="highlight-ocaml"><div class="highlight"><pre><span class="c">(* Pop all our variables from scope. *)</span>
+<span class="nn">List</span><span class="p">.</span><span class="n">iter</span> <span class="o">(</span><span class="k">fun</span> <span class="o">(</span><span class="n">var_name</span><span class="o">,</span> <span class="n">old_value</span><span class="o">)</span> <span class="o">-></span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">old_value</span>
+<span class="o">)</span> <span class="o">!</span><span class="n">old_bindings</span><span class="o">;</span>
+
+<span class="c">(* Return the body computation. *)</span>
+<span class="n">body_val</span>
+</pre></div>
+</div>
+<p>The end result of all of this is that we get properly scoped variable
+definitions, and we even (trivially) allow mutation of them :).</p>
+<p>With this, we completed what we set out to do. Our nice iterative fib
+example from the intro compiles and runs just fine. The mem2reg pass
+optimizes all of our stack variables into SSA registers, inserting PHI
+nodes where needed, and our front-end remains simple: no “iterated
+dominance frontier” computation anywhere in sight.</p>
+</div>
+<div class="section" id="id1">
+<h2><a class="toc-backref" href="#id9">7.8. Full Code Listing</a><a class="headerlink" href="#id1" title="Permalink to this headline">¶</a></h2>
+<p>Here is the complete code listing for our running example, enhanced with
+mutable variables and var/in support. To build this example, use:</p>
+<div class="highlight-bash"><div class="highlight"><pre><span class="c"># Compile</span>
+ocamlbuild toy.byte
+<span class="c"># Run</span>
+./toy.byte
+</pre></div>
+</div>
+<p>Here is the code:</p>
+<dl class="docutils">
+<dt>_tags:</dt>
+<dd><div class="first last highlight-python"><pre><{lexer,parser}.ml>: use_camlp4, pp(camlp4of)
+<*.{byte,native}>: g++, use_llvm, use_llvm_analysis
+<*.{byte,native}>: use_llvm_executionengine, use_llvm_target
+<*.{byte,native}>: use_llvm_scalar_opts, use_bindings</pre>
+</div>
+</dd>
+<dt>myocamlbuild.ml:</dt>
+<dd><div class="first last highlight-ocaml"><div class="highlight"><pre><span class="k">open</span> <span class="nc">Ocamlbuild_plugin</span><span class="o">;;</span>
+
+<span class="n">ocaml_lib</span> <span class="o">~</span><span class="n">extern</span><span class="o">:</span><span class="bp">true</span> <span class="s2">"llvm"</span><span class="o">;;</span>
+<span class="n">ocaml_lib</span> <span class="o">~</span><span class="n">extern</span><span class="o">:</span><span class="bp">true</span> <span class="s2">"llvm_analysis"</span><span class="o">;;</span>
+<span class="n">ocaml_lib</span> <span class="o">~</span><span class="n">extern</span><span class="o">:</span><span class="bp">true</span> <span class="s2">"llvm_executionengine"</span><span class="o">;;</span>
+<span class="n">ocaml_lib</span> <span class="o">~</span><span class="n">extern</span><span class="o">:</span><span class="bp">true</span> <span class="s2">"llvm_target"</span><span class="o">;;</span>
+<span class="n">ocaml_lib</span> <span class="o">~</span><span class="n">extern</span><span class="o">:</span><span class="bp">true</span> <span class="s2">"llvm_scalar_opts"</span><span class="o">;;</span>
+
+<span class="n">flag</span> <span class="o">[</span><span class="s2">"link"</span><span class="o">;</span> <span class="s2">"ocaml"</span><span class="o">;</span> <span class="s2">"g++"</span><span class="o">]</span> <span class="o">(</span><span class="nc">S</span><span class="o">[</span><span class="nc">A</span><span class="s2">"-cc"</span><span class="o">;</span> <span class="nc">A</span><span class="s2">"g++"</span><span class="o">;</span> <span class="nc">A</span><span class="s2">"-cclib"</span><span class="o">;</span> <span class="nc">A</span><span class="s2">"-rdynamic"</span><span class="o">]);;</span>
+<span class="n">dep</span> <span class="o">[</span><span class="s2">"link"</span><span class="o">;</span> <span class="s2">"ocaml"</span><span class="o">;</span> <span class="s2">"use_bindings"</span><span class="o">]</span> <span class="o">[</span><span class="s2">"bindings.o"</span><span class="o">];;</span>
+</pre></div>
+</div>
+</dd>
+<dt>token.ml:</dt>
+<dd><div class="first last highlight-ocaml"><div class="highlight"><pre><span class="c">(*===----------------------------------------------------------------------===</span>
+<span class="c"> * Lexer Tokens</span>
+<span class="c"> *===----------------------------------------------------------------------===*)</span>
+
+<span class="c">(* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of</span>
+<span class="c"> * these others for known things. *)</span>
+<span class="k">type</span> <span class="n">token</span> <span class="o">=</span>
+ <span class="c">(* commands *)</span>
+ <span class="o">|</span> <span class="nc">Def</span> <span class="o">|</span> <span class="nc">Extern</span>
+
+ <span class="c">(* primary *)</span>
+ <span class="o">|</span> <span class="nc">Ident</span> <span class="k">of</span> <span class="kt">string</span> <span class="o">|</span> <span class="nc">Number</span> <span class="k">of</span> <span class="kt">float</span>
+
+ <span class="c">(* unknown *)</span>
+ <span class="o">|</span> <span class="nc">Kwd</span> <span class="k">of</span> <span class="kt">char</span>
+
+ <span class="c">(* control *)</span>
+ <span class="o">|</span> <span class="nc">If</span> <span class="o">|</span> <span class="nc">Then</span> <span class="o">|</span> <span class="nc">Else</span>
+ <span class="o">|</span> <span class="nc">For</span> <span class="o">|</span> <span class="nc">In</span>
+
+ <span class="c">(* operators *)</span>
+ <span class="o">|</span> <span class="nc">Binary</span> <span class="o">|</span> <span class="nc">Unary</span>
+
+ <span class="c">(* var definition *)</span>
+ <span class="o">|</span> <span class="nc">Var</span>
+</pre></div>
+</div>
+</dd>
+<dt>lexer.ml:</dt>
+<dd><div class="first last highlight-ocaml"><div class="highlight"><pre><span class="c">(*===----------------------------------------------------------------------===</span>
+<span class="c"> * Lexer</span>
+<span class="c"> *===----------------------------------------------------------------------===*)</span>
+
+<span class="k">let</span> <span class="k">rec</span> <span class="n">lex</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="c">(* Skip any whitespace. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span> <span class="o">(</span><span class="sc">' '</span> <span class="o">|</span> <span class="sc">'\n'</span> <span class="o">|</span> <span class="sc">'\r'</span> <span class="o">|</span> <span class="sc">'\t'</span><span class="o">);</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">lex</span> <span class="n">stream</span>
+
+ <span class="c">(* identifier: [a-zA-Z][a-zA-Z0-9] *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span> <span class="o">(</span><span class="sc">'A'</span> <span class="o">..</span> <span class="sc">'Z'</span> <span class="o">|</span> <span class="sc">'a'</span> <span class="o">..</span> <span class="sc">'z'</span> <span class="k">as</span> <span class="n">c</span><span class="o">);</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">buffer</span> <span class="o">=</span> <span class="nn">Buffer</span><span class="p">.</span><span class="n">create</span> <span class="mi">1</span> <span class="k">in</span>
+ <span class="nn">Buffer</span><span class="p">.</span><span class="n">add_char</span> <span class="n">buffer</span> <span class="n">c</span><span class="o">;</span>
+ <span class="n">lex_ident</span> <span class="n">buffer</span> <span class="n">stream</span>
+
+ <span class="c">(* number: [0-9.]+ *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span> <span class="o">(</span><span class="sc">'0'</span> <span class="o">..</span> <span class="sc">'9'</span> <span class="k">as</span> <span class="n">c</span><span class="o">);</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">buffer</span> <span class="o">=</span> <span class="nn">Buffer</span><span class="p">.</span><span class="n">create</span> <span class="mi">1</span> <span class="k">in</span>
+ <span class="nn">Buffer</span><span class="p">.</span><span class="n">add_char</span> <span class="n">buffer</span> <span class="n">c</span><span class="o">;</span>
+ <span class="n">lex_number</span> <span class="n">buffer</span> <span class="n">stream</span>
+
+ <span class="c">(* Comment until end of line. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span> <span class="o">(</span><span class="sc">'#'</span><span class="o">);</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="n">lex_comment</span> <span class="n">stream</span>
+
+ <span class="c">(* Otherwise, just return the character as its ascii value. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="n">c</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="n">c</span><span class="o">;</span> <span class="n">lex</span> <span class="n">stream</span> <span class="o">>]</span>
+
+ <span class="c">(* end of stream. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="o">[<</span> <span class="o">>]</span>
+
+<span class="ow">and</span> <span class="n">lex_number</span> <span class="n">buffer</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span> <span class="o">(</span><span class="sc">'0'</span> <span class="o">..</span> <span class="sc">'9'</span> <span class="o">|</span> <span class="sc">'.'</span> <span class="k">as</span> <span class="n">c</span><span class="o">);</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Buffer</span><span class="p">.</span><span class="n">add_char</span> <span class="n">buffer</span> <span class="n">c</span><span class="o">;</span>
+ <span class="n">lex_number</span> <span class="n">buffer</span> <span class="n">stream</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="n">stream</span><span class="o">=</span><span class="n">lex</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Number</span> <span class="o">(</span><span class="n">float_of_string</span> <span class="o">(</span><span class="nn">Buffer</span><span class="p">.</span><span class="n">contents</span> <span class="n">buffer</span><span class="o">));</span> <span class="n">stream</span> <span class="o">>]</span>
+
+<span class="ow">and</span> <span class="n">lex_ident</span> <span class="n">buffer</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span> <span class="o">(</span><span class="sc">'A'</span> <span class="o">..</span> <span class="sc">'Z'</span> <span class="o">|</span> <span class="sc">'a'</span> <span class="o">..</span> <span class="sc">'z'</span> <span class="o">|</span> <span class="sc">'0'</span> <span class="o">..</span> <span class="sc">'9'</span> <span class="k">as</span> <span class="n">c</span><span class="o">);</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Buffer</span><span class="p">.</span><span class="n">add_char</span> <span class="n">buffer</span> <span class="n">c</span><span class="o">;</span>
+ <span class="n">lex_ident</span> <span class="n">buffer</span> <span class="n">stream</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="n">stream</span><span class="o">=</span><span class="n">lex</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">match</span> <span class="nn">Buffer</span><span class="p">.</span><span class="n">contents</span> <span class="n">buffer</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="s2">"def"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Def</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"extern"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Extern</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"if"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">If</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"then"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Then</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"else"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Else</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"for"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">For</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"in"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">In</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"binary"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Binary</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"unary"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Unary</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="s2">"var"</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Var</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+ <span class="o">|</span> <span class="n">id</span> <span class="o">-></span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span>
+
+<span class="ow">and</span> <span class="n">lex_comment</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span> <span class="o">(</span><span class="sc">'\n'</span><span class="o">);</span> <span class="n">stream</span><span class="o">=</span><span class="n">lex</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">stream</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="n">c</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">lex_comment</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">e</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="o">[<</span> <span class="o">>]</span>
+</pre></div>
+</div>
+</dd>
+<dt>ast.ml:</dt>
+<dd><div class="first last highlight-ocaml"><div class="highlight"><pre><span class="c">(*===----------------------------------------------------------------------===</span>
+<span class="c"> * Abstract Syntax Tree (aka Parse Tree)</span>
+<span class="c"> *===----------------------------------------------------------------------===*)</span>
+
+<span class="c">(* expr - Base type for all expression nodes. *)</span>
+<span class="k">type</span> <span class="n">expr</span> <span class="o">=</span>
+ <span class="c">(* variant for numeric literals like "1.0". *)</span>
+ <span class="o">|</span> <span class="nc">Number</span> <span class="k">of</span> <span class="kt">float</span>
+
+ <span class="c">(* variant for referencing a variable, like "a". *)</span>
+ <span class="o">|</span> <span class="nc">Variable</span> <span class="k">of</span> <span class="kt">string</span>
+
+ <span class="c">(* variant for a unary operator. *)</span>
+ <span class="o">|</span> <span class="nc">Unary</span> <span class="k">of</span> <span class="kt">char</span> <span class="o">*</span> <span class="n">expr</span>
+
+ <span class="c">(* variant for a binary operator. *)</span>
+ <span class="o">|</span> <span class="nc">Binary</span> <span class="k">of</span> <span class="kt">char</span> <span class="o">*</span> <span class="n">expr</span> <span class="o">*</span> <span class="n">expr</span>
+
+ <span class="c">(* variant for function calls. *)</span>
+ <span class="o">|</span> <span class="nc">Call</span> <span class="k">of</span> <span class="kt">string</span> <span class="o">*</span> <span class="n">expr</span> <span class="kt">array</span>
+
+ <span class="c">(* variant for if/then/else. *)</span>
+ <span class="o">|</span> <span class="nc">If</span> <span class="k">of</span> <span class="n">expr</span> <span class="o">*</span> <span class="n">expr</span> <span class="o">*</span> <span class="n">expr</span>
+
+ <span class="c">(* variant for for/in. *)</span>
+ <span class="o">|</span> <span class="nc">For</span> <span class="k">of</span> <span class="kt">string</span> <span class="o">*</span> <span class="n">expr</span> <span class="o">*</span> <span class="n">expr</span> <span class="o">*</span> <span class="n">expr</span> <span class="n">option</span> <span class="o">*</span> <span class="n">expr</span>
+
+ <span class="c">(* variant for var/in. *)</span>
+ <span class="o">|</span> <span class="nc">Var</span> <span class="k">of</span> <span class="o">(</span><span class="kt">string</span> <span class="o">*</span> <span class="n">expr</span> <span class="n">option</span><span class="o">)</span> <span class="kt">array</span> <span class="o">*</span> <span class="n">expr</span>
+
+<span class="c">(* proto - This type represents the "prototype" for a function, which captures</span>
+<span class="c"> * its name, and its argument names (thus implicitly the number of arguments the</span>
+<span class="c"> * function takes). *)</span>
+<span class="k">type</span> <span class="n">proto</span> <span class="o">=</span>
+ <span class="o">|</span> <span class="nc">Prototype</span> <span class="k">of</span> <span class="kt">string</span> <span class="o">*</span> <span class="kt">string</span> <span class="kt">array</span>
+ <span class="o">|</span> <span class="nc">BinOpPrototype</span> <span class="k">of</span> <span class="kt">string</span> <span class="o">*</span> <span class="kt">string</span> <span class="kt">array</span> <span class="o">*</span> <span class="kt">int</span>
+
+<span class="c">(* func - This type represents a function definition itself. *)</span>
+<span class="k">type</span> <span class="n">func</span> <span class="o">=</span> <span class="nc">Function</span> <span class="k">of</span> <span class="n">proto</span> <span class="o">*</span> <span class="n">expr</span>
+</pre></div>
+</div>
+</dd>
+<dt>parser.ml:</dt>
+<dd><div class="first last highlight-ocaml"><div class="highlight"><pre><span class="c">(*===---------------------------------------------------------------------===</span>
+<span class="c"> * Parser</span>
+<span class="c"> *===---------------------------------------------------------------------===*)</span>
+
+<span class="c">(* binop_precedence - This holds the precedence for each binary operator that is</span>
+<span class="c"> * defined *)</span>
+<span class="k">let</span> <span class="n">binop_precedence</span><span class="o">:(</span><span class="kt">char</span><span class="o">,</span> <span class="kt">int</span><span class="o">)</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">t</span> <span class="o">=</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">create</span> <span class="mi">10</span>
+
+<span class="c">(* precedence - Get the precedence of the pending binary operator token. *)</span>
+<span class="k">let</span> <span class="n">precedence</span> <span class="n">c</span> <span class="o">=</span> <span class="k">try</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">binop_precedence</span> <span class="n">c</span> <span class="k">with</span> <span class="nc">Not_found</span> <span class="o">-></span> <span class="o">-</span><span class="mi">1</span>
+
+<span class="c">(* primary</span>
+<span class="c"> * ::= identifier</span>
+<span class="c"> * ::= numberexpr</span>
+<span class="c"> * ::= parenexpr</span>
+<span class="c"> * ::= ifexpr</span>
+<span class="c"> * ::= forexpr</span>
+<span class="c"> * ::= varexpr *)</span>
+<span class="k">let</span> <span class="k">rec</span> <span class="n">parse_primary</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="c">(* numberexpr ::= number *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Number</span> <span class="n">n</span> <span class="o">>]</span> <span class="o">-></span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Number</span> <span class="n">n</span>
+
+ <span class="c">(* parenexpr ::= '(' expression ')' *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">'('</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_expr</span><span class="o">;</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">')'</span> <span class="o">??</span> <span class="s2">"expected ')'"</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">e</span>
+
+ <span class="c">(* identifierexpr</span>
+<span class="c"> * ::= identifier</span>
+<span class="c"> * ::= identifier '(' argumentexpr ')' *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="k">rec</span> <span class="n">parse_args</span> <span class="n">accumulator</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_expr</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">begin</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">','</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_args</span> <span class="o">(</span><span class="n">e</span> <span class="o">::</span> <span class="n">accumulator</span><span class="o">)</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">e</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">e</span> <span class="o">::</span> <span class="n">accumulator</span>
+ <span class="k">end</span> <span class="n">stream</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">accumulator</span>
+ <span class="k">in</span>
+ <span class="k">let</span> <span class="k">rec</span> <span class="n">parse_ident</span> <span class="n">id</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="c">(* Call. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">'('</span><span class="o">;</span>
+ <span class="n">args</span><span class="o">=</span><span class="n">parse_args</span> <span class="bp">[]</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">')'</span> <span class="o">??</span> <span class="s2">"expected ')'"</span><span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">Call</span> <span class="o">(</span><span class="n">id</span><span class="o">,</span> <span class="nn">Array</span><span class="p">.</span><span class="n">of_list</span> <span class="o">(</span><span class="nn">List</span><span class="p">.</span><span class="n">rev</span> <span class="n">args</span><span class="o">))</span>
+
+ <span class="c">(* Simple variable ref. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Variable</span> <span class="n">id</span>
+ <span class="k">in</span>
+ <span class="n">parse_ident</span> <span class="n">id</span> <span class="n">stream</span>
+
+ <span class="c">(* ifexpr ::= 'if' expr 'then' expr 'else' expr *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">If</span><span class="o">;</span> <span class="n">c</span><span class="o">=</span><span class="n">parse_expr</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Then</span> <span class="o">??</span> <span class="s2">"expected 'then'"</span><span class="o">;</span> <span class="n">t</span><span class="o">=</span><span class="n">parse_expr</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Else</span> <span class="o">??</span> <span class="s2">"expected 'else'"</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">If</span> <span class="o">(</span><span class="n">c</span><span class="o">,</span> <span class="n">t</span><span class="o">,</span> <span class="n">e</span><span class="o">)</span>
+
+ <span class="c">(* forexpr</span>
+<span class="c"> ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">For</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span> <span class="o">??</span> <span class="s2">"expected identifier after for"</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">'='</span> <span class="o">??</span> <span class="s2">"expected '=' after for"</span><span class="o">;</span>
+ <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">begin</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span>
+ <span class="n">start</span><span class="o">=</span><span class="n">parse_expr</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">','</span> <span class="o">??</span> <span class="s2">"expected ',' after for"</span><span class="o">;</span>
+ <span class="n">end_</span><span class="o">=</span><span class="n">parse_expr</span><span class="o">;</span>
+ <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">step</span> <span class="o">=</span>
+ <span class="k">begin</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">','</span><span class="o">;</span> <span class="n">step</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span> <span class="nc">Some</span> <span class="n">step</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="nc">None</span>
+ <span class="k">end</span> <span class="n">stream</span>
+ <span class="k">in</span>
+ <span class="k">begin</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">In</span><span class="o">;</span> <span class="n">body</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">For</span> <span class="o">(</span><span class="n">id</span><span class="o">,</span> <span class="n">start</span><span class="o">,</span> <span class="n">end_</span><span class="o">,</span> <span class="n">step</span><span class="o">,</span> <span class="n">body</span><span class="o">)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">raise</span> <span class="o">(</span><span class="nn">Stream</span><span class="p">.</span><span class="nc">Error</span> <span class="s2">"expected 'in' after for"</span><span class="o">)</span>
+ <span class="k">end</span> <span class="n">stream</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">raise</span> <span class="o">(</span><span class="nn">Stream</span><span class="p">.</span><span class="nc">Error</span> <span class="s2">"expected '=' after for"</span><span class="o">)</span>
+ <span class="k">end</span> <span class="n">stream</span>
+
+ <span class="c">(* varexpr</span>
+<span class="c"> * ::= 'var' identifier ('=' expression?</span>
+<span class="c"> * (',' identifier ('=' expression)?)* 'in' expression *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Var</span><span class="o">;</span>
+ <span class="c">(* At least one variable name is required. *)</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span> <span class="o">??</span> <span class="s2">"expected identifier after var"</span><span class="o">;</span>
+ <span class="n">init</span><span class="o">=</span><span class="n">parse_var_init</span><span class="o">;</span>
+ <span class="n">var_names</span><span class="o">=</span><span class="n">parse_var_names</span> <span class="o">[(</span><span class="n">id</span><span class="o">,</span> <span class="n">init</span><span class="o">)];</span>
+ <span class="c">(* At this point, we have to have 'in'. *)</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">In</span> <span class="o">??</span> <span class="s2">"expected 'in' keyword after 'var'"</span><span class="o">;</span>
+ <span class="n">body</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">Var</span> <span class="o">(</span><span class="nn">Array</span><span class="p">.</span><span class="n">of_list</span> <span class="o">(</span><span class="nn">List</span><span class="p">.</span><span class="n">rev</span> <span class="n">var_names</span><span class="o">),</span> <span class="n">body</span><span class="o">)</span>
+
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nn">Stream</span><span class="p">.</span><span class="nc">Error</span> <span class="s2">"unknown token when expecting an expression."</span><span class="o">)</span>
+
+<span class="c">(* unary</span>
+<span class="c"> * ::= primary</span>
+<span class="c"> * ::= '!' unary *)</span>
+<span class="ow">and</span> <span class="n">parse_unary</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="c">(* If this is a unary operator, read it. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="n">op</span> <span class="k">when</span> <span class="n">op</span> <span class="o">!=</span> <span class="sc">'('</span> <span class="o">&&</span> <span class="n">op</span> <span class="o">!=</span> <span class="sc">')'</span><span class="o">;</span> <span class="n">operand</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">Unary</span> <span class="o">(</span><span class="n">op</span><span class="o">,</span> <span class="n">operand</span><span class="o">)</span>
+
+ <span class="c">(* If the current token is not an operator, it must be a primary expr. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">parse_primary</span> <span class="n">stream</span>
+
+<span class="c">(* binoprhs</span>
+<span class="c"> * ::= ('+' primary)* *)</span>
+<span class="ow">and</span> <span class="n">parse_bin_rhs</span> <span class="n">expr_prec</span> <span class="n">lhs</span> <span class="n">stream</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="nn">Stream</span><span class="p">.</span><span class="n">peek</span> <span class="n">stream</span> <span class="k">with</span>
+ <span class="c">(* If this is a binop, find its precedence. *)</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="o">(</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="n">c</span><span class="o">)</span> <span class="k">when</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">mem</span> <span class="n">binop_precedence</span> <span class="n">c</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">token_prec</span> <span class="o">=</span> <span class="n">precedence</span> <span class="n">c</span> <span class="k">in</span>
+
+ <span class="c">(* If this is a binop that binds at least as tightly as the current binop,</span>
+<span class="c"> * consume it, otherwise we are done. *)</span>
+ <span class="k">if</span> <span class="n">token_prec</span> <span class="o"><</span> <span class="n">expr_prec</span> <span class="k">then</span> <span class="n">lhs</span> <span class="k">else</span> <span class="k">begin</span>
+ <span class="c">(* Eat the binop. *)</span>
+ <span class="nn">Stream</span><span class="p">.</span><span class="n">junk</span> <span class="n">stream</span><span class="o">;</span>
+
+ <span class="c">(* Parse the primary expression after the binary operator. *)</span>
+ <span class="k">let</span> <span class="n">rhs</span> <span class="o">=</span> <span class="n">parse_unary</span> <span class="n">stream</span> <span class="k">in</span>
+
+ <span class="c">(* Okay, we know this is a binop. *)</span>
+ <span class="k">let</span> <span class="n">rhs</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="nn">Stream</span><span class="p">.</span><span class="n">peek</span> <span class="n">stream</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="o">(</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="n">c2</span><span class="o">)</span> <span class="o">-></span>
+ <span class="c">(* If BinOp binds less tightly with rhs than the operator after</span>
+<span class="c"> * rhs, let the pending operator take rhs as its lhs. *)</span>
+ <span class="k">let</span> <span class="n">next_prec</span> <span class="o">=</span> <span class="n">precedence</span> <span class="n">c2</span> <span class="k">in</span>
+ <span class="k">if</span> <span class="n">token_prec</span> <span class="o"><</span> <span class="n">next_prec</span>
+ <span class="k">then</span> <span class="n">parse_bin_rhs</span> <span class="o">(</span><span class="n">token_prec</span> <span class="o">+</span> <span class="mi">1</span><span class="o">)</span> <span class="n">rhs</span> <span class="n">stream</span>
+ <span class="k">else</span> <span class="n">rhs</span>
+ <span class="o">|</span> <span class="o">_</span> <span class="o">-></span> <span class="n">rhs</span>
+ <span class="k">in</span>
+
+ <span class="c">(* Merge lhs/rhs. *)</span>
+ <span class="k">let</span> <span class="n">lhs</span> <span class="o">=</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Binary</span> <span class="o">(</span><span class="n">c</span><span class="o">,</span> <span class="n">lhs</span><span class="o">,</span> <span class="n">rhs</span><span class="o">)</span> <span class="k">in</span>
+ <span class="n">parse_bin_rhs</span> <span class="n">expr_prec</span> <span class="n">lhs</span> <span class="n">stream</span>
+ <span class="k">end</span>
+ <span class="o">|</span> <span class="o">_</span> <span class="o">-></span> <span class="n">lhs</span>
+
+<span class="ow">and</span> <span class="n">parse_var_init</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="c">(* read in the optional initializer. *)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">'='</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span> <span class="nc">Some</span> <span class="n">e</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="nc">None</span>
+
+<span class="ow">and</span> <span class="n">parse_var_names</span> <span class="n">accumulator</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">','</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span> <span class="o">??</span> <span class="s2">"expected identifier list after var"</span><span class="o">;</span>
+ <span class="n">init</span><span class="o">=</span><span class="n">parse_var_init</span><span class="o">;</span>
+ <span class="n">e</span><span class="o">=</span><span class="n">parse_var_names</span> <span class="o">((</span><span class="n">id</span><span class="o">,</span> <span class="n">init</span><span class="o">)</span> <span class="o">::</span> <span class="n">accumulator</span><span class="o">)</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">e</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">accumulator</span>
+
+<span class="c">(* expression</span>
+<span class="c"> * ::= primary binoprhs *)</span>
+<span class="ow">and</span> <span class="n">parse_expr</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="n">lhs</span><span class="o">=</span><span class="n">parse_unary</span><span class="o">;</span> <span class="n">stream</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">parse_bin_rhs</span> <span class="mi">0</span> <span class="n">lhs</span> <span class="n">stream</span>
+
+<span class="c">(* prototype</span>
+<span class="c"> * ::= id '(' id* ')'</span>
+<span class="c"> * ::= binary LETTER number? (id, id)</span>
+<span class="c"> * ::= unary LETTER number? (id) *)</span>
+<span class="k">let</span> <span class="n">parse_prototype</span> <span class="o">=</span>
+ <span class="k">let</span> <span class="k">rec</span> <span class="n">parse_args</span> <span class="n">accumulator</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_args</span> <span class="o">(</span><span class="n">id</span><span class="o">::</span><span class="n">accumulator</span><span class="o">)</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">e</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">accumulator</span>
+ <span class="k">in</span>
+ <span class="k">let</span> <span class="n">parse_operator</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Unary</span> <span class="o">>]</span> <span class="o">-></span> <span class="s2">"unary"</span><span class="o">,</span> <span class="mi">1</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Binary</span> <span class="o">>]</span> <span class="o">-></span> <span class="s2">"binary"</span><span class="o">,</span> <span class="mi">2</span>
+ <span class="k">in</span>
+ <span class="k">let</span> <span class="n">parse_binary_precedence</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Number</span> <span class="n">n</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">int_of_float</span> <span class="n">n</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span> <span class="mi">30</span>
+ <span class="k">in</span>
+ <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Ident</span> <span class="n">id</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">'('</span> <span class="o">??</span> <span class="s2">"expected '(' in prototype"</span><span class="o">;</span>
+ <span class="n">args</span><span class="o">=</span><span class="n">parse_args</span> <span class="bp">[]</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">')'</span> <span class="o">??</span> <span class="s2">"expected ')' in prototype"</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="c">(* success. *)</span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">Prototype</span> <span class="o">(</span><span class="n">id</span><span class="o">,</span> <span class="nn">Array</span><span class="p">.</span><span class="n">of_list</span> <span class="o">(</span><span class="nn">List</span><span class="p">.</span><span class="n">rev</span> <span class="n">args</span><span class="o">))</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">(</span><span class="n">prefix</span><span class="o">,</span> <span class="n">kind</span><span class="o">)=</span><span class="n">parse_operator</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="n">op</span> <span class="o">??</span> <span class="s2">"expected an operator"</span><span class="o">;</span>
+ <span class="c">(* Read the precedence if present. *)</span>
+ <span class="n">binary_precedence</span><span class="o">=</span><span class="n">parse_binary_precedence</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">'('</span> <span class="o">??</span> <span class="s2">"expected '(' in prototype"</span><span class="o">;</span>
+ <span class="n">args</span><span class="o">=</span><span class="n">parse_args</span> <span class="bp">[]</span><span class="o">;</span>
+ <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">')'</span> <span class="o">??</span> <span class="s2">"expected ')' in prototype"</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">name</span> <span class="o">=</span> <span class="n">prefix</span> <span class="o">^</span> <span class="o">(</span><span class="nn">String</span><span class="p">.</span><span class="n">make</span> <span class="mi">1</span> <span class="n">op</span><span class="o">)</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">args</span> <span class="o">=</span> <span class="nn">Array</span><span class="p">.</span><span class="n">of_list</span> <span class="o">(</span><span class="nn">List</span><span class="p">.</span><span class="n">rev</span> <span class="n">args</span><span class="o">)</span> <span class="k">in</span>
+
+ <span class="c">(* Verify right number of arguments for operator. *)</span>
+ <span class="k">if</span> <span class="nn">Array</span><span class="p">.</span><span class="n">length</span> <span class="n">args</span> <span class="o">!=</span> <span class="n">kind</span>
+ <span class="k">then</span> <span class="k">raise</span> <span class="o">(</span><span class="nn">Stream</span><span class="p">.</span><span class="nc">Error</span> <span class="s2">"invalid number of operands for operator"</span><span class="o">)</span>
+ <span class="k">else</span>
+ <span class="k">if</span> <span class="n">kind</span> <span class="o">==</span> <span class="mi">1</span> <span class="k">then</span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">Prototype</span> <span class="o">(</span><span class="n">name</span><span class="o">,</span> <span class="n">args</span><span class="o">)</span>
+ <span class="k">else</span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">BinOpPrototype</span> <span class="o">(</span><span class="n">name</span><span class="o">,</span> <span class="n">args</span><span class="o">,</span> <span class="n">binary_precedence</span><span class="o">)</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="k">raise</span> <span class="o">(</span><span class="nn">Stream</span><span class="p">.</span><span class="nc">Error</span> <span class="s2">"expected function name in prototype"</span><span class="o">)</span>
+
+<span class="c">(* definition ::= 'def' prototype expression *)</span>
+<span class="k">let</span> <span class="n">parse_definition</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Def</span><span class="o">;</span> <span class="n">p</span><span class="o">=</span><span class="n">parse_prototype</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">Function</span> <span class="o">(</span><span class="n">p</span><span class="o">,</span> <span class="n">e</span><span class="o">)</span>
+
+<span class="c">(* toplevelexpr ::= expression *)</span>
+<span class="k">let</span> <span class="n">parse_toplevel</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_expr</span> <span class="o">>]</span> <span class="o">-></span>
+ <span class="c">(* Make an anonymous proto. *)</span>
+ <span class="nn">Ast</span><span class="p">.</span><span class="nc">Function</span> <span class="o">(</span><span class="nn">Ast</span><span class="p">.</span><span class="nc">Prototype</span> <span class="o">(</span><span class="s2">""</span><span class="o">,</span> <span class="o">[||]),</span> <span class="n">e</span><span class="o">)</span>
+
+<span class="c">(* external ::= 'extern' prototype *)</span>
+<span class="k">let</span> <span class="n">parse_extern</span> <span class="o">=</span> <span class="n">parser</span>
+ <span class="o">|</span> <span class="o">[<</span> <span class="k">'</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Extern</span><span class="o">;</span> <span class="n">e</span><span class="o">=</span><span class="n">parse_prototype</span> <span class="o">>]</span> <span class="o">-></span> <span class="n">e</span>
+</pre></div>
+</div>
+</dd>
+<dt>codegen.ml:</dt>
+<dd><div class="first last highlight-ocaml"><div class="highlight"><pre><span class="c">(*===----------------------------------------------------------------------===</span>
+<span class="c"> * Code Generation</span>
+<span class="c"> *===----------------------------------------------------------------------===*)</span>
+
+<span class="k">open</span> <span class="nc">Llvm</span>
+
+<span class="k">exception</span> <span class="nc">Error</span> <span class="k">of</span> <span class="kt">string</span>
+
+<span class="k">let</span> <span class="n">context</span> <span class="o">=</span> <span class="n">global_context</span> <span class="bp">()</span>
+<span class="k">let</span> <span class="n">the_module</span> <span class="o">=</span> <span class="n">create_module</span> <span class="n">context</span> <span class="s2">"my cool jit"</span>
+<span class="k">let</span> <span class="n">builder</span> <span class="o">=</span> <span class="n">builder</span> <span class="n">context</span>
+<span class="k">let</span> <span class="n">named_values</span><span class="o">:(</span><span class="kt">string</span><span class="o">,</span> <span class="n">llvalue</span><span class="o">)</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">t</span> <span class="o">=</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">create</span> <span class="mi">10</span>
+<span class="k">let</span> <span class="n">double_type</span> <span class="o">=</span> <span class="n">double_type</span> <span class="n">context</span>
+
+<span class="c">(* Create an alloca instruction in the entry block of the function. This</span>
+<span class="c"> * is used for mutable variables etc. *)</span>
+<span class="k">let</span> <span class="n">create_entry_block_alloca</span> <span class="n">the_function</span> <span class="n">var_name</span> <span class="o">=</span>
+ <span class="k">let</span> <span class="n">builder</span> <span class="o">=</span> <span class="n">builder_at</span> <span class="n">context</span> <span class="o">(</span><span class="n">instr_begin</span> <span class="o">(</span><span class="n">entry_block</span> <span class="n">the_function</span><span class="o">))</span> <span class="k">in</span>
+ <span class="n">build_alloca</span> <span class="n">double_type</span> <span class="n">var_name</span> <span class="n">builder</span>
+
+<span class="k">let</span> <span class="k">rec</span> <span class="n">codegen_expr</span> <span class="o">=</span> <span class="k">function</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Number</span> <span class="n">n</span> <span class="o">-></span> <span class="n">const_float</span> <span class="n">double_type</span> <span class="n">n</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Variable</span> <span class="n">name</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">v</span> <span class="o">=</span> <span class="k">try</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">named_values</span> <span class="n">name</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Not_found</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"unknown variable name"</span><span class="o">)</span>
+ <span class="k">in</span>
+ <span class="c">(* Load the value. *)</span>
+ <span class="n">build_load</span> <span class="n">v</span> <span class="n">name</span> <span class="n">builder</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Unary</span> <span class="o">(</span><span class="n">op</span><span class="o">,</span> <span class="n">operand</span><span class="o">)</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">operand</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">operand</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">callee</span> <span class="o">=</span> <span class="s2">"unary"</span> <span class="o">^</span> <span class="o">(</span><span class="nn">String</span><span class="p">.</span><span class="n">make</span> <span class="mi">1</span> <span class="n">op</span><span class="o">)</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">callee</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">lookup_function</span> <span class="n">callee</span> <span class="n">the_module</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">callee</span> <span class="o">-></span> <span class="n">callee</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"unknown unary operator"</span><span class="o">)</span>
+ <span class="k">in</span>
+ <span class="n">build_call</span> <span class="n">callee</span> <span class="o">[|</span><span class="n">operand</span><span class="o">|]</span> <span class="s2">"unop"</span> <span class="n">builder</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Binary</span> <span class="o">(</span><span class="n">op</span><span class="o">,</span> <span class="n">lhs</span><span class="o">,</span> <span class="n">rhs</span><span class="o">)</span> <span class="o">-></span>
+ <span class="k">begin</span> <span class="k">match</span> <span class="n">op</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="sc">'='</span> <span class="o">-></span>
+ <span class="c">(* Special case '=' because we don't want to emit the LHS as an</span>
+<span class="c"> * expression. *)</span>
+ <span class="k">let</span> <span class="n">name</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">lhs</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Variable</span> <span class="n">name</span> <span class="o">-></span> <span class="n">name</span>
+ <span class="o">|</span> <span class="o">_</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"destination of '=' must be a variable"</span><span class="o">)</span>
+ <span class="k">in</span>
+
+ <span class="c">(* Codegen the rhs. *)</span>
+ <span class="k">let</span> <span class="n">val_</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">rhs</span> <span class="k">in</span>
+
+ <span class="c">(* Lookup the name. *)</span>
+ <span class="k">let</span> <span class="n">variable</span> <span class="o">=</span> <span class="k">try</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">named_values</span> <span class="n">name</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Not_found</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"unknown variable name"</span><span class="o">)</span>
+ <span class="k">in</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">val_</span> <span class="n">variable</span> <span class="n">builder</span><span class="o">);</span>
+ <span class="n">val_</span>
+ <span class="o">|</span> <span class="o">_</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">lhs_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">lhs</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">rhs_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">rhs</span> <span class="k">in</span>
+ <span class="k">begin</span>
+ <span class="k">match</span> <span class="n">op</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="sc">'+'</span> <span class="o">-></span> <span class="n">build_add</span> <span class="n">lhs_val</span> <span class="n">rhs_val</span> <span class="s2">"addtmp"</span> <span class="n">builder</span>
+ <span class="o">|</span> <span class="sc">'-'</span> <span class="o">-></span> <span class="n">build_sub</span> <span class="n">lhs_val</span> <span class="n">rhs_val</span> <span class="s2">"subtmp"</span> <span class="n">builder</span>
+ <span class="o">|</span> <span class="sc">'*'</span> <span class="o">-></span> <span class="n">build_mul</span> <span class="n">lhs_val</span> <span class="n">rhs_val</span> <span class="s2">"multmp"</span> <span class="n">builder</span>
+ <span class="o">|</span> <span class="sc">'<'</span> <span class="o">-></span>
+ <span class="c">(* Convert bool 0/1 to double 0.0 or 1.0 *)</span>
+ <span class="k">let</span> <span class="n">i</span> <span class="o">=</span> <span class="n">build_fcmp</span> <span class="nn">Fcmp</span><span class="p">.</span><span class="nc">Ult</span> <span class="n">lhs_val</span> <span class="n">rhs_val</span> <span class="s2">"cmptmp"</span> <span class="n">builder</span> <span class="k">in</span>
+ <span class="n">build_uitofp</span> <span class="n">i</span> <span class="n">double_type</span> <span class="s2">"booltmp"</span> <span class="n">builder</span>
+ <span class="o">|</span> <span class="o">_</span> <span class="o">-></span>
+ <span class="c">(* If it wasn't a builtin binary operator, it must be a user defined</span>
+<span class="c"> * one. Emit a call to it. *)</span>
+ <span class="k">let</span> <span class="n">callee</span> <span class="o">=</span> <span class="s2">"binary"</span> <span class="o">^</span> <span class="o">(</span><span class="nn">String</span><span class="p">.</span><span class="n">make</span> <span class="mi">1</span> <span class="n">op</span><span class="o">)</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">callee</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">lookup_function</span> <span class="n">callee</span> <span class="n">the_module</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">callee</span> <span class="o">-></span> <span class="n">callee</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"binary operator not found!"</span><span class="o">)</span>
+ <span class="k">in</span>
+ <span class="n">build_call</span> <span class="n">callee</span> <span class="o">[|</span><span class="n">lhs_val</span><span class="o">;</span> <span class="n">rhs_val</span><span class="o">|]</span> <span class="s2">"binop"</span> <span class="n">builder</span>
+ <span class="k">end</span>
+ <span class="k">end</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Call</span> <span class="o">(</span><span class="n">callee</span><span class="o">,</span> <span class="n">args</span><span class="o">)</span> <span class="o">-></span>
+ <span class="c">(* Look up the name in the module table. *)</span>
+ <span class="k">let</span> <span class="n">callee</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">lookup_function</span> <span class="n">callee</span> <span class="n">the_module</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">callee</span> <span class="o">-></span> <span class="n">callee</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"unknown function referenced"</span><span class="o">)</span>
+ <span class="k">in</span>
+ <span class="k">let</span> <span class="n">params</span> <span class="o">=</span> <span class="n">params</span> <span class="n">callee</span> <span class="k">in</span>
+
+ <span class="c">(* If argument mismatch error. *)</span>
+ <span class="k">if</span> <span class="nn">Array</span><span class="p">.</span><span class="n">length</span> <span class="n">params</span> <span class="o">==</span> <span class="nn">Array</span><span class="p">.</span><span class="n">length</span> <span class="n">args</span> <span class="k">then</span> <span class="bp">()</span> <span class="k">else</span>
+ <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"incorrect # arguments passed"</span><span class="o">);</span>
+ <span class="k">let</span> <span class="n">args</span> <span class="o">=</span> <span class="nn">Array</span><span class="p">.</span><span class="n">map</span> <span class="n">codegen_expr</span> <span class="n">args</span> <span class="k">in</span>
+ <span class="n">build_call</span> <span class="n">callee</span> <span class="n">args</span> <span class="s2">"calltmp"</span> <span class="n">builder</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">If</span> <span class="o">(</span><span class="n">cond</span><span class="o">,</span> <span class="n">then_</span><span class="o">,</span> <span class="n">else_</span><span class="o">)</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">cond</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">cond</span> <span class="k">in</span>
+
+ <span class="c">(* Convert condition to a bool by comparing equal to 0.0 *)</span>
+ <span class="k">let</span> <span class="n">zero</span> <span class="o">=</span> <span class="n">const_float</span> <span class="n">double_type</span> <span class="mi">0</span><span class="o">.</span><span class="mi">0</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">cond_val</span> <span class="o">=</span> <span class="n">build_fcmp</span> <span class="nn">Fcmp</span><span class="p">.</span><span class="nc">One</span> <span class="n">cond</span> <span class="n">zero</span> <span class="s2">"ifcond"</span> <span class="n">builder</span> <span class="k">in</span>
+
+ <span class="c">(* Grab the first block so that we might later add the conditional branch</span>
+<span class="c"> * to it at the end of the function. *)</span>
+ <span class="k">let</span> <span class="n">start_bb</span> <span class="o">=</span> <span class="n">insertion_block</span> <span class="n">builder</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">the_function</span> <span class="o">=</span> <span class="n">block_parent</span> <span class="n">start_bb</span> <span class="k">in</span>
+
+ <span class="k">let</span> <span class="n">then_bb</span> <span class="o">=</span> <span class="n">append_block</span> <span class="n">context</span> <span class="s2">"then"</span> <span class="n">the_function</span> <span class="k">in</span>
+
+ <span class="c">(* Emit 'then' value. *)</span>
+ <span class="n">position_at_end</span> <span class="n">then_bb</span> <span class="n">builder</span><span class="o">;</span>
+ <span class="k">let</span> <span class="n">then_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">then_</span> <span class="k">in</span>
+
+ <span class="c">(* Codegen of 'then' can change the current block, update then_bb for the</span>
+<span class="c"> * phi. We create a new name because one is used for the phi node, and the</span>
+<span class="c"> * other is used for the conditional branch. *)</span>
+ <span class="k">let</span> <span class="n">new_then_bb</span> <span class="o">=</span> <span class="n">insertion_block</span> <span class="n">builder</span> <span class="k">in</span>
+
+ <span class="c">(* Emit 'else' value. *)</span>
+ <span class="k">let</span> <span class="n">else_bb</span> <span class="o">=</span> <span class="n">append_block</span> <span class="n">context</span> <span class="s2">"else"</span> <span class="n">the_function</span> <span class="k">in</span>
+ <span class="n">position_at_end</span> <span class="n">else_bb</span> <span class="n">builder</span><span class="o">;</span>
+ <span class="k">let</span> <span class="n">else_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">else_</span> <span class="k">in</span>
+
+ <span class="c">(* Codegen of 'else' can change the current block, update else_bb for the</span>
+<span class="c"> * phi. *)</span>
+ <span class="k">let</span> <span class="n">new_else_bb</span> <span class="o">=</span> <span class="n">insertion_block</span> <span class="n">builder</span> <span class="k">in</span>
+
+ <span class="c">(* Emit merge block. *)</span>
+ <span class="k">let</span> <span class="n">merge_bb</span> <span class="o">=</span> <span class="n">append_block</span> <span class="n">context</span> <span class="s2">"ifcont"</span> <span class="n">the_function</span> <span class="k">in</span>
+ <span class="n">position_at_end</span> <span class="n">merge_bb</span> <span class="n">builder</span><span class="o">;</span>
+ <span class="k">let</span> <span class="n">incoming</span> <span class="o">=</span> <span class="o">[(</span><span class="n">then_val</span><span class="o">,</span> <span class="n">new_then_bb</span><span class="o">);</span> <span class="o">(</span><span class="n">else_val</span><span class="o">,</span> <span class="n">new_else_bb</span><span class="o">)]</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">phi</span> <span class="o">=</span> <span class="n">build_phi</span> <span class="n">incoming</span> <span class="s2">"iftmp"</span> <span class="n">builder</span> <span class="k">in</span>
+
+ <span class="c">(* Return to the start block to add the conditional branch. *)</span>
+ <span class="n">position_at_end</span> <span class="n">start_bb</span> <span class="n">builder</span><span class="o">;</span>
+ <span class="n">ignore</span> <span class="o">(</span><span class="n">build_cond_br</span> <span class="n">cond_val</span> <span class="n">then_bb</span> <span class="n">else_bb</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Set a unconditional branch at the end of the 'then' block and the</span>
+<span class="c"> * 'else' block to the 'merge' block. *)</span>
+ <span class="n">position_at_end</span> <span class="n">new_then_bb</span> <span class="n">builder</span><span class="o">;</span> <span class="n">ignore</span> <span class="o">(</span><span class="n">build_br</span> <span class="n">merge_bb</span> <span class="n">builder</span><span class="o">);</span>
+ <span class="n">position_at_end</span> <span class="n">new_else_bb</span> <span class="n">builder</span><span class="o">;</span> <span class="n">ignore</span> <span class="o">(</span><span class="n">build_br</span> <span class="n">merge_bb</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Finally, set the builder to the end of the merge block. *)</span>
+ <span class="n">position_at_end</span> <span class="n">merge_bb</span> <span class="n">builder</span><span class="o">;</span>
+
+ <span class="n">phi</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">For</span> <span class="o">(</span><span class="n">var_name</span><span class="o">,</span> <span class="n">start</span><span class="o">,</span> <span class="n">end_</span><span class="o">,</span> <span class="n">step</span><span class="o">,</span> <span class="n">body</span><span class="o">)</span> <span class="o">-></span>
+ <span class="c">(* Output this as:</span>
+<span class="c"> * var = alloca double</span>
+<span class="c"> * ...</span>
+<span class="c"> * start = startexpr</span>
+<span class="c"> * store start -> var</span>
+<span class="c"> * goto loop</span>
+<span class="c"> * loop:</span>
+<span class="c"> * ...</span>
+<span class="c"> * bodyexpr</span>
+<span class="c"> * ...</span>
+<span class="c"> * loopend:</span>
+<span class="c"> * step = stepexpr</span>
+<span class="c"> * endcond = endexpr</span>
+<span class="c"> *</span>
+<span class="c"> * curvar = load var</span>
+<span class="c"> * nextvar = curvar + step</span>
+<span class="c"> * store nextvar -> var</span>
+<span class="c"> * br endcond, loop, endloop</span>
+<span class="c"> * outloop: *)</span>
+
+ <span class="k">let</span> <span class="n">the_function</span> <span class="o">=</span> <span class="n">block_parent</span> <span class="o">(</span><span class="n">insertion_block</span> <span class="n">builder</span><span class="o">)</span> <span class="k">in</span>
+
+ <span class="c">(* Create an alloca for the variable in the entry block. *)</span>
+ <span class="k">let</span> <span class="n">alloca</span> <span class="o">=</span> <span class="n">create_entry_block_alloca</span> <span class="n">the_function</span> <span class="n">var_name</span> <span class="k">in</span>
+
+ <span class="c">(* Emit the start code first, without 'variable' in scope. *)</span>
+ <span class="k">let</span> <span class="n">start_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">start</span> <span class="k">in</span>
+
+ <span class="c">(* Store the value into the alloca. *)</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">start_val</span> <span class="n">alloca</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Make the new basic block for the loop header, inserting after current</span>
+<span class="c"> * block. *)</span>
+ <span class="k">let</span> <span class="n">loop_bb</span> <span class="o">=</span> <span class="n">append_block</span> <span class="n">context</span> <span class="s2">"loop"</span> <span class="n">the_function</span> <span class="k">in</span>
+
+ <span class="c">(* Insert an explicit fall through from the current block to the</span>
+<span class="c"> * loop_bb. *)</span>
+ <span class="n">ignore</span> <span class="o">(</span><span class="n">build_br</span> <span class="n">loop_bb</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Start insertion in loop_bb. *)</span>
+ <span class="n">position_at_end</span> <span class="n">loop_bb</span> <span class="n">builder</span><span class="o">;</span>
+
+ <span class="c">(* Within the loop, the variable is defined equal to the PHI node. If it</span>
+<span class="c"> * shadows an existing variable, we have to restore it, so save it</span>
+<span class="c"> * now. *)</span>
+ <span class="k">let</span> <span class="n">old_val</span> <span class="o">=</span>
+ <span class="k">try</span> <span class="nc">Some</span> <span class="o">(</span><span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">named_values</span> <span class="n">var_name</span><span class="o">)</span> <span class="k">with</span> <span class="nc">Not_found</span> <span class="o">-></span> <span class="nc">None</span>
+ <span class="k">in</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">alloca</span><span class="o">;</span>
+
+ <span class="c">(* Emit the body of the loop. This, like any other expr, can change the</span>
+<span class="c"> * current BB. Note that we ignore the value computed by the body, but</span>
+<span class="c"> * don't allow an error *)</span>
+ <span class="n">ignore</span> <span class="o">(</span><span class="n">codegen_expr</span> <span class="n">body</span><span class="o">);</span>
+
+ <span class="c">(* Emit the step value. *)</span>
+ <span class="k">let</span> <span class="n">step_val</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">step</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">step</span> <span class="o">-></span> <span class="n">codegen_expr</span> <span class="n">step</span>
+ <span class="c">(* If not specified, use 1.0. *)</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="n">const_float</span> <span class="n">double_type</span> <span class="mi">1</span><span class="o">.</span><span class="mi">0</span>
+ <span class="k">in</span>
+
+ <span class="c">(* Compute the end condition. *)</span>
+ <span class="k">let</span> <span class="n">end_cond</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">end_</span> <span class="k">in</span>
+
+ <span class="c">(* Reload, increment, and restore the alloca. This handles the case where</span>
+<span class="c"> * the body of the loop mutates the variable. *)</span>
+ <span class="k">let</span> <span class="n">cur_var</span> <span class="o">=</span> <span class="n">build_load</span> <span class="n">alloca</span> <span class="n">var_name</span> <span class="n">builder</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">next_var</span> <span class="o">=</span> <span class="n">build_add</span> <span class="n">cur_var</span> <span class="n">step_val</span> <span class="s2">"nextvar"</span> <span class="n">builder</span> <span class="k">in</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">next_var</span> <span class="n">alloca</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Convert condition to a bool by comparing equal to 0.0. *)</span>
+ <span class="k">let</span> <span class="n">zero</span> <span class="o">=</span> <span class="n">const_float</span> <span class="n">double_type</span> <span class="mi">0</span><span class="o">.</span><span class="mi">0</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">end_cond</span> <span class="o">=</span> <span class="n">build_fcmp</span> <span class="nn">Fcmp</span><span class="p">.</span><span class="nc">One</span> <span class="n">end_cond</span> <span class="n">zero</span> <span class="s2">"loopcond"</span> <span class="n">builder</span> <span class="k">in</span>
+
+ <span class="c">(* Create the "after loop" block and insert it. *)</span>
+ <span class="k">let</span> <span class="n">after_bb</span> <span class="o">=</span> <span class="n">append_block</span> <span class="n">context</span> <span class="s2">"afterloop"</span> <span class="n">the_function</span> <span class="k">in</span>
+
+ <span class="c">(* Insert the conditional branch into the end of loop_end_bb. *)</span>
+ <span class="n">ignore</span> <span class="o">(</span><span class="n">build_cond_br</span> <span class="n">end_cond</span> <span class="n">loop_bb</span> <span class="n">after_bb</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Any new code will be inserted in after_bb. *)</span>
+ <span class="n">position_at_end</span> <span class="n">after_bb</span> <span class="n">builder</span><span class="o">;</span>
+
+ <span class="c">(* Restore the unshadowed variable. *)</span>
+ <span class="k">begin</span> <span class="k">match</span> <span class="n">old_val</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">old_val</span> <span class="o">-></span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">old_val</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="bp">()</span>
+ <span class="k">end</span><span class="o">;</span>
+
+ <span class="c">(* for expr always returns 0.0. *)</span>
+ <span class="n">const_null</span> <span class="n">double_type</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Var</span> <span class="o">(</span><span class="n">var_names</span><span class="o">,</span> <span class="n">body</span><span class="o">)</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">old_bindings</span> <span class="o">=</span> <span class="n">ref</span> <span class="bp">[]</span> <span class="k">in</span>
+
+ <span class="k">let</span> <span class="n">the_function</span> <span class="o">=</span> <span class="n">block_parent</span> <span class="o">(</span><span class="n">insertion_block</span> <span class="n">builder</span><span class="o">)</span> <span class="k">in</span>
+
+ <span class="c">(* Register all variables and emit their initializer. *)</span>
+ <span class="nn">Array</span><span class="p">.</span><span class="n">iter</span> <span class="o">(</span><span class="k">fun</span> <span class="o">(</span><span class="n">var_name</span><span class="o">,</span> <span class="n">init</span><span class="o">)</span> <span class="o">-></span>
+ <span class="c">(* Emit the initializer before adding the variable to scope, this</span>
+<span class="c"> * prevents the initializer from referencing the variable itself, and</span>
+<span class="c"> * permits stuff like this:</span>
+<span class="c"> * var a = 1 in</span>
+<span class="c"> * var a = a in ... # refers to outer 'a'. *)</span>
+ <span class="k">let</span> <span class="n">init_val</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">init</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">init</span> <span class="o">-></span> <span class="n">codegen_expr</span> <span class="n">init</span>
+ <span class="c">(* If not specified, use 0.0. *)</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="n">const_float</span> <span class="n">double_type</span> <span class="mi">0</span><span class="o">.</span><span class="mi">0</span>
+ <span class="k">in</span>
+
+ <span class="k">let</span> <span class="n">alloca</span> <span class="o">=</span> <span class="n">create_entry_block_alloca</span> <span class="n">the_function</span> <span class="n">var_name</span> <span class="k">in</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">init_val</span> <span class="n">alloca</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Remember the old variable binding so that we can restore the binding</span>
+<span class="c"> * when we unrecurse. *)</span>
+ <span class="k">begin</span>
+ <span class="k">try</span>
+ <span class="k">let</span> <span class="n">old_value</span> <span class="o">=</span> <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">find</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="k">in</span>
+ <span class="n">old_bindings</span> <span class="o">:=</span> <span class="o">(</span><span class="n">var_name</span><span class="o">,</span> <span class="n">old_value</span><span class="o">)</span> <span class="o">::</span> <span class="o">!</span><span class="n">old_bindings</span><span class="o">;</span>
+ <span class="k">with</span> <span class="nc">Not_found</span> <span class="o">-></span> <span class="bp">()</span>
+ <span class="k">end</span><span class="o">;</span>
+
+ <span class="c">(* Remember this binding. *)</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">alloca</span><span class="o">;</span>
+ <span class="o">)</span> <span class="n">var_names</span><span class="o">;</span>
+
+ <span class="c">(* Codegen the body, now that all vars are in scope. *)</span>
+ <span class="k">let</span> <span class="n">body_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">body</span> <span class="k">in</span>
+
+ <span class="c">(* Pop all our variables from scope. *)</span>
+ <span class="nn">List</span><span class="p">.</span><span class="n">iter</span> <span class="o">(</span><span class="k">fun</span> <span class="o">(</span><span class="n">var_name</span><span class="o">,</span> <span class="n">old_value</span><span class="o">)</span> <span class="o">-></span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">old_value</span>
+ <span class="o">)</span> <span class="o">!</span><span class="n">old_bindings</span><span class="o">;</span>
+
+ <span class="c">(* Return the body computation. *)</span>
+ <span class="n">body_val</span>
+
+<span class="k">let</span> <span class="n">codegen_proto</span> <span class="o">=</span> <span class="k">function</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Prototype</span> <span class="o">(</span><span class="n">name</span><span class="o">,</span> <span class="n">args</span><span class="o">)</span> <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">BinOpPrototype</span> <span class="o">(</span><span class="n">name</span><span class="o">,</span> <span class="n">args</span><span class="o">,</span> <span class="o">_)</span> <span class="o">-></span>
+ <span class="c">(* Make the function type: double(double,double) etc. *)</span>
+ <span class="k">let</span> <span class="n">doubles</span> <span class="o">=</span> <span class="nn">Array</span><span class="p">.</span><span class="n">make</span> <span class="o">(</span><span class="nn">Array</span><span class="p">.</span><span class="n">length</span> <span class="n">args</span><span class="o">)</span> <span class="n">double_type</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">ft</span> <span class="o">=</span> <span class="n">function_type</span> <span class="n">double_type</span> <span class="n">doubles</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">f</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="n">lookup_function</span> <span class="n">name</span> <span class="n">the_module</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="n">declare_function</span> <span class="n">name</span> <span class="n">ft</span> <span class="n">the_module</span>
+
+ <span class="c">(* If 'f' conflicted, there was already something named 'name'. If it</span>
+<span class="c"> * has a body, don't allow redefinition or reextern. *)</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">f</span> <span class="o">-></span>
+ <span class="c">(* If 'f' already has a body, reject this. *)</span>
+ <span class="k">if</span> <span class="n">block_begin</span> <span class="n">f</span> <span class="o"><></span> <span class="nc">At_end</span> <span class="n">f</span> <span class="k">then</span>
+ <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"redefinition of function"</span><span class="o">);</span>
+
+ <span class="c">(* If 'f' took a different number of arguments, reject. *)</span>
+ <span class="k">if</span> <span class="n">element_type</span> <span class="o">(</span><span class="n">type_of</span> <span class="n">f</span><span class="o">)</span> <span class="o"><></span> <span class="n">ft</span> <span class="k">then</span>
+ <span class="k">raise</span> <span class="o">(</span><span class="nc">Error</span> <span class="s2">"redefinition of function with different # args"</span><span class="o">);</span>
+ <span class="n">f</span>
+ <span class="k">in</span>
+
+ <span class="c">(* Set names for all arguments. *)</span>
+ <span class="nn">Array</span><span class="p">.</span><span class="n">iteri</span> <span class="o">(</span><span class="k">fun</span> <span class="n">i</span> <span class="n">a</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">n</span> <span class="o">=</span> <span class="n">args</span><span class="o">.(</span><span class="n">i</span><span class="o">)</span> <span class="k">in</span>
+ <span class="n">set_value_name</span> <span class="n">n</span> <span class="n">a</span><span class="o">;</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">n</span> <span class="n">a</span><span class="o">;</span>
+ <span class="o">)</span> <span class="o">(</span><span class="n">params</span> <span class="n">f</span><span class="o">);</span>
+ <span class="n">f</span>
+
+<span class="c">(* Create an alloca for each argument and register the argument in the symbol</span>
+<span class="c"> * table so that references to it will succeed. *)</span>
+<span class="k">let</span> <span class="n">create_argument_allocas</span> <span class="n">the_function</span> <span class="n">proto</span> <span class="o">=</span>
+ <span class="k">let</span> <span class="n">args</span> <span class="o">=</span> <span class="k">match</span> <span class="n">proto</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Prototype</span> <span class="o">(_,</span> <span class="n">args</span><span class="o">)</span> <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">BinOpPrototype</span> <span class="o">(_,</span> <span class="n">args</span><span class="o">,</span> <span class="o">_)</span> <span class="o">-></span> <span class="n">args</span>
+ <span class="k">in</span>
+ <span class="nn">Array</span><span class="p">.</span><span class="n">iteri</span> <span class="o">(</span><span class="k">fun</span> <span class="n">i</span> <span class="n">ai</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">var_name</span> <span class="o">=</span> <span class="n">args</span><span class="o">.(</span><span class="n">i</span><span class="o">)</span> <span class="k">in</span>
+ <span class="c">(* Create an alloca for this variable. *)</span>
+ <span class="k">let</span> <span class="n">alloca</span> <span class="o">=</span> <span class="n">create_entry_block_alloca</span> <span class="n">the_function</span> <span class="n">var_name</span> <span class="k">in</span>
+
+ <span class="c">(* Store the initial value into the alloca. *)</span>
+ <span class="n">ignore</span><span class="o">(</span><span class="n">build_store</span> <span class="n">ai</span> <span class="n">alloca</span> <span class="n">builder</span><span class="o">);</span>
+
+ <span class="c">(* Add arguments to variable symbol table. *)</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="n">named_values</span> <span class="n">var_name</span> <span class="n">alloca</span><span class="o">;</span>
+ <span class="o">)</span> <span class="o">(</span><span class="n">params</span> <span class="n">the_function</span><span class="o">)</span>
+
+<span class="k">let</span> <span class="n">codegen_func</span> <span class="n">the_fpm</span> <span class="o">=</span> <span class="k">function</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">Function</span> <span class="o">(</span><span class="n">proto</span><span class="o">,</span> <span class="n">body</span><span class="o">)</span> <span class="o">-></span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">clear</span> <span class="n">named_values</span><span class="o">;</span>
+ <span class="k">let</span> <span class="n">the_function</span> <span class="o">=</span> <span class="n">codegen_proto</span> <span class="n">proto</span> <span class="k">in</span>
+
+ <span class="c">(* If this is an operator, install it. *)</span>
+ <span class="k">begin</span> <span class="k">match</span> <span class="n">proto</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nn">Ast</span><span class="p">.</span><span class="nc">BinOpPrototype</span> <span class="o">(</span><span class="n">name</span><span class="o">,</span> <span class="n">args</span><span class="o">,</span> <span class="n">prec</span><span class="o">)</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">op</span> <span class="o">=</span> <span class="n">name</span><span class="o">.[</span><span class="nn">String</span><span class="p">.</span><span class="n">length</span> <span class="n">name</span> <span class="o">-</span> <span class="mi">1</span><span class="o">]</span> <span class="k">in</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="n">op</span> <span class="n">prec</span><span class="o">;</span>
+ <span class="o">|</span> <span class="o">_</span> <span class="o">-></span> <span class="bp">()</span>
+ <span class="k">end</span><span class="o">;</span>
+
+ <span class="c">(* Create a new basic block to start insertion into. *)</span>
+ <span class="k">let</span> <span class="n">bb</span> <span class="o">=</span> <span class="n">append_block</span> <span class="n">context</span> <span class="s2">"entry"</span> <span class="n">the_function</span> <span class="k">in</span>
+ <span class="n">position_at_end</span> <span class="n">bb</span> <span class="n">builder</span><span class="o">;</span>
+
+ <span class="k">try</span>
+ <span class="c">(* Add all arguments to the symbol table and create their allocas. *)</span>
+ <span class="n">create_argument_allocas</span> <span class="n">the_function</span> <span class="n">proto</span><span class="o">;</span>
+
+ <span class="k">let</span> <span class="n">ret_val</span> <span class="o">=</span> <span class="n">codegen_expr</span> <span class="n">body</span> <span class="k">in</span>
+
+ <span class="c">(* Finish off the function. *)</span>
+ <span class="k">let</span> <span class="o">_</span> <span class="o">=</span> <span class="n">build_ret</span> <span class="n">ret_val</span> <span class="n">builder</span> <span class="k">in</span>
+
+ <span class="c">(* Validate the generated code, checking for consistency. *)</span>
+ <span class="nn">Llvm_analysis</span><span class="p">.</span><span class="n">assert_valid_function</span> <span class="n">the_function</span><span class="o">;</span>
+
+ <span class="c">(* Optimize the function. *)</span>
+ <span class="k">let</span> <span class="o">_</span> <span class="o">=</span> <span class="nn">PassManager</span><span class="p">.</span><span class="n">run_function</span> <span class="n">the_function</span> <span class="n">the_fpm</span> <span class="k">in</span>
+
+ <span class="n">the_function</span>
+ <span class="k">with</span> <span class="n">e</span> <span class="o">-></span>
+ <span class="n">delete_function</span> <span class="n">the_function</span><span class="o">;</span>
+ <span class="k">raise</span> <span class="n">e</span>
+</pre></div>
+</div>
+</dd>
+<dt>toplevel.ml:</dt>
+<dd><div class="first last highlight-ocaml"><div class="highlight"><pre><span class="c">(*===----------------------------------------------------------------------===</span>
+<span class="c"> * Top-Level parsing and JIT Driver</span>
+<span class="c"> *===----------------------------------------------------------------------===*)</span>
+
+<span class="k">open</span> <span class="nc">Llvm</span>
+<span class="k">open</span> <span class="nc">Llvm_executionengine</span>
+
+<span class="c">(* top ::= definition | external | expression | ';' *)</span>
+<span class="k">let</span> <span class="k">rec</span> <span class="n">main_loop</span> <span class="n">the_fpm</span> <span class="n">the_execution_engine</span> <span class="n">stream</span> <span class="o">=</span>
+ <span class="k">match</span> <span class="nn">Stream</span><span class="p">.</span><span class="n">peek</span> <span class="n">stream</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nc">None</span> <span class="o">-></span> <span class="bp">()</span>
+
+ <span class="c">(* ignore top-level semicolons. *)</span>
+ <span class="o">|</span> <span class="nc">Some</span> <span class="o">(</span><span class="nn">Token</span><span class="p">.</span><span class="nc">Kwd</span> <span class="sc">';'</span><span class="o">)</span> <span class="o">-></span>
+ <span class="nn">Stream</span><span class="p">.</span><span class="n">junk</span> <span class="n">stream</span><span class="o">;</span>
+ <span class="n">main_loop</span> <span class="n">the_fpm</span> <span class="n">the_execution_engine</span> <span class="n">stream</span>
+
+ <span class="o">|</span> <span class="nc">Some</span> <span class="n">token</span> <span class="o">-></span>
+ <span class="k">begin</span>
+ <span class="k">try</span> <span class="k">match</span> <span class="n">token</span> <span class="k">with</span>
+ <span class="o">|</span> <span class="nn">Token</span><span class="p">.</span><span class="nc">Def</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">e</span> <span class="o">=</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">parse_definition</span> <span class="n">stream</span> <span class="k">in</span>
+ <span class="n">print_endline</span> <span class="s2">"parsed a function definition."</span><span class="o">;</span>
+ <span class="n">dump_value</span> <span class="o">(</span><span class="nn">Codegen</span><span class="p">.</span><span class="n">codegen_func</span> <span class="n">the_fpm</span> <span class="n">e</span><span class="o">);</span>
+ <span class="o">|</span> <span class="nn">Token</span><span class="p">.</span><span class="nc">Extern</span> <span class="o">-></span>
+ <span class="k">let</span> <span class="n">e</span> <span class="o">=</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">parse_extern</span> <span class="n">stream</span> <span class="k">in</span>
+ <span class="n">print_endline</span> <span class="s2">"parsed an extern."</span><span class="o">;</span>
+ <span class="n">dump_value</span> <span class="o">(</span><span class="nn">Codegen</span><span class="p">.</span><span class="n">codegen_proto</span> <span class="n">e</span><span class="o">);</span>
+ <span class="o">|</span> <span class="o">_</span> <span class="o">-></span>
+ <span class="c">(* Evaluate a top-level expression into an anonymous function. *)</span>
+ <span class="k">let</span> <span class="n">e</span> <span class="o">=</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">parse_toplevel</span> <span class="n">stream</span> <span class="k">in</span>
+ <span class="n">print_endline</span> <span class="s2">"parsed a top-level expr"</span><span class="o">;</span>
+ <span class="k">let</span> <span class="n">the_function</span> <span class="o">=</span> <span class="nn">Codegen</span><span class="p">.</span><span class="n">codegen_func</span> <span class="n">the_fpm</span> <span class="n">e</span> <span class="k">in</span>
+ <span class="n">dump_value</span> <span class="n">the_function</span><span class="o">;</span>
+
+ <span class="c">(* JIT the function, returning a function pointer. *)</span>
+ <span class="k">let</span> <span class="n">result</span> <span class="o">=</span> <span class="nn">ExecutionEngine</span><span class="p">.</span><span class="n">run_function</span> <span class="n">the_function</span> <span class="o">[||]</span>
+ <span class="n">the_execution_engine</span> <span class="k">in</span>
+
+ <span class="n">print_string</span> <span class="s2">"Evaluated to "</span><span class="o">;</span>
+ <span class="n">print_float</span> <span class="o">(</span><span class="nn">GenericValue</span><span class="p">.</span><span class="n">as_float</span> <span class="nn">Codegen</span><span class="p">.</span><span class="n">double_type</span> <span class="n">result</span><span class="o">);</span>
+ <span class="n">print_newline</span> <span class="bp">()</span><span class="o">;</span>
+ <span class="k">with</span> <span class="nn">Stream</span><span class="p">.</span><span class="nc">Error</span> <span class="n">s</span> <span class="o">|</span> <span class="nn">Codegen</span><span class="p">.</span><span class="nc">Error</span> <span class="n">s</span> <span class="o">-></span>
+ <span class="c">(* Skip token for error recovery. *)</span>
+ <span class="nn">Stream</span><span class="p">.</span><span class="n">junk</span> <span class="n">stream</span><span class="o">;</span>
+ <span class="n">print_endline</span> <span class="n">s</span><span class="o">;</span>
+ <span class="k">end</span><span class="o">;</span>
+ <span class="n">print_string</span> <span class="s2">"ready> "</span><span class="o">;</span> <span class="n">flush</span> <span class="n">stdout</span><span class="o">;</span>
+ <span class="n">main_loop</span> <span class="n">the_fpm</span> <span class="n">the_execution_engine</span> <span class="n">stream</span>
+</pre></div>
+</div>
+</dd>
+<dt>toy.ml:</dt>
+<dd><div class="first last highlight-ocaml"><div class="highlight"><pre><span class="c">(*===----------------------------------------------------------------------===</span>
+<span class="c"> * Main driver code.</span>
+<span class="c"> *===----------------------------------------------------------------------===*)</span>
+
+<span class="k">open</span> <span class="nc">Llvm</span>
+<span class="k">open</span> <span class="nc">Llvm_executionengine</span>
+<span class="k">open</span> <span class="nc">Llvm_target</span>
+<span class="k">open</span> <span class="nc">Llvm_scalar_opts</span>
+
+<span class="k">let</span> <span class="n">main</span> <span class="bp">()</span> <span class="o">=</span>
+ <span class="n">ignore</span> <span class="o">(</span><span class="n">initialize_native_target</span> <span class="bp">()</span><span class="o">);</span>
+
+ <span class="c">(* Install standard binary operators.</span>
+<span class="c"> * 1 is the lowest precedence. *)</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'='</span> <span class="mi">2</span><span class="o">;</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'<'</span> <span class="mi">10</span><span class="o">;</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'+'</span> <span class="mi">20</span><span class="o">;</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'-'</span> <span class="mi">20</span><span class="o">;</span>
+ <span class="nn">Hashtbl</span><span class="p">.</span><span class="n">add</span> <span class="nn">Parser</span><span class="p">.</span><span class="n">binop_precedence</span> <span class="sc">'*'</span> <span class="mi">40</span><span class="o">;</span> <span class="c">(* highest. *)</span>
+
+ <span class="c">(* Prime the first token. *)</span>
+ <span class="n">print_string</span> <span class="s2">"ready> "</span><span class="o">;</span> <span class="n">flush</span> <span class="n">stdout</span><span class="o">;</span>
+ <span class="k">let</span> <span class="n">stream</span> <span class="o">=</span> <span class="nn">Lexer</span><span class="p">.</span><span class="n">lex</span> <span class="o">(</span><span class="nn">Stream</span><span class="p">.</span><span class="n">of_channel</span> <span class="n">stdin</span><span class="o">)</span> <span class="k">in</span>
+
+ <span class="c">(* Create the JIT. *)</span>
+ <span class="k">let</span> <span class="n">the_execution_engine</span> <span class="o">=</span> <span class="nn">ExecutionEngine</span><span class="p">.</span><span class="n">create</span> <span class="nn">Codegen</span><span class="p">.</span><span class="n">the_module</span> <span class="k">in</span>
+ <span class="k">let</span> <span class="n">the_fpm</span> <span class="o">=</span> <span class="nn">PassManager</span><span class="p">.</span><span class="n">create_function</span> <span class="nn">Codegen</span><span class="p">.</span><span class="n">the_module</span> <span class="k">in</span>
+
+ <span class="c">(* Set up the optimizer pipeline. Start with registering info about how the</span>
+<span class="c"> * target lays out data structures. *)</span>
+ <span class="nn">DataLayout</span><span class="p">.</span><span class="n">add</span> <span class="o">(</span><span class="nn">ExecutionEngine</span><span class="p">.</span><span class="n">target_data</span> <span class="n">the_execution_engine</span><span class="o">)</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="c">(* Promote allocas to registers. *)</span>
+ <span class="n">add_memory_to_register_promotion</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="c">(* Do simple "peephole" optimizations and bit-twiddling optzn. *)</span>
+ <span class="n">add_instruction_combination</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="c">(* reassociate expressions. *)</span>
+ <span class="n">add_reassociation</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="c">(* Eliminate Common SubExpressions. *)</span>
+ <span class="n">add_gvn</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="c">(* Simplify the control flow graph (deleting unreachable blocks, etc). *)</span>
+ <span class="n">add_cfg_simplification</span> <span class="n">the_fpm</span><span class="o">;</span>
+
+ <span class="n">ignore</span> <span class="o">(</span><span class="nn">PassManager</span><span class="p">.</span><span class="n">initialize</span> <span class="n">the_fpm</span><span class="o">);</span>
+
+ <span class="c">(* Run the main "interpreter loop" now. *)</span>
+ <span class="nn">Toplevel</span><span class="p">.</span><span class="n">main_loop</span> <span class="n">the_fpm</span> <span class="n">the_execution_engine</span> <span class="n">stream</span><span class="o">;</span>
+
+ <span class="c">(* Print out all the generated code. *)</span>
+ <span class="n">dump_module</span> <span class="nn">Codegen</span><span class="p">.</span><span class="n">the_module</span>
+<span class="o">;;</span>
+
+<span class="n">main</span> <span class="bp">()</span>
+</pre></div>
+</div>
+</dd>
+<dt>bindings.c</dt>
+<dd><div class="first last highlight-c"><div class="highlight"><pre><span class="cp">#include <stdio.h></span>
+
+<span class="cm">/* putchard - putchar that takes a double and returns 0. */</span>
+<span class="k">extern</span> <span class="kt">double</span> <span class="nf">putchard</span><span class="p">(</span><span class="kt">double</span> <span class="n">X</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">putchar</span><span class="p">((</span><span class="kt">char</span><span class="p">)</span><span class="n">X</span><span class="p">);</span>
+ <span class="k">return</span> <span class="mi">0</span><span class="p">;</span>
+<span class="p">}</span>
+
+<span class="cm">/* printd - printf that takes a double prints it as "%f\n", returning 0. */</span>
+<span class="k">extern</span> <span class="kt">double</span> <span class="nf">printd</span><span class="p">(</span><span class="kt">double</span> <span class="n">X</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">printf</span><span class="p">(</span><span class="s">"%f</span><span class="se">\n</span><span class="s">"</span><span class="p">,</span> <span class="n">X</span><span class="p">);</span>
+ <span class="k">return</span> <span class="mi">0</span><span class="p">;</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</dd>
+</dl>
+<p><a class="reference external" href="OCamlLangImpl8.html">Next: Conclusion and other useful LLVM tidbits</a></p>
+</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="OCamlLangImpl8.html" title="8. Kaleidoscope: Conclusion and other useful LLVM tidbits"
+ >next</a> |</li>
+ <li class="right" >
+ <a href="OCamlLangImpl6.html" title="6. Kaleidoscope: Extending the Language: User-defined Operators"
+ >previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="../index.html">Documentation</a>»</li>
+
+ <li><a href="index.html" >LLVM Tutorial: Table of Contents</a> »</li>
+ </ul>
+ </div>
+ <div class="footer">
+ © Copyright 2003-2018, LLVM Project.
+ Last updated on 2018-12-21.
+ Created using <a href="http://sphinx.pocoo.org/">Sphinx</a> 1.1.3.
+ </div>
+ </body>
+</html>
\ No newline at end of file
Added: www-releases/trunk/7.0.1/docs/tutorial/OCamlLangImpl8.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/7.0.1/docs/tutorial/OCamlLangImpl8.html?rev=349965&view=auto
==============================================================================
--- www-releases/trunk/7.0.1/docs/tutorial/OCamlLangImpl8.html (added)
+++ www-releases/trunk/7.0.1/docs/tutorial/OCamlLangImpl8.html Fri Dec 21 13:53:02 2018
@@ -0,0 +1,353 @@
+
+
+<!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>8. Kaleidoscope: Conclusion and other useful LLVM tidbits — LLVM 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: '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 7 documentation" href="../index.html" />
+ <link rel="up" title="LLVM Tutorial: Table of Contents" href="index.html" />
+ <link rel="next" title="1. Building a JIT: Starting out with KaleidoscopeJIT" href="BuildingAJIT1.html" />
+ <link rel="prev" title="7. Kaleidoscope: Extending the Language: Mutable Variables" href="OCamlLangImpl7.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="BuildingAJIT1.html" title="1. Building a JIT: Starting out with KaleidoscopeJIT"
+ accesskey="N">next</a> |</li>
+ <li class="right" >
+ <a href="OCamlLangImpl7.html" title="7. Kaleidoscope: Extending the Language: Mutable Variables"
+ accesskey="P">previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="../index.html">Documentation</a>»</li>
+
+ <li><a href="index.html" accesskey="U">LLVM Tutorial: Table of Contents</a> »</li>
+ </ul>
+ </div>
+
+
+ <div class="document">
+ <div class="documentwrapper">
+ <div class="body">
+
+ <div class="section" id="kaleidoscope-conclusion-and-other-useful-llvm-tidbits">
+<h1>8. Kaleidoscope: Conclusion and other useful LLVM tidbits<a class="headerlink" href="#kaleidoscope-conclusion-and-other-useful-llvm-tidbits" title="Permalink to this headline">¶</a></h1>
+<div class="contents local topic" id="contents">
+<ul class="simple">
+<li><a class="reference internal" href="#tutorial-conclusion" id="id2">Tutorial Conclusion</a></li>
+<li><a class="reference internal" href="#properties-of-the-llvm-ir" id="id3">Properties of the LLVM IR</a><ul>
+<li><a class="reference internal" href="#target-independence" id="id4">Target Independence</a></li>
+<li><a class="reference internal" href="#safety-guarantees" id="id5">Safety Guarantees</a></li>
+<li><a class="reference internal" href="#language-specific-optimizations" id="id6">Language-Specific Optimizations</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#tips-and-tricks" id="id7">Tips and Tricks</a><ul>
+<li><a class="reference internal" href="#implementing-portable-offsetof-sizeof" id="id8">Implementing portable offsetof/sizeof</a></li>
+<li><a class="reference internal" href="#garbage-collected-stack-frames" id="id9">Garbage Collected Stack Frames</a></li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="section" id="tutorial-conclusion">
+<h2><a class="toc-backref" href="#id2">8.1. Tutorial Conclusion</a><a class="headerlink" href="#tutorial-conclusion" title="Permalink to this headline">¶</a></h2>
+<p>Welcome to the final chapter of the “<a class="reference external" href="index.html">Implementing a language with
+LLVM</a>” tutorial. In the course of this tutorial, we have
+grown our little Kaleidoscope language from being a useless toy, to
+being a semi-interesting (but probably still useless) toy. :)</p>
+<p>It is interesting to see how far we’ve come, and how little code it has
+taken. We built the entire lexer, parser, AST, code generator, and an
+interactive run-loop (with a JIT!) by-hand in under 700 lines of
+(non-comment/non-blank) code.</p>
+<p>Our little language supports a couple of interesting features: it
+supports user defined binary and unary operators, it uses JIT
+compilation for immediate evaluation, and it supports a few control flow
+constructs with SSA construction.</p>
+<p>Part of the idea of this tutorial was to show you how easy and fun it
+can be to define, build, and play with languages. Building a compiler
+need not be a scary or mystical process! Now that you’ve seen some of
+the basics, I strongly encourage you to take the code and hack on it.
+For example, try adding:</p>
+<ul class="simple">
+<li><strong>global variables</strong> - While global variables have questional value
+in modern software engineering, they are often useful when putting
+together quick little hacks like the Kaleidoscope compiler itself.
+Fortunately, our current setup makes it very easy to add global
+variables: just have value lookup check to see if an unresolved
+variable is in the global variable symbol table before rejecting it.
+To create a new global variable, make an instance of the LLVM
+<tt class="docutils literal"><span class="pre">GlobalVariable</span></tt> class.</li>
+<li><strong>typed variables</strong> - Kaleidoscope currently only supports variables
+of type double. This gives the language a very nice elegance, because
+only supporting one type means that you never have to specify types.
+Different languages have different ways of handling this. The easiest
+way is to require the user to specify types for every variable
+definition, and record the type of the variable in the symbol table
+along with its Value*.</li>
+<li><strong>arrays, structs, vectors, etc</strong> - Once you add types, you can start
+extending the type system in all sorts of interesting ways. Simple
+arrays are very easy and are quite useful for many different
+applications. Adding them is mostly an exercise in learning how the
+LLVM <a class="reference external" href="../LangRef.html#getelementptr-instruction">getelementptr</a> instruction
+works: it is so nifty/unconventional, it <a class="reference external" href="../GetElementPtr.html">has its own
+FAQ</a>! If you add support for recursive types
+(e.g. linked lists), make sure to read the <a class="reference external" href="../ProgrammersManual.html#TypeResolve">section in the LLVM
+Programmer’s Manual</a> that
+describes how to construct them.</li>
+<li><strong>standard runtime</strong> - Our current language allows the user to access
+arbitrary external functions, and we use it for things like “printd”
+and “putchard”. As you extend the language to add higher-level
+constructs, often these constructs make the most sense if they are
+lowered to calls into a language-supplied runtime. For example, if
+you add hash tables to the language, it would probably make sense to
+add the routines to a runtime, instead of inlining them all the way.</li>
+<li><strong>memory management</strong> - Currently we can only access the stack in
+Kaleidoscope. It would also be useful to be able to allocate heap
+memory, either with calls to the standard libc malloc/free interface
+or with a garbage collector. If you would like to use garbage
+collection, note that LLVM fully supports <a class="reference external" href="../GarbageCollection.html">Accurate Garbage
+Collection</a> including algorithms that
+move objects and need to scan/update the stack.</li>
+<li><strong>debugger support</strong> - LLVM supports generation of <a class="reference external" href="../SourceLevelDebugging.html">DWARF Debug
+info</a> which is understood by common
+debuggers like GDB. Adding support for debug info is fairly
+straightforward. The best way to understand it is to compile some
+C/C++ code with “<tt class="docutils literal"><span class="pre">clang</span> <span class="pre">-g</span> <span class="pre">-O0</span></tt>” and taking a look at what it
+produces.</li>
+<li><strong>exception handling support</strong> - LLVM supports generation of <a class="reference external" href="../ExceptionHandling.html">zero
+cost exceptions</a> which interoperate with
+code compiled in other languages. You could also generate code by
+implicitly making every function return an error value and checking
+it. You could also make explicit use of setjmp/longjmp. There are
+many different ways to go here.</li>
+<li><strong>object orientation, generics, database access, complex numbers,
+geometric programming, ...</strong> - Really, there is no end of crazy
+features that you can add to the language.</li>
+<li><strong>unusual domains</strong> - We’ve been talking about applying LLVM to a
+domain that many people are interested in: building a compiler for a
+specific language. However, there are many other domains that can use
+compiler technology that are not typically considered. For example,
+LLVM has been used to implement OpenGL graphics acceleration,
+translate C++ code to ActionScript, and many other cute and clever
+things. Maybe you will be the first to JIT compile a regular
+expression interpreter into native code with LLVM?</li>
+</ul>
+<p>Have fun - try doing something crazy and unusual. Building a language
+like everyone else always has, is much less fun than trying something a
+little crazy or off the wall and seeing how it turns out. If you get
+stuck or want to talk about it, feel free to email the <a class="reference external" href="http://lists.llvm.org/mailman/listinfo/llvm-dev">llvm-dev mailing
+list</a>: it has lots
+of people who are interested in languages and are often willing to help
+out.</p>
+<p>Before we end this tutorial, I want to talk about some “tips and tricks”
+for generating LLVM IR. These are some of the more subtle things that
+may not be obvious, but are very useful if you want to take advantage of
+LLVM’s capabilities.</p>
+</div>
+<div class="section" id="properties-of-the-llvm-ir">
+<h2><a class="toc-backref" href="#id3">8.2. Properties of the LLVM IR</a><a class="headerlink" href="#properties-of-the-llvm-ir" title="Permalink to this headline">¶</a></h2>
+<p>We have a couple common questions about code in the LLVM IR form - lets
+just get these out of the way right now, shall we?</p>
+<div class="section" id="target-independence">
+<h3><a class="toc-backref" href="#id4">8.2.1. Target Independence</a><a class="headerlink" href="#target-independence" title="Permalink to this headline">¶</a></h3>
+<p>Kaleidoscope is an example of a “portable language”: any program written
+in Kaleidoscope will work the same way on any target that it runs on.
+Many other languages have this property, e.g. lisp, java, haskell,
+javascript, python, etc (note that while these languages are portable,
+not all their libraries are).</p>
+<p>One nice aspect of LLVM is that it is often capable of preserving target
+independence in the IR: you can take the LLVM IR for a
+Kaleidoscope-compiled program and run it on any target that LLVM
+supports, even emitting C code and compiling that on targets that LLVM
+doesn’t support natively. You can trivially tell that the Kaleidoscope
+compiler generates target-independent code because it never queries for
+any target-specific information when generating code.</p>
+<p>The fact that LLVM provides a compact, target-independent,
+representation for code gets a lot of people excited. Unfortunately,
+these people are usually thinking about C or a language from the C
+family when they are asking questions about language portability. I say
+“unfortunately”, because there is really no way to make (fully general)
+C code portable, other than shipping the source code around (and of
+course, C source code is not actually portable in general either - ever
+port a really old application from 32- to 64-bits?).</p>
+<p>The problem with C (again, in its full generality) is that it is heavily
+laden with target specific assumptions. As one simple example, the
+preprocessor often destructively removes target-independence from the
+code when it processes the input text:</p>
+<div class="highlight-c"><div class="highlight"><pre><span class="cp">#ifdef __i386__</span>
+ <span class="kt">int</span> <span class="n">X</span> <span class="o">=</span> <span class="mi">1</span><span class="p">;</span>
+<span class="cp">#else</span>
+ <span class="kt">int</span> <span class="n">X</span> <span class="o">=</span> <span class="mi">42</span><span class="p">;</span>
+<span class="cp">#endif</span>
+</pre></div>
+</div>
+<p>While it is possible to engineer more and more complex solutions to
+problems like this, it cannot be solved in full generality in a way that
+is better than shipping the actual source code.</p>
+<p>That said, there are interesting subsets of C that can be made portable.
+If you are willing to fix primitive types to a fixed size (say int =
+32-bits, and long = 64-bits), don’t care about ABI compatibility with
+existing binaries, and are willing to give up some other minor features,
+you can have portable code. This can make sense for specialized domains
+such as an in-kernel language.</p>
+</div>
+<div class="section" id="safety-guarantees">
+<h3><a class="toc-backref" href="#id5">8.2.2. Safety Guarantees</a><a class="headerlink" href="#safety-guarantees" title="Permalink to this headline">¶</a></h3>
+<p>Many of the languages above are also “safe” languages: it is impossible
+for a program written in Java to corrupt its address space and crash the
+process (assuming the JVM has no bugs). Safety is an interesting
+property that requires a combination of language design, runtime
+support, and often operating system support.</p>
+<p>It is certainly possible to implement a safe language in LLVM, but LLVM
+IR does not itself guarantee safety. The LLVM IR allows unsafe pointer
+casts, use after free bugs, buffer over-runs, and a variety of other
+problems. Safety needs to be implemented as a layer on top of LLVM and,
+conveniently, several groups have investigated this. Ask on the <a class="reference external" href="http://lists.llvm.org/mailman/listinfo/llvm-dev">llvm-dev
+mailing list</a> if
+you are interested in more details.</p>
+</div>
+<div class="section" id="language-specific-optimizations">
+<h3><a class="toc-backref" href="#id6">8.2.3. Language-Specific Optimizations</a><a class="headerlink" href="#language-specific-optimizations" title="Permalink to this headline">¶</a></h3>
+<p>One thing about LLVM that turns off many people is that it does not
+solve all the world’s problems in one system (sorry ‘world hunger’,
+someone else will have to solve you some other day). One specific
+complaint is that people perceive LLVM as being incapable of performing
+high-level language-specific optimization: LLVM “loses too much
+information”.</p>
+<p>Unfortunately, this is really not the place to give you a full and
+unified version of “Chris Lattner’s theory of compiler design”. Instead,
+I’ll make a few observations:</p>
+<p>First, you’re right that LLVM does lose information. For example, as of
+this writing, there is no way to distinguish in the LLVM IR whether an
+SSA-value came from a C “int” or a C “long” on an ILP32 machine (other
+than debug info). Both get compiled down to an ‘i32’ value and the
+information about what it came from is lost. The more general issue
+here, is that the LLVM type system uses “structural equivalence” instead
+of “name equivalence”. Another place this surprises people is if you
+have two types in a high-level language that have the same structure
+(e.g. two different structs that have a single int field): these types
+will compile down into a single LLVM type and it will be impossible to
+tell what it came from.</p>
+<p>Second, while LLVM does lose information, LLVM is not a fixed target: we
+continue to enhance and improve it in many different ways. In addition
+to adding new features (LLVM did not always support exceptions or debug
+info), we also extend the IR to capture important information for
+optimization (e.g. whether an argument is sign or zero extended,
+information about pointers aliasing, etc). Many of the enhancements are
+user-driven: people want LLVM to include some specific feature, so they
+go ahead and extend it.</p>
+<p>Third, it is <em>possible and easy</em> to add language-specific optimizations,
+and you have a number of choices in how to do it. As one trivial
+example, it is easy to add language-specific optimization passes that
+“know” things about code compiled for a language. In the case of the C
+family, there is an optimization pass that “knows” about the standard C
+library functions. If you call “exit(0)” in main(), it knows that it is
+safe to optimize that into “return 0;” because C specifies what the
+‘exit’ function does.</p>
+<p>In addition to simple library knowledge, it is possible to embed a
+variety of other language-specific information into the LLVM IR. If you
+have a specific need and run into a wall, please bring the topic up on
+the llvm-dev list. At the very worst, you can always treat LLVM as if it
+were a “dumb code generator” and implement the high-level optimizations
+you desire in your front-end, on the language-specific AST.</p>
+</div>
+</div>
+<div class="section" id="tips-and-tricks">
+<h2><a class="toc-backref" href="#id7">8.3. Tips and Tricks</a><a class="headerlink" href="#tips-and-tricks" title="Permalink to this headline">¶</a></h2>
+<p>There is a variety of useful tips and tricks that you come to know after
+working on/with LLVM that aren’t obvious at first glance. Instead of
+letting everyone rediscover them, this section talks about some of these
+issues.</p>
+<div class="section" id="implementing-portable-offsetof-sizeof">
+<h3><a class="toc-backref" href="#id8">8.3.1. Implementing portable offsetof/sizeof</a><a class="headerlink" href="#implementing-portable-offsetof-sizeof" title="Permalink to this headline">¶</a></h3>
+<p>One interesting thing that comes up, if you are trying to keep the code
+generated by your compiler “target independent”, is that you often need
+to know the size of some LLVM type or the offset of some field in an
+llvm structure. For example, you might need to pass the size of a type
+into a function that allocates memory.</p>
+<p>Unfortunately, this can vary widely across targets: for example the
+width of a pointer is trivially target-specific. However, there is a
+<a class="reference external" href="http://nondot.org/sabre/LLVMNotes/SizeOf-OffsetOf-VariableSizedStructs.txt">clever way to use the getelementptr
+instruction</a>
+that allows you to compute this in a portable way.</p>
+</div>
+<div class="section" id="garbage-collected-stack-frames">
+<h3><a class="toc-backref" href="#id9">8.3.2. Garbage Collected Stack Frames</a><a class="headerlink" href="#garbage-collected-stack-frames" title="Permalink to this headline">¶</a></h3>
+<p>Some languages want to explicitly manage their stack frames, often so
+that they are garbage collected or to allow easy implementation of
+closures. There are often better ways to implement these features than
+explicit stack frames, but <a class="reference external" href="http://nondot.org/sabre/LLVMNotes/ExplicitlyManagedStackFrames.txt">LLVM does support
+them,</a>
+if you want. It requires your front-end to convert the code into
+<a class="reference external" href="http://en.wikipedia.org/wiki/Continuation-passing_style">Continuation Passing
+Style</a> and
+the use of tail calls (which LLVM also supports).</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="BuildingAJIT1.html" title="1. Building a JIT: Starting out with KaleidoscopeJIT"
+ >next</a> |</li>
+ <li class="right" >
+ <a href="OCamlLangImpl7.html" title="7. Kaleidoscope: Extending the Language: Mutable Variables"
+ >previous</a> |</li>
+ <li><a href="http://llvm.org/">LLVM Home</a> | </li>
+ <li><a href="../index.html">Documentation</a>»</li>
+
+ <li><a href="index.html" >LLVM Tutorial: Table of Contents</a> »</li>
+ </ul>
+ </div>
+ <div class="footer">
+ © Copyright 2003-2018, LLVM Project.
+ Last updated on 2018-12-21.
+ Created using <a href="http://sphinx.pocoo.org/">Sphinx</a> 1.1.3.
+ </div>
+ </body>
+</html>
\ No newline at end of file
Added: www-releases/trunk/7.0.1/docs/tutorial/index.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/7.0.1/docs/tutorial/index.html?rev=349965&view=auto
==============================================================================
--- www-releases/trunk/7.0.1/docs/tutorial/index.html (added)
+++ www-releases/trunk/7.0.1/docs/tutorial/index.html Fri Dec 21 13:53:02 2018
@@ -0,0 +1,207 @@
+
+
+<!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 Tutorial: Table of Contents — LLVM 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: '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 7 documentation" href="../index.html" />
+ <link rel="next" title="1. Kaleidoscope: Tutorial Introduction and the Lexer" href="LangImpl01.html" />
+ <link rel="prev" title="LLVM test-suite Guide" href="../TestSuiteMakefileGuide.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="LangImpl01.html" title="1. Kaleidoscope: Tutorial Introduction and the Lexer"
+ accesskey="N">next</a> |</li>
+ <li class="right" >
+ <a href="../TestSuiteMakefileGuide.html" title="LLVM test-suite Guide"
+ 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-tutorial-table-of-contents">
+<h1>LLVM Tutorial: Table of Contents<a class="headerlink" href="#llvm-tutorial-table-of-contents" title="Permalink to this headline">¶</a></h1>
+<div class="section" id="kaleidoscope-implementing-a-language-with-llvm">
+<h2>Kaleidoscope: Implementing a Language with LLVM<a class="headerlink" href="#kaleidoscope-implementing-a-language-with-llvm" title="Permalink to this headline">¶</a></h2>
+<div class="toctree-wrapper compound">
+<ul>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl01.html">1. Kaleidoscope: Tutorial Introduction and the Lexer</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl02.html">2. Kaleidoscope: Implementing a Parser and AST</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl03.html">3. Kaleidoscope: Code generation to LLVM IR</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl04.html">4. Kaleidoscope: Adding JIT and Optimizer Support</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl05.html">5. Kaleidoscope: Extending the Language: Control Flow</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl06.html">6. Kaleidoscope: Extending the Language: User-defined Operators</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl07.html">7. Kaleidoscope: Extending the Language: Mutable Variables</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl08.html">8. Kaleidoscope: Compiling to Object Code</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl09.html">9. Kaleidoscope: Adding Debug Information</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="LangImpl10.html">10. Kaleidoscope: Conclusion and other useful LLVM tidbits</a><ul class="simple">
+</ul>
+</li>
+</ul>
+</div>
+</div>
+<div class="section" id="kaleidoscope-implementing-a-language-with-llvm-in-objective-caml">
+<h2>Kaleidoscope: Implementing a Language with LLVM in Objective Caml<a class="headerlink" href="#kaleidoscope-implementing-a-language-with-llvm-in-objective-caml" title="Permalink to this headline">¶</a></h2>
+<div class="toctree-wrapper compound">
+<ul>
+<li class="toctree-l1"><a class="reference internal" href="OCamlLangImpl1.html">1. Kaleidoscope: Tutorial Introduction and the Lexer</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="OCamlLangImpl2.html">2. Kaleidoscope: Implementing a Parser and AST</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="OCamlLangImpl3.html">3. Kaleidoscope: Code generation to LLVM IR</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="OCamlLangImpl4.html">4. Kaleidoscope: Adding JIT and Optimizer Support</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="OCamlLangImpl5.html">5. Kaleidoscope: Extending the Language: Control Flow</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="OCamlLangImpl6.html">6. Kaleidoscope: Extending the Language: User-defined Operators</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="OCamlLangImpl7.html">7. Kaleidoscope: Extending the Language: Mutable Variables</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="OCamlLangImpl8.html">8. Kaleidoscope: Conclusion and other useful LLVM tidbits</a><ul class="simple">
+</ul>
+</li>
+</ul>
+</div>
+</div>
+<div class="section" id="building-a-jit-in-llvm">
+<h2>Building a JIT in LLVM<a class="headerlink" href="#building-a-jit-in-llvm" title="Permalink to this headline">¶</a></h2>
+<div class="toctree-wrapper compound">
+<ul>
+<li class="toctree-l1"><a class="reference internal" href="BuildingAJIT1.html">1. Building a JIT: Starting out with KaleidoscopeJIT</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="BuildingAJIT2.html">2. Building a JIT: Adding Optimizations – An introduction to ORC Layers</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="BuildingAJIT3.html">3. Building a JIT: Per-function Lazy Compilation</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="BuildingAJIT4.html">4. Building a JIT: Extreme Laziness - Using Compile Callbacks to JIT from ASTs</a><ul class="simple">
+</ul>
+</li>
+<li class="toctree-l1"><a class="reference internal" href="BuildingAJIT5.html">5. Building a JIT: Remote-JITing – Process Isolation and Laziness at a Distance</a><ul class="simple">
+</ul>
+</li>
+</ul>
+</div>
+</div>
+<div class="section" id="external-tutorials">
+<h2>External Tutorials<a class="headerlink" href="#external-tutorials" title="Permalink to this headline">¶</a></h2>
+<dl class="docutils">
+<dt><a class="reference external" href="http://jonathan2251.github.com/lbd/">Tutorial: Creating an LLVM Backend for the Cpu0 Architecture</a></dt>
+<dd>A step-by-step tutorial for developing an LLVM backend. Under
+active development at <a class="reference external" href="https://github.com/Jonathan2251/lbd">https://github.com/Jonathan2251/lbd</a> (please
+contribute!).</dd>
+<dt><a class="reference external" href="http://www.embecosm.com/appnotes/ean10/ean10-howto-llvmas-1.0.html">Howto: Implementing LLVM Integrated Assembler</a></dt>
+<dd>A simple guide for how to implement an LLVM integrated assembler for an
+architecture.</dd>
+</dl>
+</div>
+<div class="section" id="advanced-topics">
+<h2>Advanced Topics<a class="headerlink" href="#advanced-topics" title="Permalink to this headline">¶</a></h2>
+<ol class="arabic simple">
+<li><a class="reference external" href="http://llvm.org/pubs/2004-09-22-LCPCLLVMTutorial.html">Writing an Optimization for LLVM</a></li>
+</ol>
+</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="LangImpl01.html" title="1. Kaleidoscope: Tutorial Introduction and the Lexer"
+ >next</a> |</li>
+ <li class="right" >
+ <a href="../TestSuiteMakefileGuide.html" title="LLVM test-suite Guide"
+ >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 2003-2018, LLVM Project.
+ Last updated on 2018-12-21.
+ Created using <a href="http://sphinx.pocoo.org/">Sphinx</a> 1.1.3.
+ </div>
+ </body>
+</html>
\ No newline at end of file
Added: www-releases/trunk/7.0.1/docs/yaml2obj.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/7.0.1/docs/yaml2obj.html?rev=349965&view=auto
==============================================================================
--- www-releases/trunk/7.0.1/docs/yaml2obj.html (added)
+++ www-releases/trunk/7.0.1/docs/yaml2obj.html Fri Dec 21 13:53:02 2018
@@ -0,0 +1,308 @@
+
+
+<!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>yaml2obj — LLVM 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: '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 7 documentation" href="index.html" />
+ <link rel="next" title="How to submit an LLVM bug report" href="HowToSubmitABug.html" />
+ <link rel="prev" title="How To Add Your Build Configuration To LLVM Buildbot Infrastructure" href="HowToAddABuilder.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="HowToSubmitABug.html" title="How to submit an LLVM bug report"
+ accesskey="N">next</a> |</li>
+ <li class="right" >
+ <a href="HowToAddABuilder.html" title="How To Add Your Build Configuration To LLVM Buildbot Infrastructure"
+ 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="yaml2obj">
+<h1>yaml2obj<a class="headerlink" href="#yaml2obj" title="Permalink to this headline">¶</a></h1>
+<p>yaml2obj takes a YAML description of an object file and converts it to a binary
+file.</p>
+<blockquote>
+<div>$ yaml2obj input-file</div></blockquote>
+<p>Outputs the binary to stdout.</p>
+<div class="section" id="coff-syntax">
+<h2>COFF Syntax<a class="headerlink" href="#coff-syntax" title="Permalink to this headline">¶</a></h2>
+<p>Here’s a sample COFF file.</p>
+<div class="highlight-yaml"><div class="highlight"><pre><span class="l-Scalar-Plain">header</span><span class="p-Indicator">:</span>
+ <span class="l-Scalar-Plain">Machine</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">IMAGE_FILE_MACHINE_I386</span> <span class="c1"># (0x14C)</span>
+
+<span class="l-Scalar-Plain">sections</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">Name</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">.text</span>
+ <span class="l-Scalar-Plain">Characteristics</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="nv">IMAGE_SCN_CNT_CODE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_16BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_EXECUTE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_READ</span>
+ <span class="p-Indicator">]</span> <span class="c1"># 0x60500020</span>
+ <span class="l-Scalar-Plain">SectionData</span><span class="p-Indicator">:</span>
+ <span class="s">"</span><span class="se">\x83\xEC\x0C\xC7\x44\x24\x08\x00\x00\x00\x00\xC7\x04\x24\x00\x00\x00\x00\xE8\x00\x00\x00\x00\xE8\x00\x00\x00\x00\x8B\x44\x24\x08\x83\xC4\x0C\xC3</span><span class="s">"</span> <span class="c1"># |....D$.......$...............D$.....|</span>
+
+<span class="l-Scalar-Plain">symbols</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">Name</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">.text</span>
+ <span class="l-Scalar-Plain">Value</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">0</span>
+ <span class="l-Scalar-Plain">SectionNumber</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">1</span>
+ <span class="l-Scalar-Plain">SimpleType</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">IMAGE_SYM_TYPE_NULL</span> <span class="c1"># (0)</span>
+ <span class="l-Scalar-Plain">ComplexType</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">IMAGE_SYM_DTYPE_NULL</span> <span class="c1"># (0)</span>
+ <span class="l-Scalar-Plain">StorageClass</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">IMAGE_SYM_CLASS_STATIC</span> <span class="c1"># (3)</span>
+ <span class="l-Scalar-Plain">NumberOfAuxSymbols</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">1</span>
+ <span class="l-Scalar-Plain">AuxiliaryData</span><span class="p-Indicator">:</span>
+ <span class="s">"</span><span class="se">\x24\x00\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x01\x00\x00\x00\x00\x00</span><span class="s">"</span> <span class="c1"># |$.................|</span>
+
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">Name</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">_main</span>
+ <span class="l-Scalar-Plain">Value</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">0</span>
+ <span class="l-Scalar-Plain">SectionNumber</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">1</span>
+ <span class="l-Scalar-Plain">SimpleType</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">IMAGE_SYM_TYPE_NULL</span> <span class="c1"># (0)</span>
+ <span class="l-Scalar-Plain">ComplexType</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">IMAGE_SYM_DTYPE_NULL</span> <span class="c1"># (0)</span>
+ <span class="l-Scalar-Plain">StorageClass</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">IMAGE_SYM_CLASS_EXTERNAL</span> <span class="c1"># (2)</span>
+</pre></div>
+</div>
+<p>Here’s a simplified <a class="reference external" href="http://www.kuwata-lab.com/kwalify/ruby/users-guide.html">Kwalify</a> schema with an extension to allow alternate types.</p>
+<div class="highlight-yaml"><div class="highlight"><pre><span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">map</span>
+ <span class="l-Scalar-Plain">mapping</span><span class="p-Indicator">:</span>
+ <span class="l-Scalar-Plain">header</span><span class="p-Indicator">:</span>
+ <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">map</span>
+ <span class="l-Scalar-Plain">mapping</span><span class="p-Indicator">:</span>
+ <span class="l-Scalar-Plain">Machine</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">str</span><span class="p-Indicator">,</span> <span class="nv">enum</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">[</span> <span class="nv">IMAGE_FILE_MACHINE_UNKNOWN</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_AM33</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_AMD64</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_ARM</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_ARMNT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_ARM64</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_EBC</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_I386</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_IA64</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_M32R</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_MIPS16</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_MIPSFPU</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_MIPSFPU16</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_POWERPC</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_POWERPCFP</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_R4000</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_SH3</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_SH3DSP</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_SH4</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_SH5</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_THUMB</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_MACHINE_WCEMIPSV2</span>
+ <span class="p-Indicator">]}</span>
+ <span class="p-Indicator">,</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">int</span><span class="p-Indicator">}</span>
+ <span class="p-Indicator">]</span>
+ <span class="l-Scalar-Plain">Characteristics</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">seq</span>
+ <span class="l-Scalar-Plain">sequence</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">str</span>
+ <span class="l-Scalar-Plain">enum</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="nv">IMAGE_FILE_RELOCS_STRIPPED</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_EXECUTABLE_IMAGE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_LINE_NUMS_STRIPPED</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_LOCAL_SYMS_STRIPPED</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_AGGRESSIVE_WS_TRIM</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_LARGE_ADDRESS_AWARE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_BYTES_REVERSED_LO</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_32BIT_MACHINE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_DEBUG_STRIPPED</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_REMOVABLE_RUN_FROM_SWAP</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_NET_RUN_FROM_SWAP</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_SYSTEM</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_DLL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_UP_SYSTEM_ONLY</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_FILE_BYTES_REVERSED_HI</span>
+ <span class="p-Indicator">]</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">int</span>
+ <span class="l-Scalar-Plain">sections</span><span class="p-Indicator">:</span>
+ <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">seq</span>
+ <span class="l-Scalar-Plain">sequence</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">map</span>
+ <span class="l-Scalar-Plain">mapping</span><span class="p-Indicator">:</span>
+ <span class="l-Scalar-Plain">Name</span><span class="p-Indicator">:</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">str</span><span class="p-Indicator">}</span>
+ <span class="l-Scalar-Plain">Characteristics</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">seq</span>
+ <span class="l-Scalar-Plain">sequence</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">str</span>
+ <span class="l-Scalar-Plain">enum</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="nv">IMAGE_SCN_TYPE_NO_PAD</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_CNT_CODE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_CNT_INITIALIZED_DATA</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_CNT_UNINITIALIZED_DATA</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_LNK_OTHER</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_LNK_INFO</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_LNK_REMOVE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_LNK_COMDAT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_GPREL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_PURGEABLE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_16BIT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_LOCKED</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_PRELOAD</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_1BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_2BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_4BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_8BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_16BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_32BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_64BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_128BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_256BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_512BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_1024BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_2048BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_4096BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_ALIGN_8192BYTES</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_LNK_NRELOC_OVFL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_DISCARDABLE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_NOT_CACHED</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_NOT_PAGED</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_SHARED</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_EXECUTE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_READ</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SCN_MEM_WRITE</span>
+ <span class="p-Indicator">]</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">int</span>
+ <span class="l-Scalar-Plain">SectionData</span><span class="p-Indicator">:</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">str</span><span class="p-Indicator">}</span>
+ <span class="l-Scalar-Plain">symbols</span><span class="p-Indicator">:</span>
+ <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">seq</span>
+ <span class="l-Scalar-Plain">sequence</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">-</span> <span class="l-Scalar-Plain">type</span><span class="p-Indicator">:</span> <span class="l-Scalar-Plain">map</span>
+ <span class="l-Scalar-Plain">mapping</span><span class="p-Indicator">:</span>
+ <span class="l-Scalar-Plain">Name</span><span class="p-Indicator">:</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">str</span><span class="p-Indicator">}</span>
+ <span class="l-Scalar-Plain">Value</span><span class="p-Indicator">:</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">int</span><span class="p-Indicator">}</span>
+ <span class="l-Scalar-Plain">SectionNumber</span><span class="p-Indicator">:</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">int</span><span class="p-Indicator">}</span>
+ <span class="l-Scalar-Plain">SimpleType</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">str</span><span class="p-Indicator">,</span> <span class="nv">enum</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="nv">IMAGE_SYM_TYPE_NULL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_VOID</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_CHAR</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_SHORT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_INT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_LONG</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_FLOAT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_DOUBLE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_STRUCT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_UNION</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_ENUM</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_MOE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_BYTE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_WORD</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_UINT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_TYPE_DWORD</span>
+ <span class="p-Indicator">]}</span>
+ <span class="p-Indicator">,</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">int</span><span class="p-Indicator">}</span>
+ <span class="p-Indicator">]</span>
+ <span class="l-Scalar-Plain">ComplexType</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">str</span><span class="p-Indicator">,</span> <span class="nv">enum</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="nv">IMAGE_SYM_DTYPE_NULL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_DTYPE_POINTER</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_DTYPE_FUNCTION</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_DTYPE_ARRAY</span>
+ <span class="p-Indicator">]}</span>
+ <span class="p-Indicator">,</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">int</span><span class="p-Indicator">}</span>
+ <span class="p-Indicator">]</span>
+ <span class="l-Scalar-Plain">StorageClass</span><span class="p-Indicator">:</span> <span class="p-Indicator">[</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">str</span><span class="p-Indicator">,</span> <span class="nv">enum</span><span class="p-Indicator">:</span>
+ <span class="p-Indicator">[</span> <span class="nv">IMAGE_SYM_CLASS_END_OF_FUNCTION</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_NULL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_AUTOMATIC</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_EXTERNAL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_STATIC</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_REGISTER</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_EXTERNAL_DEF</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_LABEL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_UNDEFINED_LABEL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_MEMBER_OF_STRUCT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_ARGUMENT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_STRUCT_TAG</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_MEMBER_OF_UNION</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_UNION_TAG</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_TYPE_DEFINITION</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_UNDEFINED_STATIC</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_ENUM_TAG</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_MEMBER_OF_ENUM</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_REGISTER_PARAM</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_BIT_FIELD</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_BLOCK</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_FUNCTION</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_END_OF_STRUCT</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_FILE</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_SECTION</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_WEAK_EXTERNAL</span>
+ <span class="p-Indicator">,</span> <span class="nv">IMAGE_SYM_CLASS_CLR_TOKEN</span>
+ <span class="p-Indicator">]}</span>
+ <span class="p-Indicator">,</span> <span class="p-Indicator">{</span><span class="nv">type</span><span class="p-Indicator">:</span> <span class="nv">int</span><span class="p-Indicator">}</span>
+ <span class="p-Indicator">]</span>
+</pre></div>
+</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="HowToSubmitABug.html" title="How to submit an LLVM bug report"
+ >next</a> |</li>
+ <li class="right" >
+ <a href="HowToAddABuilder.html" title="How To Add Your Build Configuration To LLVM Buildbot Infrastructure"
+ >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 2003-2018, LLVM Project.
+ Last updated on 2018-12-21.
+ Created using <a href="http://sphinx.pocoo.org/">Sphinx</a> 1.1.3.
+ </div>
+ </body>
+</html>
\ No newline at end of file
Added: www-releases/trunk/7.0.1/tools/clang/docs/AddressSanitizer.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/7.0.1/tools/clang/docs/AddressSanitizer.html?rev=349965&view=auto
==============================================================================
--- www-releases/trunk/7.0.1/tools/clang/docs/AddressSanitizer.html (added)
+++ www-releases/trunk/7.0.1/tools/clang/docs/AddressSanitizer.html Fri Dec 21 13:53:02 2018
@@ -0,0 +1,350 @@
+
+<!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="X-UA-Compatible" content="IE=Edge" />
+ <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+ <title>AddressSanitizer — Clang 7 documentation</title>
+ <link rel="stylesheet" href="_static/haiku.css" type="text/css" />
+ <link rel="stylesheet" href="_static/pygments.css" type="text/css" />
+ <script type="text/javascript" id="documentation_options" data-url_root="./" src="_static/documentation_options.js"></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>
+ <script type="text/javascript" src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.1/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
+ <link rel="index" title="Index" href="genindex.html" />
+ <link rel="search" title="Search" href="search.html" />
+ <link rel="next" title="ThreadSanitizer" href="ThreadSanitizer.html" />
+ <link rel="prev" title="Thread Safety Analysis" href="ThreadSafetyAnalysis.html" />
+ </head><body>
+ <div class="header" role="banner"><h1 class="heading"><a href="index.html">
+ <span>Clang 7 documentation</span></a></h1>
+ <h2 class="heading"><span>AddressSanitizer</span></h2>
+ </div>
+ <div class="topnav" role="navigation" aria-label="top navigation">
+
+ <p>
+ « <a href="ThreadSafetyAnalysis.html">Thread Safety Analysis</a>
+ ::
+ <a class="uplink" href="index.html">Contents</a>
+ ::
+ <a href="ThreadSanitizer.html">ThreadSanitizer</a> »
+ </p>
+
+ </div>
+ <div class="content">
+
+
+ <div class="section" id="addresssanitizer">
+<h1>AddressSanitizer<a class="headerlink" href="#addresssanitizer" 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="id1">Introduction</a></li>
+<li><a class="reference internal" href="#how-to-build" id="id2">How to build</a></li>
+<li><a class="reference internal" href="#usage" id="id3">Usage</a></li>
+<li><a class="reference internal" href="#symbolizing-the-reports" id="id4">Symbolizing the Reports</a></li>
+<li><a class="reference internal" href="#additional-checks" id="id5">Additional Checks</a><ul>
+<li><a class="reference internal" href="#initialization-order-checking" id="id6">Initialization order checking</a></li>
+<li><a class="reference internal" href="#memory-leak-detection" id="id7">Memory leak detection</a></li>
+<li><a class="reference internal" href="#writable-executable-paging-detection" id="id8">Writable/Executable paging detection</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#issue-suppression" id="id9">Issue Suppression</a><ul>
+<li><a class="reference internal" href="#suppressing-reports-in-external-libraries" id="id10">Suppressing Reports in External Libraries</a></li>
+<li><a class="reference internal" href="#conditional-compilation-with-has-feature-address-sanitizer" id="id11">Conditional Compilation with <code class="docutils literal notranslate"><span class="pre">__has_feature(address_sanitizer)</span></code></a></li>
+<li><a class="reference internal" href="#disabling-instrumentation-with-attribute-no-sanitize-address" id="id12">Disabling Instrumentation with <code class="docutils literal notranslate"><span class="pre">__attribute__((no_sanitize("address")))</span></code></a></li>
+<li><a class="reference internal" href="#suppressing-errors-in-recompiled-code-blacklist" id="id13">Suppressing Errors in Recompiled Code (Blacklist)</a></li>
+<li><a class="reference internal" href="#suppressing-memory-leaks" id="id14">Suppressing memory leaks</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#limitations" id="id15">Limitations</a></li>
+<li><a class="reference internal" href="#supported-platforms" id="id16">Supported Platforms</a></li>
+<li><a class="reference internal" href="#current-status" id="id17">Current Status</a></li>
+<li><a class="reference internal" href="#more-information" id="id18">More Information</a></li>
+</ul>
+</div>
+<div class="section" id="introduction">
+<h2><a class="toc-backref" href="#id1">Introduction</a><a class="headerlink" href="#introduction" title="Permalink to this headline">¶</a></h2>
+<p>AddressSanitizer is a fast memory error detector. It consists of a compiler
+instrumentation module and a run-time library. The tool can detect the
+following types of bugs:</p>
+<ul class="simple">
+<li>Out-of-bounds accesses to heap, stack and globals</li>
+<li>Use-after-free</li>
+<li>Use-after-return (runtime flag <cite>ASAN_OPTIONS=detect_stack_use_after_return=1</cite>)</li>
+<li>Use-after-scope (clang flag <cite>-fsanitize-address-use-after-scope</cite>)</li>
+<li>Double-free, invalid free</li>
+<li>Memory leaks (experimental)</li>
+</ul>
+<p>Typical slowdown introduced by AddressSanitizer is <strong>2x</strong>.</p>
+</div>
+<div class="section" id="how-to-build">
+<h2><a class="toc-backref" href="#id2">How to build</a><a class="headerlink" href="#how-to-build" title="Permalink to this headline">¶</a></h2>
+<p>Build LLVM/Clang with <a class="reference external" href="http://llvm.org/docs/CMake.html">CMake</a>.</p>
+</div>
+<div class="section" id="usage">
+<h2><a class="toc-backref" href="#id3">Usage</a><a class="headerlink" href="#usage" title="Permalink to this headline">¶</a></h2>
+<p>Simply compile and link your program with <code class="docutils literal notranslate"><span class="pre">-fsanitize=address</span></code> flag. The
+AddressSanitizer run-time library should be linked to the final executable, so
+make sure to use <code class="docutils literal notranslate"><span class="pre">clang</span></code> (not <code class="docutils literal notranslate"><span class="pre">ld</span></code>) for the final link step. When linking
+shared libraries, the AddressSanitizer run-time is not linked, so
+<code class="docutils literal notranslate"><span class="pre">-Wl,-z,defs</span></code> may cause link errors (donât use it with AddressSanitizer). To
+get a reasonable performance add <code class="docutils literal notranslate"><span class="pre">-O1</span></code> or higher. To get nicer stack traces
+in error messages add <code class="docutils literal notranslate"><span class="pre">-fno-omit-frame-pointer</span></code>. To get perfect stack traces
+you may need to disable inlining (just use <code class="docutils literal notranslate"><span class="pre">-O1</span></code>) and tail call elimination
+(<code class="docutils literal notranslate"><span class="pre">-fno-optimize-sibling-calls</span></code>).</p>
+<div class="highlight-console notranslate"><div class="highlight"><pre><span></span><span class="gp">%</span> cat example_UseAfterFree.cc
+<span class="go">int main(int argc, char **argv) {</span>
+<span class="go"> int *array = new int[100];</span>
+<span class="go"> delete [] array;</span>
+<span class="go"> return array[argc]; // BOOM</span>
+<span class="go">}</span>
+
+<span class="gp">#</span> Compile and link
+<span class="gp">%</span> clang++ -O1 -g -fsanitize<span class="o">=</span>address -fno-omit-frame-pointer example_UseAfterFree.cc
+</pre></div>
+</div>
+<p>or:</p>
+<div class="highlight-console notranslate"><div class="highlight"><pre><span></span><span class="gp">#</span> Compile
+<span class="gp">%</span> clang++ -O1 -g -fsanitize<span class="o">=</span>address -fno-omit-frame-pointer -c example_UseAfterFree.cc
+<span class="gp">#</span> Link
+<span class="gp">%</span> clang++ -g -fsanitize<span class="o">=</span>address example_UseAfterFree.o
+</pre></div>
+</div>
+<p>If a bug is detected, the program will print an error message to stderr and
+exit with a non-zero exit code. AddressSanitizer exits on the first detected error.
+This is by design:</p>
+<ul class="simple">
+<li>This approach allows AddressSanitizer to produce faster and smaller generated code
+(both by ~5%).</li>
+<li>Fixing bugs becomes unavoidable. AddressSanitizer does not produce
+false alarms. Once a memory corruption occurs, the program is in an inconsistent
+state, which could lead to confusing results and potentially misleading
+subsequent reports.</li>
+</ul>
+<p>If your process is sandboxed and you are running on OS X 10.10 or earlier, you
+will need to set <code class="docutils literal notranslate"><span class="pre">DYLD_INSERT_LIBRARIES</span></code> environment variable and point it to
+the ASan library that is packaged with the compiler used to build the
+executable. (You can find the library by searching for dynamic libraries with
+<code class="docutils literal notranslate"><span class="pre">asan</span></code> in their name.) If the environment variable is not set, the process will
+try to re-exec. Also keep in mind that when moving the executable to another machine,
+the ASan library will also need to be copied over.</p>
+</div>
+<div class="section" id="symbolizing-the-reports">
+<h2><a class="toc-backref" href="#id4">Symbolizing the Reports</a><a class="headerlink" href="#symbolizing-the-reports" title="Permalink to this headline">¶</a></h2>
+<p>To make AddressSanitizer symbolize its output
+you need to set the <code class="docutils literal notranslate"><span class="pre">ASAN_SYMBOLIZER_PATH</span></code> environment variable to point to
+the <code class="docutils literal notranslate"><span class="pre">llvm-symbolizer</span></code> binary (or make sure <code class="docutils literal notranslate"><span class="pre">llvm-symbolizer</span></code> is in your
+<code class="docutils literal notranslate"><span class="pre">$PATH</span></code>):</p>
+<div class="highlight-console notranslate"><div class="highlight"><pre><span></span><span class="gp">%</span> <span class="nv">ASAN_SYMBOLIZER_PATH</span><span class="o">=</span>/usr/local/bin/llvm-symbolizer ./a.out
+<span class="go">==9442== ERROR: AddressSanitizer heap-use-after-free on address 0x7f7ddab8c084 at pc 0x403c8c bp 0x7fff87fb82d0 sp 0x7fff87fb82c8</span>
+<span class="go">READ of size 4 at 0x7f7ddab8c084 thread T0</span>
+<span class="gp"> #</span><span class="m">0</span> 0x403c8c in main example_UseAfterFree.cc:4
+<span class="gp"> #</span><span class="m">1</span> 0x7f7ddabcac4d in __libc_start_main ??:0
+<span class="go">0x7f7ddab8c084 is located 4 bytes inside of 400-byte region [0x7f7ddab8c080,0x7f7ddab8c210)</span>
+<span class="go">freed by thread T0 here:</span>
+<span class="gp"> #</span><span class="m">0</span> 0x404704 in operator delete<span class="o">[](</span>void*<span class="o">)</span> ??:0
+<span class="gp"> #</span><span class="m">1</span> 0x403c53 in main example_UseAfterFree.cc:4
+<span class="gp"> #</span><span class="m">2</span> 0x7f7ddabcac4d in __libc_start_main ??:0
+<span class="go">previously allocated by thread T0 here:</span>
+<span class="gp"> #</span><span class="m">0</span> 0x404544 in operator new<span class="o">[](</span>unsigned long<span class="o">)</span> ??:0
+<span class="gp"> #</span><span class="m">1</span> 0x403c43 in main example_UseAfterFree.cc:2
+<span class="gp"> #</span><span class="m">2</span> 0x7f7ddabcac4d in __libc_start_main ??:0
+<span class="go">==9442== ABORTING</span>
+</pre></div>
+</div>
+<p>If that does not work for you (e.g. your process is sandboxed), you can use a
+separate script to symbolize the result offline (online symbolization can be
+force disabled by setting <code class="docutils literal notranslate"><span class="pre">ASAN_OPTIONS=symbolize=0</span></code>):</p>
+<div class="highlight-console notranslate"><div class="highlight"><pre><span></span><span class="gp">%</span> <span class="nv">ASAN_OPTIONS</span><span class="o">=</span><span class="nv">symbolize</span><span class="o">=</span><span class="m">0</span> ./a.out <span class="m">2</span>> log
+<span class="gp">%</span> projects/compiler-rt/lib/asan/scripts/asan_symbolize.py / < log <span class="p">|</span> c++filt
+<span class="go">==9442== ERROR: AddressSanitizer heap-use-after-free on address 0x7f7ddab8c084 at pc 0x403c8c bp 0x7fff87fb82d0 sp 0x7fff87fb82c8</span>
+<span class="go">READ of size 4 at 0x7f7ddab8c084 thread T0</span>
+<span class="gp"> #</span><span class="m">0</span> 0x403c8c in main example_UseAfterFree.cc:4
+<span class="gp"> #</span><span class="m">1</span> 0x7f7ddabcac4d in __libc_start_main ??:0
+<span class="go">...</span>
+</pre></div>
+</div>
+<p>Note that on OS X you may need to run <code class="docutils literal notranslate"><span class="pre">dsymutil</span></code> on your binary to have the
+file:line info in the AddressSanitizer reports.</p>
+</div>
+<div class="section" id="additional-checks">
+<h2><a class="toc-backref" href="#id5">Additional Checks</a><a class="headerlink" href="#additional-checks" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="initialization-order-checking">
+<h3><a class="toc-backref" href="#id6">Initialization order checking</a><a class="headerlink" href="#initialization-order-checking" title="Permalink to this headline">¶</a></h3>
+<p>AddressSanitizer can optionally detect dynamic initialization order problems,
+when initialization of globals defined in one translation unit uses
+globals defined in another translation unit. To enable this check at runtime,
+you should set environment variable
+<code class="docutils literal notranslate"><span class="pre">ASAN_OPTIONS=check_initialization_order=1</span></code>.</p>
+<p>Note that this option is not supported on OS X.</p>
+</div>
+<div class="section" id="memory-leak-detection">
+<h3><a class="toc-backref" href="#id7">Memory leak detection</a><a class="headerlink" href="#memory-leak-detection" title="Permalink to this headline">¶</a></h3>
+<p>For more information on leak detector in AddressSanitizer, see
+<a class="reference internal" href="LeakSanitizer.html"><span class="doc">LeakSanitizer</span></a>. The leak detection is turned on by default on Linux,
+and can be enabled using <code class="docutils literal notranslate"><span class="pre">ASAN_OPTIONS=detect_leaks=1</span></code> on OS X;
+however, it is not yet supported on other platforms.</p>
+</div>
+<div class="section" id="writable-executable-paging-detection">
+<h3><a class="toc-backref" href="#id8">Writable/Executable paging detection</a><a class="headerlink" href="#writable-executable-paging-detection" title="Permalink to this headline">¶</a></h3>
+<p>The W^X detection is disabled by default and can be enabled using
+<code class="docutils literal notranslate"><span class="pre">ASAN_OPTIONS=detect_write_exec=1</span></code>.</p>
+</div>
+</div>
+<div class="section" id="issue-suppression">
+<h2><a class="toc-backref" href="#id9">Issue Suppression</a><a class="headerlink" href="#issue-suppression" title="Permalink to this headline">¶</a></h2>
+<p>AddressSanitizer is not expected to produce false positives. If you see one,
+look again; most likely it is a true positive!</p>
+<div class="section" id="suppressing-reports-in-external-libraries">
+<h3><a class="toc-backref" href="#id10">Suppressing Reports in External Libraries</a><a class="headerlink" href="#suppressing-reports-in-external-libraries" title="Permalink to this headline">¶</a></h3>
+<p>Runtime interposition allows AddressSanitizer to find bugs in code that is
+not being recompiled. If you run into an issue in external libraries, we
+recommend immediately reporting it to the library maintainer so that it
+gets addressed. However, you can use the following suppression mechanism
+to unblock yourself and continue on with the testing. This suppression
+mechanism should only be used for suppressing issues in external code; it
+does not work on code recompiled with AddressSanitizer. To suppress errors
+in external libraries, set the <code class="docutils literal notranslate"><span class="pre">ASAN_OPTIONS</span></code> environment variable to point
+to a suppression file. You can either specify the full path to the file or the
+path of the file relative to the location of your executable.</p>
+<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span><span class="nv">ASAN_OPTIONS</span><span class="o">=</span><span class="nv">suppressions</span><span class="o">=</span>MyASan.supp
+</pre></div>
+</div>
+<p>Use the following format to specify the names of the functions or libraries
+you want to suppress. You can see these in the error report. Remember that
+the narrower the scope of the suppression, the more bugs you will be able to
+catch.</p>
+<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span>interceptor_via_fun:NameOfCFunctionToSuppress
+interceptor_via_fun:-<span class="o">[</span>ClassName objCMethodToSuppress:<span class="o">]</span>
+interceptor_via_lib:NameOfTheLibraryToSuppress
+</pre></div>
+</div>
+</div>
+<div class="section" id="conditional-compilation-with-has-feature-address-sanitizer">
+<h3><a class="toc-backref" href="#id11">Conditional Compilation with <code class="docutils literal notranslate"><span class="pre">__has_feature(address_sanitizer)</span></code></a><a class="headerlink" href="#conditional-compilation-with-has-feature-address-sanitizer" title="Permalink to this headline">¶</a></h3>
+<p>In some cases one may need to execute different code depending on whether
+AddressSanitizer is enabled.
+<a class="reference internal" href="LanguageExtensions.html#langext-has-feature-has-extension"><span class="std std-ref">__has_feature</span></a> can be used for
+this purpose.</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="cp">#if defined(__has_feature)</span>
+<span class="cp"># if __has_feature(address_sanitizer)</span>
+<span class="c1">// code that builds only under AddressSanitizer</span>
+<span class="cp"># endif</span>
+<span class="cp">#endif</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="disabling-instrumentation-with-attribute-no-sanitize-address">
+<h3><a class="toc-backref" href="#id12">Disabling Instrumentation with <code class="docutils literal notranslate"><span class="pre">__attribute__((no_sanitize("address")))</span></code></a><a class="headerlink" href="#disabling-instrumentation-with-attribute-no-sanitize-address" title="Permalink to this headline">¶</a></h3>
+<p>Some code should not be instrumented by AddressSanitizer. One may use
+the attribute <code class="docutils literal notranslate"><span class="pre">__attribute__((no_sanitize("address")))</span></code> (which has
+deprecated synonyms <cite>no_sanitize_address</cite> and
+<cite>no_address_safety_analysis</cite>) to disable instrumentation of a
+particular function. This attribute may not be supported by other
+compilers, so we suggest to use it together with
+<code class="docutils literal notranslate"><span class="pre">__has_feature(address_sanitizer)</span></code>.</p>
+<p>The same attribute used on a global variable prevents AddressSanitizer
+from adding redzones around it and detecting out of bounds accesses.</p>
+</div>
+<div class="section" id="suppressing-errors-in-recompiled-code-blacklist">
+<h3><a class="toc-backref" href="#id13">Suppressing Errors in Recompiled Code (Blacklist)</a><a class="headerlink" href="#suppressing-errors-in-recompiled-code-blacklist" title="Permalink to this headline">¶</a></h3>
+<p>AddressSanitizer supports <code class="docutils literal notranslate"><span class="pre">src</span></code> and <code class="docutils literal notranslate"><span class="pre">fun</span></code> entity types in
+<a class="reference internal" href="SanitizerSpecialCaseList.html"><span class="doc">Sanitizer special case list</span></a>, that can be used to suppress error reports
+in the specified source files or functions. Additionally, AddressSanitizer
+introduces <code class="docutils literal notranslate"><span class="pre">global</span></code> and <code class="docutils literal notranslate"><span class="pre">type</span></code> entity types that can be used to
+suppress error reports for out-of-bound access to globals with certain
+names and types (you may only specify class or struct types).</p>
+<p>You may use an <code class="docutils literal notranslate"><span class="pre">init</span></code> category to suppress reports about initialization-order
+problems happening in certain source files or with certain global variables.</p>
+<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span><span class="c1"># Suppress error reports for code in a file or in a function:</span>
+src:bad_file.cpp
+<span class="c1"># Ignore all functions with names containing MyFooBar:</span>
+fun:*MyFooBar*
+<span class="c1"># Disable out-of-bound checks for global:</span>
+global:bad_array
+<span class="c1"># Disable out-of-bound checks for global instances of a given class ...</span>
+type:Namespace::BadClassName
+<span class="c1"># ... or a given struct. Use wildcard to deal with anonymous namespace.</span>
+type:Namespace2::*::BadStructName
+<span class="c1"># Disable initialization-order checks for globals:</span>
+global:bad_init_global<span class="o">=</span>init
+type:*BadInitClassSubstring*<span class="o">=</span>init
+src:bad/init/files/*<span class="o">=</span>init
+</pre></div>
+</div>
+</div>
+<div class="section" id="suppressing-memory-leaks">
+<h3><a class="toc-backref" href="#id14">Suppressing memory leaks</a><a class="headerlink" href="#suppressing-memory-leaks" title="Permalink to this headline">¶</a></h3>
+<p>Memory leak reports produced by <a class="reference internal" href="LeakSanitizer.html"><span class="doc">LeakSanitizer</span></a> (if it is run as a part
+of AddressSanitizer) can be suppressed by a separate file passed as</p>
+<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span><span class="nv">LSAN_OPTIONS</span><span class="o">=</span><span class="nv">suppressions</span><span class="o">=</span>MyLSan.supp
+</pre></div>
+</div>
+<p>which contains lines of the form <cite>leak:<pattern></cite>. Memory leak will be
+suppressed if pattern matches any function name, source file name, or
+library name in the symbolized stack trace of the leak report. See
+<a class="reference external" href="https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer#suppressions">full documentation</a>
+for more details.</p>
+</div>
+</div>
+<div class="section" id="limitations">
+<h2><a class="toc-backref" href="#id15">Limitations</a><a class="headerlink" href="#limitations" title="Permalink to this headline">¶</a></h2>
+<ul class="simple">
+<li>AddressSanitizer uses more real memory than a native run. Exact overhead
+depends on the allocations sizes. The smaller the allocations you make the
+bigger the overhead is.</li>
+<li>AddressSanitizer uses more stack memory. We have seen up to 3x increase.</li>
+<li>On 64-bit platforms AddressSanitizer maps (but not reserves) 16+ Terabytes of
+virtual address space. This means that tools like <code class="docutils literal notranslate"><span class="pre">ulimit</span></code> may not work as
+usually expected.</li>
+<li>Static linking is not supported.</li>
+</ul>
+</div>
+<div class="section" id="supported-platforms">
+<h2><a class="toc-backref" href="#id16">Supported Platforms</a><a class="headerlink" href="#supported-platforms" title="Permalink to this headline">¶</a></h2>
+<p>AddressSanitizer is supported on:</p>
+<ul class="simple">
+<li>Linux i386/x86_64 (tested on Ubuntu 12.04)</li>
+<li>OS X 10.7 - 10.11 (i386/x86_64)</li>
+<li>iOS Simulator</li>
+<li>Android ARM</li>
+<li>NetBSD i386/x86_64</li>
+<li>FreeBSD i386/x86_64 (tested on FreeBSD 11-current)</li>
+</ul>
+<p>Ports to various other platforms are in progress.</p>
+</div>
+<div class="section" id="current-status">
+<h2><a class="toc-backref" href="#id17">Current Status</a><a class="headerlink" href="#current-status" title="Permalink to this headline">¶</a></h2>
+<p>AddressSanitizer is fully functional on supported platforms starting from LLVM
+3.1. The test suite is integrated into CMake build and can be run with <code class="docutils literal notranslate"><span class="pre">make</span>
+<span class="pre">check-asan</span></code> command.</p>
+</div>
+<div class="section" id="more-information">
+<h2><a class="toc-backref" href="#id18">More Information</a><a class="headerlink" href="#more-information" title="Permalink to this headline">¶</a></h2>
+<p><a class="reference external" href="https://github.com/google/sanitizers/wiki/AddressSanitizer">https://github.com/google/sanitizers/wiki/AddressSanitizer</a></p>
+</div>
+</div>
+
+
+ </div>
+ <div class="bottomnav" role="navigation" aria-label="bottom navigation">
+
+ <p>
+ « <a href="ThreadSafetyAnalysis.html">Thread Safety Analysis</a>
+ ::
+ <a class="uplink" href="index.html">Contents</a>
+ ::
+ <a href="ThreadSanitizer.html">ThreadSanitizer</a> »
+ </p>
+
+ </div>
+
+ <div class="footer" role="contentinfo">
+ © Copyright 2007-2018, The Clang Team.
+ Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.7.6.
+ </div>
+ </body>
+</html>
\ No newline at end of file
Added: www-releases/trunk/7.0.1/tools/clang/docs/AttributeReference.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/7.0.1/tools/clang/docs/AttributeReference.html?rev=349965&view=auto
==============================================================================
--- www-releases/trunk/7.0.1/tools/clang/docs/AttributeReference.html (added)
+++ www-releases/trunk/7.0.1/tools/clang/docs/AttributeReference.html Fri Dec 21 13:53:02 2018
@@ -0,0 +1,7275 @@
+
+<!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="X-UA-Compatible" content="IE=Edge" />
+ <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+ <title>Attributes in Clang — Clang 7 documentation</title>
+ <link rel="stylesheet" href="_static/haiku.css" type="text/css" />
+ <link rel="stylesheet" href="_static/pygments.css" type="text/css" />
+ <script type="text/javascript" id="documentation_options" data-url_root="./" src="_static/documentation_options.js"></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>
+ <script type="text/javascript" src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.1/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
+ <link rel="index" title="Index" href="genindex.html" />
+ <link rel="search" title="Search" href="search.html" />
+ <link rel="next" title="Diagnostic flags in Clang" href="DiagnosticsReference.html" />
+ <link rel="prev" title="Clang command line argument reference" href="ClangCommandLineReference.html" />
+ </head><body>
+ <div class="header" role="banner"><h1 class="heading"><a href="index.html">
+ <span>Clang 7 documentation</span></a></h1>
+ <h2 class="heading"><span>Attributes in Clang</span></h2>
+ </div>
+ <div class="topnav" role="navigation" aria-label="top navigation">
+
+ <p>
+ « <a href="ClangCommandLineReference.html">Clang command line argument reference</a>
+ ::
+ <a class="uplink" href="index.html">Contents</a>
+ ::
+ <a href="DiagnosticsReference.html">Diagnostic flags in Clang</a> »
+ </p>
+
+ </div>
+ <div class="content">
+
+
+ <div class="section" id="attributes-in-clang">
+<h1>Attributes in Clang<a class="headerlink" href="#attributes-in-clang" 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="id137">Introduction</a></li>
+<li><a class="reference internal" href="#function-attributes" id="id138">Function Attributes</a><ul>
+<li><a class="reference internal" href="#pragma-omp-declare-simd" id="id139">#pragma omp declare simd</a></li>
+<li><a class="reference internal" href="#pragma-omp-declare-target" id="id140">#pragma omp declare target</a></li>
+<li><a class="reference internal" href="#noreturn" id="id141">_Noreturn</a></li>
+<li><a class="reference internal" href="#abi-tag-gnu-abi-tag" id="id142">abi_tag (gnu::abi_tag)</a></li>
+<li><a class="reference internal" href="#acquire-capability-acquire-shared-capability-clang-acquire-capability-clang-acquire-shared-capability" id="id143">acquire_capability (acquire_shared_capability, clang::acquire_capability, clang::acquire_shared_capability)</a></li>
+<li><a class="reference internal" href="#alloc-align-gnu-alloc-align" id="id144">alloc_align (gnu::alloc_align)</a></li>
+<li><a class="reference internal" href="#alloc-size-gnu-alloc-size" id="id145">alloc_size (gnu::alloc_size)</a></li>
+<li><a class="reference internal" href="#artificial-gnu-artificial" id="id146">artificial (gnu::artificial)</a></li>
+<li><a class="reference internal" href="#assert-capability-assert-shared-capability-clang-assert-capability-clang-assert-shared-capability" id="id147">assert_capability (assert_shared_capability, clang::assert_capability, clang::assert_shared_capability)</a></li>
+<li><a class="reference internal" href="#assume-aligned-gnu-assume-aligned" id="id148">assume_aligned (gnu::assume_aligned)</a></li>
+<li><a class="reference internal" href="#availability-clang-availability-clang-availability" id="id149">availability (clang::availability, clang::availability)</a></li>
+<li><a class="reference internal" href="#carries-dependency" id="id150">carries_dependency</a></li>
+<li><a class="reference internal" href="#code-seg" id="id151">code_seg</a></li>
+<li><a class="reference internal" href="#convergent-clang-convergent-clang-convergent" id="id152">convergent (clang::convergent, clang::convergent)</a></li>
+<li><a class="reference internal" href="#cpu-dispatch-clang-cpu-dispatch-clang-cpu-dispatch" id="id153">cpu_dispatch (clang::cpu_dispatch, clang::cpu_dispatch)</a></li>
+<li><a class="reference internal" href="#cpu-specific-clang-cpu-specific-clang-cpu-specific" id="id154">cpu_specific (clang::cpu_specific, clang::cpu_specific)</a></li>
+<li><a class="reference internal" href="#deprecated-gnu-deprecated" id="id155">deprecated (gnu::deprecated)</a></li>
+<li><a class="reference internal" href="#diagnose-if" id="id156">diagnose_if</a></li>
+<li><a class="reference internal" href="#disable-tail-calls-clang-disable-tail-calls-clang-disable-tail-calls" id="id157">disable_tail_calls (clang::disable_tail_calls, clang::disable_tail_calls)</a></li>
+<li><a class="reference internal" href="#enable-if" id="id158">enable_if</a></li>
+<li><a class="reference internal" href="#external-source-symbol-clang-external-source-symbol-clang-external-source-symbol" id="id159">external_source_symbol (clang::external_source_symbol, clang::external_source_symbol)</a></li>
+<li><a class="reference internal" href="#flatten-gnu-flatten" id="id160">flatten (gnu::flatten)</a></li>
+<li><a class="reference internal" href="#force-align-arg-pointer-gnu-force-align-arg-pointer" id="id161">force_align_arg_pointer (gnu::force_align_arg_pointer)</a></li>
+<li><a class="reference internal" href="#format-gnu-format" id="id162">format (gnu::format)</a></li>
+<li><a class="reference internal" href="#ifunc-gnu-ifunc" id="id163">ifunc (gnu::ifunc)</a></li>
+<li><a class="reference internal" href="#internal-linkage-clang-internal-linkage-clang-internal-linkage" id="id164">internal_linkage (clang::internal_linkage, clang::internal_linkage)</a></li>
+<li><a class="reference internal" href="#interrupt-arm" id="id165">interrupt (ARM)</a></li>
+<li><a class="reference internal" href="#interrupt-avr" id="id166">interrupt (AVR)</a></li>
+<li><a class="reference internal" href="#interrupt-mips" id="id167">interrupt (MIPS)</a></li>
+<li><a class="reference internal" href="#interrupt-riscv" id="id168">interrupt (RISCV)</a></li>
+<li><a class="reference internal" href="#kernel" id="id169">kernel</a></li>
+<li><a class="reference internal" href="#lifetimebound-clang-lifetimebound" id="id170">lifetimebound (clang::lifetimebound)</a></li>
+<li><a class="reference internal" href="#long-call-gnu-long-call-gnu-far" id="id171">long_call (gnu::long_call, gnu::far)</a></li>
+<li><a class="reference internal" href="#micromips-gnu-micromips" id="id172">micromips (gnu::micromips)</a></li>
+<li><a class="reference internal" href="#min-vector-width-clang-min-vector-width-clang-min-vector-width" id="id173">min_vector_width (clang::min_vector_width, clang::min_vector_width)</a></li>
+<li><a class="reference internal" href="#no-caller-saved-registers-gnu-no-caller-saved-registers" id="id174">no_caller_saved_registers (gnu::no_caller_saved_registers)</a></li>
+<li><a class="reference internal" href="#no-sanitize-clang-no-sanitize-clang-no-sanitize" id="id175">no_sanitize (clang::no_sanitize, clang::no_sanitize)</a></li>
+<li><a class="reference internal" href="#no-sanitize-address-no-address-safety-analysis-gnu-no-address-safety-analysis-gnu-no-sanitize-address" id="id176">no_sanitize_address (no_address_safety_analysis, gnu::no_address_safety_analysis, gnu::no_sanitize_address)</a></li>
+<li><a class="reference internal" href="#no-sanitize-memory" id="id177">no_sanitize_memory</a></li>
+<li><a class="reference internal" href="#no-sanitize-thread" id="id178">no_sanitize_thread</a></li>
+<li><a class="reference internal" href="#no-split-stack-gnu-no-split-stack" id="id179">no_split_stack (gnu::no_split_stack)</a></li>
+<li><a class="reference internal" href="#no-stack-protector-clang-no-stack-protector-clang-no-stack-protector" id="id180">no_stack_protector (clang::no_stack_protector, clang::no_stack_protector)</a></li>
+<li><a class="reference internal" href="#noalias" id="id181">noalias</a></li>
+<li><a class="reference internal" href="#nocf-check-gnu-nocf-check" id="id182">nocf_check (gnu::nocf_check)</a></li>
+<li><a class="reference internal" href="#nodiscard-warn-unused-result-clang-warn-unused-result-gnu-warn-unused-result" id="id183">nodiscard, warn_unused_result, clang::warn_unused_result, gnu::warn_unused_result</a></li>
+<li><a class="reference internal" href="#noduplicate-clang-noduplicate-clang-noduplicate" id="id184">noduplicate (clang::noduplicate, clang::noduplicate)</a></li>
+<li><a class="reference internal" href="#nomicromips-gnu-nomicromips" id="id185">nomicromips (gnu::nomicromips)</a></li>
+<li><a class="reference internal" href="#id1" id="id186">noreturn</a></li>
+<li><a class="reference internal" href="#not-tail-called-clang-not-tail-called-clang-not-tail-called" id="id187">not_tail_called (clang::not_tail_called, clang::not_tail_called)</a></li>
+<li><a class="reference internal" href="#nothrow-gnu-nothrow" id="id188">nothrow (gnu::nothrow)</a></li>
+<li><a class="reference internal" href="#objc-boxable-clang-objc-boxable-clang-objc-boxable" id="id189">objc_boxable (clang::objc_boxable, clang::objc_boxable)</a></li>
+<li><a class="reference internal" href="#objc-method-family-clang-objc-method-family-clang-objc-method-family" id="id190">objc_method_family (clang::objc_method_family, clang::objc_method_family)</a></li>
+<li><a class="reference internal" href="#objc-requires-super-clang-objc-requires-super-clang-objc-requires-super" id="id191">objc_requires_super (clang::objc_requires_super, clang::objc_requires_super)</a></li>
+<li><a class="reference internal" href="#objc-runtime-name-clang-objc-runtime-name-clang-objc-runtime-name" id="id192">objc_runtime_name (clang::objc_runtime_name, clang::objc_runtime_name)</a></li>
+<li><a class="reference internal" href="#objc-runtime-visible-clang-objc-runtime-visible-clang-objc-runtime-visible" id="id193">objc_runtime_visible (clang::objc_runtime_visible, clang::objc_runtime_visible)</a></li>
+<li><a class="reference internal" href="#optnone-clang-optnone-clang-optnone" id="id194">optnone (clang::optnone, clang::optnone)</a></li>
+<li><a class="reference internal" href="#overloadable-clang-overloadable-clang-overloadable" id="id195">overloadable (clang::overloadable, clang::overloadable)</a></li>
+<li><a class="reference internal" href="#release-capability-release-shared-capability-clang-release-capability-clang-release-shared-capability" id="id196">release_capability (release_shared_capability, clang::release_capability, clang::release_shared_capability)</a></li>
+<li><a class="reference internal" href="#short-call-gnu-short-call-gnu-near" id="id197">short_call (gnu::short_call, gnu::near)</a></li>
+<li><a class="reference internal" href="#signal-gnu-signal" id="id198">signal (gnu::signal)</a></li>
+<li><a class="reference internal" href="#target-gnu-target" id="id199">target (gnu::target)</a></li>
+<li><a class="reference internal" href="#try-acquire-capability-try-acquire-shared-capability-clang-try-acquire-capability-clang-try-acquire-shared-capability" id="id200">try_acquire_capability (try_acquire_shared_capability, clang::try_acquire_capability, clang::try_acquire_shared_capability)</a></li>
+<li><a class="reference internal" href="#xray-always-instrument-clang-xray-always-instrument-xray-never-instrument-clang-xray-never-instrument-xray-log-args-clang-xray-log-args" id="id201">xray_always_instrument (clang::xray_always_instrument), xray_never_instrument (clang::xray_never_instrument), xray_log_args (clang::xray_log_args)</a></li>
+<li><a class="reference internal" href="#id2" id="id202">xray_always_instrument (clang::xray_always_instrument), xray_never_instrument (clang::xray_never_instrument), xray_log_args (clang::xray_log_args)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#variable-attributes" id="id203">Variable Attributes</a><ul>
+<li><a class="reference internal" href="#dllexport-gnu-dllexport" id="id204">dllexport (gnu::dllexport)</a></li>
+<li><a class="reference internal" href="#dllimport-gnu-dllimport" id="id205">dllimport (gnu::dllimport)</a></li>
+<li><a class="reference internal" href="#init-seg" id="id206">init_seg</a></li>
+<li><a class="reference internal" href="#maybe-unused-unused-gnu-unused" id="id207">maybe_unused, unused, gnu::unused</a></li>
+<li><a class="reference internal" href="#nodebug-gnu-nodebug" id="id208">nodebug (gnu::nodebug)</a></li>
+<li><a class="reference internal" href="#noescape-clang-noescape-clang-noescape" id="id209">noescape (clang::noescape, clang::noescape)</a></li>
+<li><a class="reference internal" href="#nosvm" id="id210">nosvm</a></li>
+<li><a class="reference internal" href="#pass-object-size-clang-pass-object-size-clang-pass-object-size" id="id211">pass_object_size (clang::pass_object_size, clang::pass_object_size)</a></li>
+<li><a class="reference internal" href="#require-constant-initialization-clang-require-constant-initialization" id="id212">require_constant_initialization (clang::require_constant_initialization)</a></li>
+<li><a class="reference internal" href="#section-gnu-section-declspec-allocate" id="id213">section (gnu::section, __declspec(allocate))</a></li>
+<li><a class="reference internal" href="#swift-context-clang-swift-context-clang-swift-context" id="id214">swift_context (clang::swift_context, clang::swift_context)</a></li>
+<li><a class="reference internal" href="#swift-error-result-clang-swift-error-result-clang-swift-error-result" id="id215">swift_error_result (clang::swift_error_result, clang::swift_error_result)</a></li>
+<li><a class="reference internal" href="#swift-indirect-result-clang-swift-indirect-result-clang-swift-indirect-result" id="id216">swift_indirect_result (clang::swift_indirect_result, clang::swift_indirect_result)</a></li>
+<li><a class="reference internal" href="#swiftcall-clang-swiftcall-clang-swiftcall" id="id217">swiftcall (clang::swiftcall, clang::swiftcall)</a></li>
+<li><a class="reference internal" href="#thread" id="id218">thread</a></li>
+<li><a class="reference internal" href="#tls-model-gnu-tls-model" id="id219">tls_model (gnu::tls_model)</a></li>
+<li><a class="reference internal" href="#trivial-abi-clang-trivial-abi" id="id220">trivial_abi (clang::trivial_abi)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#type-attributes" id="id221">Type Attributes</a><ul>
+<li><a class="reference internal" href="#single-inhertiance-multiple-inheritance-virtual-inheritance" id="id222">__single_inhertiance, __multiple_inheritance, __virtual_inheritance</a></li>
+<li><a class="reference internal" href="#align-value" id="id223">align_value</a></li>
+<li><a class="reference internal" href="#empty-bases" id="id224">empty_bases</a></li>
+<li><a class="reference internal" href="#enum-extensibility-clang-enum-extensibility-clang-enum-extensibility" id="id225">enum_extensibility (clang::enum_extensibility, clang::enum_extensibility)</a></li>
+<li><a class="reference internal" href="#flag-enum-clang-flag-enum-clang-flag-enum" id="id226">flag_enum (clang::flag_enum, clang::flag_enum)</a></li>
+<li><a class="reference internal" href="#layout-version" id="id227">layout_version</a></li>
+<li><a class="reference internal" href="#lto-visibility-public-clang-lto-visibility-public-clang-lto-visibility-public" id="id228">lto_visibility_public (clang::lto_visibility_public, clang::lto_visibility_public)</a></li>
+<li><a class="reference internal" href="#novtable" id="id229">novtable</a></li>
+<li><a class="reference internal" href="#objc-subclassing-restricted-clang-objc-subclassing-restricted-clang-objc-subclassing-restricted" id="id230">objc_subclassing_restricted (clang::objc_subclassing_restricted, clang::objc_subclassing_restricted)</a></li>
+<li><a class="reference internal" href="#selectany-gnu-selectany" id="id231">selectany (gnu::selectany)</a></li>
+<li><a class="reference internal" href="#transparent-union-gnu-transparent-union" id="id232">transparent_union (gnu::transparent_union)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#statement-attributes" id="id233">Statement Attributes</a><ul>
+<li><a class="reference internal" href="#pragma-clang-loop" id="id234">#pragma clang loop</a></li>
+<li><a class="reference internal" href="#pragma-unroll-pragma-nounroll" id="id235">#pragma unroll, #pragma nounroll</a></li>
+<li><a class="reference internal" href="#attribute-intel-reqd-sub-group-size" id="id236">__attribute__((intel_reqd_sub_group_size))</a></li>
+<li><a class="reference internal" href="#attribute-opencl-unroll-hint" id="id237">__attribute__((opencl_unroll_hint))</a></li>
+<li><a class="reference internal" href="#read-only-write-only-read-write-read-only-write-only-read-write" id="id238">__read_only, __write_only, __read_write (read_only, write_only, read_write)</a></li>
+<li><a class="reference internal" href="#fallthrough-clang-fallthrough" id="id239">fallthrough, clang::fallthrough</a></li>
+<li><a class="reference internal" href="#suppress-gsl-suppress" id="id240">suppress (gsl::suppress)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#amd-gpu-attributes" id="id241">AMD GPU Attributes</a><ul>
+<li><a class="reference internal" href="#amdgpu-flat-work-group-size-clang-amdgpu-flat-work-group-size" id="id242">amdgpu_flat_work_group_size (clang::amdgpu_flat_work_group_size)</a></li>
+<li><a class="reference internal" href="#amdgpu-num-sgpr-clang-amdgpu-num-sgpr" id="id243">amdgpu_num_sgpr (clang::amdgpu_num_sgpr)</a></li>
+<li><a class="reference internal" href="#amdgpu-num-vgpr-clang-amdgpu-num-vgpr" id="id244">amdgpu_num_vgpr (clang::amdgpu_num_vgpr)</a></li>
+<li><a class="reference internal" href="#amdgpu-waves-per-eu-clang-amdgpu-waves-per-eu" id="id245">amdgpu_waves_per_eu (clang::amdgpu_waves_per_eu)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#calling-conventions" id="id246">Calling Conventions</a><ul>
+<li><a class="reference internal" href="#fastcall-gnu-fastcall-fastcall-fastcall" id="id247">fastcall (gnu::fastcall, __fastcall, _fastcall)</a></li>
+<li><a class="reference internal" href="#ms-abi-gnu-ms-abi" id="id248">ms_abi (gnu::ms_abi)</a></li>
+<li><a class="reference internal" href="#pcs-gnu-pcs" id="id249">pcs (gnu::pcs)</a></li>
+<li><a class="reference internal" href="#preserve-all-clang-preserve-all-clang-preserve-all" id="id250">preserve_all (clang::preserve_all, clang::preserve_all)</a></li>
+<li><a class="reference internal" href="#preserve-most-clang-preserve-most-clang-preserve-most" id="id251">preserve_most (clang::preserve_most, clang::preserve_most)</a></li>
+<li><a class="reference internal" href="#regcall-gnu-regcall-regcall" id="id252">regcall (gnu::regcall, __regcall)</a></li>
+<li><a class="reference internal" href="#regparm-gnu-regparm" id="id253">regparm (gnu::regparm)</a></li>
+<li><a class="reference internal" href="#stdcall-gnu-stdcall-stdcall-stdcall" id="id254">stdcall (gnu::stdcall, __stdcall, _stdcall)</a></li>
+<li><a class="reference internal" href="#thiscall-gnu-thiscall-thiscall-thiscall" id="id255">thiscall (gnu::thiscall, __thiscall, _thiscall)</a></li>
+<li><a class="reference internal" href="#vectorcall-clang-vectorcall-clang-vectorcall-vectorcall-vectorcall" id="id256">vectorcall (clang::vectorcall, clang::vectorcall, __vectorcall, _vectorcall)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#consumed-annotation-checking" id="id257">Consumed Annotation Checking</a><ul>
+<li><a class="reference internal" href="#callable-when-clang-callable-when" id="id258">callable_when (clang::callable_when)</a></li>
+<li><a class="reference internal" href="#consumable-clang-consumable" id="id259">consumable (clang::consumable)</a></li>
+<li><a class="reference internal" href="#param-typestate-clang-param-typestate" id="id260">param_typestate (clang::param_typestate)</a></li>
+<li><a class="reference internal" href="#return-typestate-clang-return-typestate" id="id261">return_typestate (clang::return_typestate)</a></li>
+<li><a class="reference internal" href="#set-typestate-clang-set-typestate" id="id262">set_typestate (clang::set_typestate)</a></li>
+<li><a class="reference internal" href="#test-typestate-clang-test-typestate" id="id263">test_typestate (clang::test_typestate)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#type-safety-checking" id="id264">Type Safety Checking</a><ul>
+<li><a class="reference internal" href="#argument-with-type-tag" id="id265">argument_with_type_tag</a></li>
+<li><a class="reference internal" href="#pointer-with-type-tag" id="id266">pointer_with_type_tag</a></li>
+<li><a class="reference internal" href="#type-tag-for-datatype-clang-type-tag-for-datatype-clang-type-tag-for-datatype" id="id267">type_tag_for_datatype (clang::type_tag_for_datatype, clang::type_tag_for_datatype)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#opencl-address-spaces" id="id268">OpenCL Address Spaces</a><ul>
+<li><a class="reference internal" href="#constant-constant" id="id269">constant (__constant)</a></li>
+<li><a class="reference internal" href="#generic-generic" id="id270">generic (__generic)</a></li>
+<li><a class="reference internal" href="#global-global" id="id271">global (__global)</a></li>
+<li><a class="reference internal" href="#local-local" id="id272">local (__local)</a></li>
+<li><a class="reference internal" href="#private-private" id="id273">private (__private)</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#nullability-attributes" id="id274">Nullability Attributes</a><ul>
+<li><a class="reference internal" href="#nonnull" id="id275">_Nonnull</a></li>
+<li><a class="reference internal" href="#null-unspecified" id="id276">_Null_unspecified</a></li>
+<li><a class="reference internal" href="#nullable" id="id277">_Nullable</a></li>
+<li><a class="reference internal" href="#nonnull-gnu-nonnull" id="id278">nonnull (gnu::nonnull)</a></li>
+<li><a class="reference internal" href="#returns-nonnull-gnu-returns-nonnull" id="id279">returns_nonnull (gnu::returns_nonnull)</a></li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="section" id="introduction">
+<h2><a class="toc-backref" href="#id137">Introduction</a><a class="headerlink" href="#introduction" title="Permalink to this headline">¶</a></h2>
+<p>This page lists the attributes currently supported by Clang.</p>
+</div>
+<div class="section" id="function-attributes">
+<h2><a class="toc-backref" href="#id138">Function Attributes</a><a class="headerlink" href="#function-attributes" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="pragma-omp-declare-simd">
+<h3><a class="toc-backref" href="#id139">#pragma omp declare simd</a><a class="headerlink" href="#pragma-omp-declare-simd" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id5">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id5" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <cite>declare simd</cite> construct can be applied to a function to enable the creation
+of one or more versions that can process multiple arguments using SIMD
+instructions from a single invocation in a SIMD loop. The <cite>declare simd</cite>
+directive is a declarative directive. There may be multiple <cite>declare simd</cite>
+directives for a function. The use of a <cite>declare simd</cite> construct on a function
+enables the creation of SIMD versions of the associated function that can be
+used to process multiple arguments from a single invocation from a SIMD loop
+concurrently.
+The syntax of the <cite>declare simd</cite> construct is as follows:</p>
+<blockquote>
+<div><div class="highlight-none notranslate"><div class="highlight"><pre><span></span>#pragma omp declare simd [clause[[,] clause] ...] new-line
+[#pragma omp declare simd [clause[[,] clause] ...] new-line]
+[...]
+function definition or declaration
+</pre></div>
+</div>
+</div></blockquote>
+<p>where clause is one of the following:</p>
+<blockquote>
+<div><div class="highlight-none notranslate"><div class="highlight"><pre><span></span>simdlen(length)
+linear(argument-list[:constant-linear-step])
+aligned(argument-list[:alignment])
+uniform(argument-list)
+inbranch
+notinbranch
+</pre></div>
+</div>
+</div></blockquote>
+</div>
+<div class="section" id="pragma-omp-declare-target">
+<h3><a class="toc-backref" href="#id140">#pragma omp declare target</a><a class="headerlink" href="#pragma-omp-declare-target" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id6">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id6" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <cite>declare target</cite> directive specifies that variables and functions are mapped
+to a device for OpenMP offload mechanism.</p>
+<p>The syntax of the declare target directive is as follows:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="cp">#pragma omp declare target new-line</span>
+<span class="n">declarations</span><span class="o">-</span><span class="n">definition</span><span class="o">-</span><span class="n">seq</span>
+<span class="cp">#pragma omp end declare target new-line</span>
+</pre></div>
+</div>
+</div></blockquote>
+</div>
+<div class="section" id="noreturn">
+<h3><a class="toc-backref" href="#id141">_Noreturn</a><a class="headerlink" href="#noreturn" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id7">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id7" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>A function declared as <code class="docutils literal notranslate"><span class="pre">_Noreturn</span></code> shall not return to its caller. The
+compiler will generate a diagnostic for a function declared as <code class="docutils literal notranslate"><span class="pre">_Noreturn</span></code>
+that appears to be capable of returning to its caller.</p>
+</div>
+<div class="section" id="abi-tag-gnu-abi-tag">
+<h3><a class="toc-backref" href="#id142">abi_tag (gnu::abi_tag)</a><a class="headerlink" href="#abi-tag-gnu-abi-tag" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id8">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id8" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">abi_tag</span></code> attribute can be applied to a function, variable, class or
+inline namespace declaration to modify the mangled name of the entity. It gives
+the ability to distinguish between different versions of the same entity but
+with different ABI versions supported. For example, a newer version of a class
+could have a different set of data members and thus have a different size. Using
+the <code class="docutils literal notranslate"><span class="pre">abi_tag</span></code> attribute, it is possible to have different mangled names for
+a global variable of the class type. Therefore, the old code could keep using
+the old manged name and the new code will use the new mangled name with tags.</p>
+</div>
+<div class="section" id="acquire-capability-acquire-shared-capability-clang-acquire-capability-clang-acquire-shared-capability">
+<h3><a class="toc-backref" href="#id143">acquire_capability (acquire_shared_capability, clang::acquire_capability, clang::acquire_shared_capability)</a><a class="headerlink" href="#acquire-capability-acquire-shared-capability-clang-acquire-capability-clang-acquire-shared-capability" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id9">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id9" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Marks a function as acquiring a capability.</p>
+</div>
+<div class="section" id="alloc-align-gnu-alloc-align">
+<h3><a class="toc-backref" href="#id144">alloc_align (gnu::alloc_align)</a><a class="headerlink" href="#alloc-align-gnu-alloc-align" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id10">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id10" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Use <code class="docutils literal notranslate"><span class="pre">__attribute__((alloc_align(<alignment>))</span></code> on a function
+declaration to specify that the return value of the function (which must be a
+pointer type) is at least as aligned as the value of the indicated parameter. The
+parameter is given by its index in the list of formal parameters; the first
+parameter has index 1 unless the function is a C++ non-static member function,
+in which case the first parameter has index 2 to account for the implicit <code class="docutils literal notranslate"><span class="pre">this</span></code>
+parameter.</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="c1">// The returned pointer has the alignment specified by the first parameter.</span>
+<span class="kt">void</span> <span class="o">*</span><span class="nf">a</span><span class="p">(</span><span class="kt">size_t</span> <span class="n">align</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">alloc_align</span><span class="p">(</span><span class="mi">1</span><span class="p">)));</span>
+
+<span class="c1">// The returned pointer has the alignment specified by the second parameter.</span>
+<span class="kt">void</span> <span class="o">*</span><span class="nf">b</span><span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">v</span><span class="p">,</span> <span class="kt">size_t</span> <span class="n">align</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">alloc_align</span><span class="p">(</span><span class="mi">2</span><span class="p">)));</span>
+
+<span class="c1">// The returned pointer has the alignment specified by the second visible</span>
+<span class="c1">// parameter, however it must be adjusted for the implicit 'this' parameter.</span>
+<span class="kt">void</span> <span class="o">*</span><span class="n">Foo</span><span class="o">::</span><span class="n">b</span><span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">v</span><span class="p">,</span> <span class="kt">size_t</span> <span class="n">align</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">alloc_align</span><span class="p">(</span><span class="mi">3</span><span class="p">)));</span>
+</pre></div>
+</div>
+<p>Note that this attribute merely informs the compiler that a function always
+returns a sufficiently aligned pointer. It does not cause the compiler to
+emit code to enforce that alignment. The behavior is undefined if the returned
+poitner is not sufficiently aligned.</p>
+</div>
+<div class="section" id="alloc-size-gnu-alloc-size">
+<h3><a class="toc-backref" href="#id145">alloc_size (gnu::alloc_size)</a><a class="headerlink" href="#alloc-size-gnu-alloc-size" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id11">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id11" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">alloc_size</span></code> attribute can be placed on functions that return pointers in
+order to hint to the compiler how many bytes of memory will be available at the
+returned pointer. <code class="docutils literal notranslate"><span class="pre">alloc_size</span></code> takes one or two arguments.</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">alloc_size(N)</span></code> implies that argument number N equals the number of
+available bytes at the returned pointer.</li>
+<li><code class="docutils literal notranslate"><span class="pre">alloc_size(N,</span> <span class="pre">M)</span></code> implies that the product of argument number N and
+argument number M equals the number of available bytes at the returned
+pointer.</li>
+</ul>
+<p>Argument numbers are 1-based.</p>
+<p>An example of how to use <code class="docutils literal notranslate"><span class="pre">alloc_size</span></code></p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="o">*</span><span class="nf">my_malloc</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">alloc_size</span><span class="p">(</span><span class="mi">1</span><span class="p">)));</span>
+<span class="kt">void</span> <span class="o">*</span><span class="nf">my_calloc</span><span class="p">(</span><span class="kt">int</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="n">__attribute__</span><span class="p">((</span><span class="n">alloc_size</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">)));</span>
+
+<span class="kt">int</span> <span class="nf">main</span><span class="p">()</span> <span class="p">{</span>
+ <span class="kt">void</span> <span class="o">*</span><span class="k">const</span> <span class="n">p</span> <span class="o">=</span> <span class="n">my_malloc</span><span class="p">(</span><span class="mi">100</span><span class="p">);</span>
+ <span class="n">assert</span><span class="p">(</span><span class="n">__builtin_object_size</span><span class="p">(</span><span class="n">p</span><span class="p">,</span> <span class="mi">0</span><span class="p">)</span> <span class="o">==</span> <span class="mi">100</span><span class="p">);</span>
+ <span class="kt">void</span> <span class="o">*</span><span class="k">const</span> <span class="n">a</span> <span class="o">=</span> <span class="n">my_calloc</span><span class="p">(</span><span class="mi">20</span><span class="p">,</span> <span class="mi">5</span><span class="p">);</span>
+ <span class="n">assert</span><span class="p">(</span><span class="n">__builtin_object_size</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="mi">0</span><span class="p">)</span> <span class="o">==</span> <span class="mi">100</span><span class="p">);</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<div class="admonition note">
+<p class="first admonition-title">Note</p>
+<p class="last">This attribute works differently in clang than it does in GCC.
+Specifically, clang will only trace <code class="docutils literal notranslate"><span class="pre">const</span></code> pointers (as above); we give up
+on pointers that are not marked as <code class="docutils literal notranslate"><span class="pre">const</span></code>. In the vast majority of cases,
+this is unimportant, because LLVM has support for the <code class="docutils literal notranslate"><span class="pre">alloc_size</span></code>
+attribute. However, this may cause mildly unintuitive behavior when used with
+other attributes, such as <code class="docutils literal notranslate"><span class="pre">enable_if</span></code>.</p>
+</div>
+</div>
+<div class="section" id="artificial-gnu-artificial">
+<h3><a class="toc-backref" href="#id146">artificial (gnu::artificial)</a><a class="headerlink" href="#artificial-gnu-artificial" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id12">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id12" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">artificial</span></code> attribute can be applied to an inline function. If such a
+function is inlined, the attribute indicates that debuggers should associate
+the resulting instructions with the call site, rather than with the
+corresponding line within the inlined callee.</p>
+</div>
+<div class="section" id="assert-capability-assert-shared-capability-clang-assert-capability-clang-assert-shared-capability">
+<h3><a class="toc-backref" href="#id147">assert_capability (assert_shared_capability, clang::assert_capability, clang::assert_shared_capability)</a><a class="headerlink" href="#assert-capability-assert-shared-capability-clang-assert-capability-clang-assert-shared-capability" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id13">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id13" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Marks a function that dynamically tests whether a capability is held, and halts
+the program if it is not held.</p>
+</div>
+<div class="section" id="assume-aligned-gnu-assume-aligned">
+<h3><a class="toc-backref" href="#id148">assume_aligned (gnu::assume_aligned)</a><a class="headerlink" href="#assume-aligned-gnu-assume-aligned" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id14">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id14" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Use <code class="docutils literal notranslate"><span class="pre">__attribute__((assume_aligned(<alignment>[,<offset>]))</span></code> on a function
+declaration to specify that the return value of the function (which must be a
+pointer type) has the specified offset, in bytes, from an address with the
+specified alignment. The offset is taken to be zero if omitted.</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="c1">// The returned pointer value has 32-byte alignment.</span>
+<span class="kt">void</span> <span class="o">*</span><span class="nf">a</span><span class="p">()</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">assume_aligned</span> <span class="p">(</span><span class="mi">32</span><span class="p">)));</span>
+
+<span class="c1">// The returned pointer value is 4 bytes greater than an address having</span>
+<span class="c1">// 32-byte alignment.</span>
+<span class="kt">void</span> <span class="o">*</span><span class="nf">b</span><span class="p">()</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">assume_aligned</span> <span class="p">(</span><span class="mi">32</span><span class="p">,</span> <span class="mi">4</span><span class="p">)));</span>
+</pre></div>
+</div>
+<p>Note that this attribute provides information to the compiler regarding a
+condition that the code already ensures is true. It does not cause the compiler
+to enforce the provided alignment assumption.</p>
+</div>
+<div class="section" id="availability-clang-availability-clang-availability">
+<h3><a class="toc-backref" href="#id149">availability (clang::availability, clang::availability)</a><a class="headerlink" href="#availability-clang-availability-clang-availability" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id15">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id15" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">availability</span></code> attribute can be placed on declarations to describe the
+lifecycle of that declaration relative to operating system versions. Consider
+the function declaration for a hypothetical function <code class="docutils literal notranslate"><span class="pre">f</span></code>:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">f</span><span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">macos</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">10.4</span><span class="p">,</span><span class="n">deprecated</span><span class="o">=</span><span class="mf">10.6</span><span class="p">,</span><span class="n">obsoleted</span><span class="o">=</span><span class="mf">10.7</span><span class="p">)));</span>
+</pre></div>
+</div>
+<p>The availability attribute states that <code class="docutils literal notranslate"><span class="pre">f</span></code> was introduced in macOS 10.4,
+deprecated in macOS 10.6, and obsoleted in macOS 10.7. This information
+is used by Clang to determine when it is safe to use <code class="docutils literal notranslate"><span class="pre">f</span></code>: for example, if
+Clang is instructed to compile code for macOS 10.5, a call to <code class="docutils literal notranslate"><span class="pre">f()</span></code>
+succeeds. If Clang is instructed to compile code for macOS 10.6, the call
+succeeds but Clang emits a warning specifying that the function is deprecated.
+Finally, if Clang is instructed to compile code for macOS 10.7, the call
+fails because <code class="docutils literal notranslate"><span class="pre">f()</span></code> is no longer available.</p>
+<p>The availability attribute is a comma-separated list starting with the
+platform name and then including clauses specifying important milestones in the
+declarationâs lifetime (in any order) along with additional information. Those
+clauses can be:</p>
+<dl class="docutils">
+<dt>introduced=<em>version</em></dt>
+<dd>The first version in which this declaration was introduced.</dd>
+<dt>deprecated=<em>version</em></dt>
+<dd>The first version in which this declaration was deprecated, meaning that
+users should migrate away from this API.</dd>
+<dt>obsoleted=<em>version</em></dt>
+<dd>The first version in which this declaration was obsoleted, meaning that it
+was removed completely and can no longer be used.</dd>
+<dt>unavailable</dt>
+<dd>This declaration is never available on this platform.</dd>
+<dt>message=<em>string-literal</em></dt>
+<dd>Additional message text that Clang will provide when emitting a warning or
+error about use of a deprecated or obsoleted declaration. Useful to direct
+users to replacement APIs.</dd>
+<dt>replacement=<em>string-literal</em></dt>
+<dd>Additional message text that Clang will use to provide Fix-It when emitting
+a warning about use of a deprecated declaration. The Fix-It will replace
+the deprecated declaration with the new declaration specified.</dd>
+</dl>
+<p>Multiple availability attributes can be placed on a declaration, which may
+correspond to different platforms. Only the availability attribute with the
+platform corresponding to the target platform will be used; any others will be
+ignored. If no availability attribute specifies availability for the current
+target platform, the availability attributes are ignored. Supported platforms
+are:</p>
+<dl class="docutils">
+<dt><code class="docutils literal notranslate"><span class="pre">ios</span></code></dt>
+<dd>Appleâs iOS operating system. The minimum deployment target is specified by
+the <code class="docutils literal notranslate"><span class="pre">-mios-version-min=*version*</span></code> or <code class="docutils literal notranslate"><span class="pre">-miphoneos-version-min=*version*</span></code>
+command-line arguments.</dd>
+<dt><code class="docutils literal notranslate"><span class="pre">macos</span></code></dt>
+<dd>Appleâs macOS operating system. The minimum deployment target is
+specified by the <code class="docutils literal notranslate"><span class="pre">-mmacosx-version-min=*version*</span></code> command-line argument.
+<code class="docutils literal notranslate"><span class="pre">macosx</span></code> is supported for backward-compatibility reasons, but it is
+deprecated.</dd>
+<dt><code class="docutils literal notranslate"><span class="pre">tvos</span></code></dt>
+<dd>Appleâs tvOS operating system. The minimum deployment target is specified by
+the <code class="docutils literal notranslate"><span class="pre">-mtvos-version-min=*version*</span></code> command-line argument.</dd>
+<dt><code class="docutils literal notranslate"><span class="pre">watchos</span></code></dt>
+<dd>Appleâs watchOS operating system. The minimum deployment target is specified by
+the <code class="docutils literal notranslate"><span class="pre">-mwatchos-version-min=*version*</span></code> command-line argument.</dd>
+</dl>
+<p>A declaration can typically be used even when deploying back to a platform
+version prior to when the declaration was introduced. When this happens, the
+declaration is <a class="reference external" href="https://developer.apple.com/library/mac/#documentation/MacOSX/Conceptual/BPFrameworks/Concepts/WeakLinking.html">weakly linked</a>,
+as if the <code class="docutils literal notranslate"><span class="pre">weak_import</span></code> attribute were added to the declaration. A
+weakly-linked declaration may or may not be present a run-time, and a program
+can determine whether the declaration is present by checking whether the
+address of that declaration is non-NULL.</p>
+<p>The flag <code class="docutils literal notranslate"><span class="pre">strict</span></code> disallows using API when deploying back to a
+platform version prior to when the declaration was introduced. An
+attempt to use such API before its introduction causes a hard error.
+Weakly-linking is almost always a better API choice, since it allows
+users to query availability at runtime.</p>
+<p>If there are multiple declarations of the same entity, the availability
+attributes must either match on a per-platform basis or later
+declarations must not have availability attributes for that
+platform. For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">g</span><span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">macos</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">10.4</span><span class="p">)));</span>
+<span class="kt">void</span> <span class="nf">g</span><span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">macos</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">10.4</span><span class="p">)));</span> <span class="c1">// okay, matches</span>
+<span class="kt">void</span> <span class="nf">g</span><span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">ios</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">4.0</span><span class="p">)));</span> <span class="c1">// okay, adds a new platform</span>
+<span class="kt">void</span> <span class="nf">g</span><span class="p">(</span><span class="kt">void</span><span class="p">);</span> <span class="c1">// okay, inherits both macos and ios availability from above.</span>
+<span class="kt">void</span> <span class="nf">g</span><span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">macos</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">10.5</span><span class="p">)));</span> <span class="c1">// error: mismatch</span>
+</pre></div>
+</div>
+<p>When one method overrides another, the overriding method can be more widely available than the overridden method, e.g.,:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="k">@interface</span> <span class="nc">A</span>
+<span class="p">-</span> <span class="p">(</span><span class="kt">id</span><span class="p">)</span><span class="nf">method</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">macos</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">10.4</span><span class="p">)));</span>
+<span class="p">-</span> <span class="p">(</span><span class="kt">id</span><span class="p">)</span><span class="nf">method2</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">macos</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">10.4</span><span class="p">)));</span>
+<span class="k">@end</span>
+
+<span class="k">@interface</span> <span class="nc">B</span> : <span class="nc">A</span>
+<span class="p">-</span> <span class="p">(</span><span class="kt">id</span><span class="p">)</span><span class="nf">method</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">macos</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">10.3</span><span class="p">)));</span> <span class="c1">// okay: method moved into base class later</span>
+<span class="p">-</span> <span class="p">(</span><span class="kt">id</span><span class="p">)</span><span class="nf">method</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">availability</span><span class="p">(</span><span class="n">macos</span><span class="p">,</span><span class="n">introduced</span><span class="o">=</span><span class="mf">10.5</span><span class="p">)));</span> <span class="c1">// error: this method was available via the base class in 10.4</span>
+<span class="k">@end</span>
+</pre></div>
+</div>
+<p>Starting with the macOS 10.12 SDK, the <code class="docutils literal notranslate"><span class="pre">API_AVAILABLE</span></code> macro from
+<code class="docutils literal notranslate"><span class="pre"><os/availability.h></span></code> can simplify the spelling:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="k">@interface</span> <span class="nc">A</span>
+<span class="p">-</span> <span class="p">(</span><span class="kt">id</span><span class="p">)</span><span class="nf">method</span> <span class="n">API_AVAILABLE</span><span class="p">(</span><span class="n">macos</span><span class="p">(</span><span class="mf">10.11</span><span class="p">)));</span>
+<span class="p">-</span> <span class="p">(</span><span class="kt">id</span><span class="p">)</span><span class="nf">otherMethod</span> <span class="n">API_AVAILABLE</span><span class="p">(</span><span class="n">macos</span><span class="p">(</span><span class="mf">10.11</span><span class="p">),</span> <span class="n">ios</span><span class="p">(</span><span class="mf">11.0</span><span class="p">));</span>
+<span class="k">@end</span>
+</pre></div>
+</div>
+<p>Also see the documentation for <a class="reference external" href="http://clang.llvm.org/docs/LanguageExtensions.html#objective-c-available">@available</a></p>
+</div>
+<div class="section" id="carries-dependency">
+<h3><a class="toc-backref" href="#id150">carries_dependency</a><a class="headerlink" href="#carries-dependency" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id16">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id16" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">carries_dependency</span></code> attribute specifies dependency propagation into and
+out of functions.</p>
+<p>When specified on a function or Objective-C method, the <code class="docutils literal notranslate"><span class="pre">carries_dependency</span></code>
+attribute means that the return value carries a dependency out of the function,
+so that the implementation need not constrain ordering upon return from that
+function. Implementations of the function and its caller may choose to preserve
+dependencies instead of emitting memory ordering instructions such as fences.</p>
+<p>Note, this attribute does not change the meaning of the program, but may result
+in generation of more efficient code.</p>
+</div>
+<div class="section" id="code-seg">
+<h3><a class="toc-backref" href="#id151">code_seg</a><a class="headerlink" href="#code-seg" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id17">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id17" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">__declspec(code_seg)</span></code> attribute enables the placement of code into separate
+named segments that can be paged or locked in memory individually. This attribute
+is used to control the placement of instantiated templates and compiler-generated
+code. See the documentation for <a class="reference external" href="http://msdn.microsoft.com/en-us/library/dn636922.aspx">__declspec(code_seg)</a> on MSDN.</p>
+</div>
+<div class="section" id="convergent-clang-convergent-clang-convergent">
+<h3><a class="toc-backref" href="#id152">convergent (clang::convergent, clang::convergent)</a><a class="headerlink" href="#convergent-clang-convergent-clang-convergent" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id18">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id18" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">convergent</span></code> attribute can be placed on a function declaration. It is
+translated into the LLVM <code class="docutils literal notranslate"><span class="pre">convergent</span></code> attribute, which indicates that the call
+instructions of a function with this attribute cannot be made control-dependent
+on any additional values.</p>
+<p>In languages designed for SPMD/SIMT programming model, e.g. OpenCL or CUDA,
+the call instructions of a function with this attribute must be executed by
+all work items or threads in a work group or sub group.</p>
+<p>This attribute is different from <code class="docutils literal notranslate"><span class="pre">noduplicate</span></code> because it allows duplicating
+function calls if it can be proved that the duplicated function calls are
+not made control-dependent on any additional values, e.g., unrolling a loop
+executed by all work items.</p>
+<p>Sample usage:
+.. code-block:: c</p>
+<blockquote>
+<div>void convfunc(void) __attribute__((convergent));
+// Setting it as a C++11 attribute is also valid in a C++ program.
+// void convfunc(void) [[clang::convergent]];</div></blockquote>
+</div>
+<div class="section" id="cpu-dispatch-clang-cpu-dispatch-clang-cpu-dispatch">
+<h3><a class="toc-backref" href="#id153">cpu_dispatch (clang::cpu_dispatch, clang::cpu_dispatch)</a><a class="headerlink" href="#cpu-dispatch-clang-cpu-dispatch-clang-cpu-dispatch" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id19">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id19" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">cpu_specific</span></code> and <code class="docutils literal notranslate"><span class="pre">cpu_dispatch</span></code> attributes are used to define and
+resolve multiversioned functions. This form of multiversioning provides a
+mechanism for declaring versions across translation units and manually
+specifying the resolved function list. A specified CPU defines a set of minimum
+features that are required for the function to be called. The result of this is
+that future processors execute the most restrictive version of the function the
+new processor can execute.</p>
+<p>Function versions are defined with <code class="docutils literal notranslate"><span class="pre">cpu_specific</span></code>, which takes one or more CPU
+names as a parameter. For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="c1">// Declares and defines the ivybridge version of single_cpu.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_specific</span><span class="p">(</span><span class="n">ivybridge</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">single_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+
+<span class="c1">// Declares and defines the atom version of single_cpu.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_specific</span><span class="p">(</span><span class="n">atom</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">single_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+
+<span class="c1">// Declares and defines both the ivybridge and atom version of multi_cpu.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_specific</span><span class="p">(</span><span class="n">ivybridge</span><span class="p">,</span> <span class="n">atom</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">multi_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+</pre></div>
+</div>
+<p>A dispatching (or resolving) function can be declared anywhere in a projectâs
+source code with <code class="docutils literal notranslate"><span class="pre">cpu_dispatch</span></code>. This attribute takes one or more CPU names
+as a parameter (like <code class="docutils literal notranslate"><span class="pre">cpu_specific</span></code>). Functions marked with <code class="docutils literal notranslate"><span class="pre">cpu_dispatch</span></code>
+are not expected to be defined, only declared. If such a marked function has a
+definition, any side effects of the function are ignored; trivial function
+bodies are permissible for ICC compatibility.</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="c1">// Creates a resolver for single_cpu above.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_dispatch</span><span class="p">(</span><span class="n">ivybridge</span><span class="p">,</span> <span class="n">atom</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">single_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+
+<span class="c1">// Creates a resolver for multi_cpu, but adds a 3rd version defined in another</span>
+<span class="c1">// translation unit.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_dispatch</span><span class="p">(</span><span class="n">ivybridge</span><span class="p">,</span> <span class="n">atom</span><span class="p">,</span> <span class="n">sandybridge</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">multi_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+</pre></div>
+</div>
+<p>Note that it is possible to have a resolving function that dispatches based on
+more or fewer options than are present in the program. Specifying fewer will
+result in the omitted options not being considered during resolution. Specifying
+a version for resolution that isnât defined in the program will result in a
+linking failure.</p>
+<p>It is also possible to specify a CPU name of <code class="docutils literal notranslate"><span class="pre">generic</span></code> which will be resolved
+if the executing processor doesnât satisfy the features required in the CPU
+name. The behavior of a program executing on a processor that doesnât satisfy
+any option of a multiversioned function is undefined.</p>
+</div>
+<div class="section" id="cpu-specific-clang-cpu-specific-clang-cpu-specific">
+<h3><a class="toc-backref" href="#id154">cpu_specific (clang::cpu_specific, clang::cpu_specific)</a><a class="headerlink" href="#cpu-specific-clang-cpu-specific-clang-cpu-specific" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id20">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id20" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">cpu_specific</span></code> and <code class="docutils literal notranslate"><span class="pre">cpu_dispatch</span></code> attributes are used to define and
+resolve multiversioned functions. This form of multiversioning provides a
+mechanism for declaring versions across translation units and manually
+specifying the resolved function list. A specified CPU defines a set of minimum
+features that are required for the function to be called. The result of this is
+that future processors execute the most restrictive version of the function the
+new processor can execute.</p>
+<p>Function versions are defined with <code class="docutils literal notranslate"><span class="pre">cpu_specific</span></code>, which takes one or more CPU
+names as a parameter. For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="c1">// Declares and defines the ivybridge version of single_cpu.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_specific</span><span class="p">(</span><span class="n">ivybridge</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">single_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+
+<span class="c1">// Declares and defines the atom version of single_cpu.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_specific</span><span class="p">(</span><span class="n">atom</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">single_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+
+<span class="c1">// Declares and defines both the ivybridge and atom version of multi_cpu.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_specific</span><span class="p">(</span><span class="n">ivybridge</span><span class="p">,</span> <span class="n">atom</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">multi_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+</pre></div>
+</div>
+<p>A dispatching (or resolving) function can be declared anywhere in a projectâs
+source code with <code class="docutils literal notranslate"><span class="pre">cpu_dispatch</span></code>. This attribute takes one or more CPU names
+as a parameter (like <code class="docutils literal notranslate"><span class="pre">cpu_specific</span></code>). Functions marked with <code class="docutils literal notranslate"><span class="pre">cpu_dispatch</span></code>
+are not expected to be defined, only declared. If such a marked function has a
+definition, any side effects of the function are ignored; trivial function
+bodies are permissible for ICC compatibility.</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="c1">// Creates a resolver for single_cpu above.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_dispatch</span><span class="p">(</span><span class="n">ivybridge</span><span class="p">,</span> <span class="n">atom</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">single_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+
+<span class="c1">// Creates a resolver for multi_cpu, but adds a 3rd version defined in another</span>
+<span class="c1">// translation unit.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">cpu_dispatch</span><span class="p">(</span><span class="n">ivybridge</span><span class="p">,</span> <span class="n">atom</span><span class="p">,</span> <span class="n">sandybridge</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">multi_cpu</span><span class="p">(</span><span class="kt">void</span><span class="p">){}</span>
+</pre></div>
+</div>
+<p>Note that it is possible to have a resolving function that dispatches based on
+more or fewer options than are present in the program. Specifying fewer will
+result in the omitted options not being considered during resolution. Specifying
+a version for resolution that isnât defined in the program will result in a
+linking failure.</p>
+<p>It is also possible to specify a CPU name of <code class="docutils literal notranslate"><span class="pre">generic</span></code> which will be resolved
+if the executing processor doesnât satisfy the features required in the CPU
+name. The behavior of a program executing on a processor that doesnât satisfy
+any option of a multiversioned function is undefined.</p>
+</div>
+<div class="section" id="deprecated-gnu-deprecated">
+<h3><a class="toc-backref" href="#id155">deprecated (gnu::deprecated)</a><a class="headerlink" href="#deprecated-gnu-deprecated" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id21">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id21" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">deprecated</span></code> attribute can be applied to a function, a variable, or a
+type. This is useful when identifying functions, variables, or types that are
+expected to be removed in a future version of a program.</p>
+<p>Consider the function declaration for a hypothetical function <code class="docutils literal notranslate"><span class="pre">f</span></code>:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">f</span><span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">deprecated</span><span class="p">(</span><span class="s">"message"</span><span class="p">,</span> <span class="s">"replacement"</span><span class="p">)));</span>
+</pre></div>
+</div>
+<p>When spelled as <cite>__attribute__((deprecated))</cite>, the deprecated attribute can have
+two optional string arguments. The first one is the message to display when
+emitting the warning; the second one enables the compiler to provide a Fix-It
+to replace the deprecated name with a new name. Otherwise, when spelled as
+<cite>[[gnu::deprecated]] or [[deprecated]]</cite>, the attribute can have one optional
+string argument which is the message to display when emitting the warning.</p>
+</div>
+<div class="section" id="diagnose-if">
+<h3><a class="toc-backref" href="#id156">diagnose_if</a><a class="headerlink" href="#diagnose-if" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id22">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id22" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">diagnose_if</span></code> attribute can be placed on function declarations to emit
+warnings or errors at compile-time if calls to the attributed function meet
+certain user-defined criteria. For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">abs</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">diagnose_if</span><span class="p">(</span><span class="n">a</span> <span class="o">>=</span> <span class="mi">0</span><span class="p">,</span> <span class="s">"Redundant abs call"</span><span class="p">,</span> <span class="s">"warning"</span><span class="p">)));</span>
+<span class="kt">void</span> <span class="nf">must_abs</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">diagnose_if</span><span class="p">(</span><span class="n">a</span> <span class="o">>=</span> <span class="mi">0</span><span class="p">,</span> <span class="s">"Redundant abs call"</span><span class="p">,</span> <span class="s">"error"</span><span class="p">)));</span>
+
+<span class="kt">int</span> <span class="n">val</span> <span class="o">=</span> <span class="n">abs</span><span class="p">(</span><span class="mi">1</span><span class="p">);</span> <span class="c1">// warning: Redundant abs call</span>
+<span class="kt">int</span> <span class="n">val2</span> <span class="o">=</span> <span class="n">must_abs</span><span class="p">(</span><span class="mi">1</span><span class="p">);</span> <span class="c1">// error: Redundant abs call</span>
+<span class="kt">int</span> <span class="n">val3</span> <span class="o">=</span> <span class="n">abs</span><span class="p">(</span><span class="n">val</span><span class="p">);</span>
+<span class="kt">int</span> <span class="n">val4</span> <span class="o">=</span> <span class="n">must_abs</span><span class="p">(</span><span class="n">val</span><span class="p">);</span> <span class="c1">// Because run-time checks are not emitted for</span>
+ <span class="c1">// diagnose_if attributes, this executes without</span>
+ <span class="c1">// issue.</span>
+</pre></div>
+</div>
+<p><code class="docutils literal notranslate"><span class="pre">diagnose_if</span></code> is closely related to <code class="docutils literal notranslate"><span class="pre">enable_if</span></code>, with a few key differences:</p>
+<ul class="simple">
+<li>Overload resolution is not aware of <code class="docutils literal notranslate"><span class="pre">diagnose_if</span></code> attributes: theyâre
+considered only after we select the best candidate from a given candidate set.</li>
+<li>Function declarations that differ only in their <code class="docutils literal notranslate"><span class="pre">diagnose_if</span></code> attributes are
+considered to be redeclarations of the same function (not overloads).</li>
+<li>If the condition provided to <code class="docutils literal notranslate"><span class="pre">diagnose_if</span></code> cannot be evaluated, no
+diagnostic will be emitted.</li>
+</ul>
+<p>Otherwise, <code class="docutils literal notranslate"><span class="pre">diagnose_if</span></code> is essentially the logical negation of <code class="docutils literal notranslate"><span class="pre">enable_if</span></code>.</p>
+<p>As a result of bullet number two, <code class="docutils literal notranslate"><span class="pre">diagnose_if</span></code> attributes will stack on the
+same function. For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">foo</span><span class="p">()</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">diagnose_if</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="s">"diag1"</span><span class="p">,</span> <span class="s">"warning"</span><span class="p">)));</span>
+<span class="kt">int</span> <span class="nf">foo</span><span class="p">()</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">diagnose_if</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="s">"diag2"</span><span class="p">,</span> <span class="s">"warning"</span><span class="p">)));</span>
+
+<span class="kt">int</span> <span class="n">bar</span> <span class="o">=</span> <span class="n">foo</span><span class="p">();</span> <span class="c1">// warning: diag1</span>
+ <span class="c1">// warning: diag2</span>
+<span class="kt">int</span> <span class="p">(</span><span class="o">*</span><span class="n">fooptr</span><span class="p">)(</span><span class="kt">void</span><span class="p">)</span> <span class="o">=</span> <span class="n">foo</span><span class="p">;</span> <span class="c1">// warning: diag1</span>
+ <span class="c1">// warning: diag2</span>
+
+<span class="n">constexpr</span> <span class="kt">int</span> <span class="nf">supportsAPILevel</span><span class="p">(</span><span class="kt">int</span> <span class="n">N</span><span class="p">)</span> <span class="p">{</span> <span class="k">return</span> <span class="n">N</span> <span class="o"><</span> <span class="mi">5</span><span class="p">;</span> <span class="p">}</span>
+<span class="kt">int</span> <span class="nf">baz</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">diagnose_if</span><span class="p">(</span><span class="o">!</span><span class="n">supportsAPILevel</span><span class="p">(</span><span class="mi">10</span><span class="p">),</span>
+ <span class="s">"Upgrade to API level 10 to use baz"</span><span class="p">,</span> <span class="s">"error"</span><span class="p">)));</span>
+<span class="kt">int</span> <span class="nf">baz</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">diagnose_if</span><span class="p">(</span><span class="o">!</span><span class="n">a</span><span class="p">,</span> <span class="s">"0 is not recommended."</span><span class="p">,</span> <span class="s">"warning"</span><span class="p">)));</span>
+
+<span class="kt">int</span> <span class="p">(</span><span class="o">*</span><span class="n">bazptr</span><span class="p">)(</span><span class="kt">int</span><span class="p">)</span> <span class="o">=</span> <span class="n">baz</span><span class="p">;</span> <span class="c1">// error: Upgrade to API level 10 to use baz</span>
+<span class="kt">int</span> <span class="n">v</span> <span class="o">=</span> <span class="n">baz</span><span class="p">(</span><span class="mi">0</span><span class="p">);</span> <span class="c1">// error: Upgrade to API level 10 to use baz</span>
+</pre></div>
+</div>
+<p>Query for this feature with <code class="docutils literal notranslate"><span class="pre">__has_attribute(diagnose_if)</span></code>.</p>
+</div>
+<div class="section" id="disable-tail-calls-clang-disable-tail-calls-clang-disable-tail-calls">
+<h3><a class="toc-backref" href="#id157">disable_tail_calls (clang::disable_tail_calls, clang::disable_tail_calls)</a><a class="headerlink" href="#disable-tail-calls-clang-disable-tail-calls-clang-disable-tail-calls" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id23">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id23" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">disable_tail_calls</span></code> attribute instructs the backend to not perform tail call optimization inside the marked function.</p>
+<p>For example:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">callee</span><span class="p">(</span><span class="kt">int</span><span class="p">);</span>
+
+<span class="kt">int</span> <span class="nf">foo</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">disable_tail_calls</span><span class="p">))</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="n">callee</span><span class="p">(</span><span class="n">a</span><span class="p">);</span> <span class="c1">// This call is not tail-call optimized.</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>Marking virtual functions as <code class="docutils literal notranslate"><span class="pre">disable_tail_calls</span></code> is legal.</p>
+<blockquote>
+<div><div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">callee</span><span class="p">(</span><span class="kt">int</span><span class="p">);</span>
+
+<span class="k">class</span> <span class="nc">Base</span> <span class="p">{</span>
+<span class="k">public</span><span class="o">:</span>
+ <span class="p">[[</span><span class="n">clang</span><span class="o">::</span><span class="n">disable_tail_calls</span><span class="p">]]</span> <span class="k">virtual</span> <span class="kt">int</span> <span class="n">foo1</span><span class="p">()</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="n">callee</span><span class="p">();</span> <span class="c1">// This call is not tail-call optimized.</span>
+ <span class="p">}</span>
+<span class="p">};</span>
+
+<span class="k">class</span> <span class="nc">Derived1</span> <span class="o">:</span> <span class="k">public</span> <span class="n">Base</span> <span class="p">{</span>
+<span class="k">public</span><span class="o">:</span>
+ <span class="kt">int</span> <span class="n">foo1</span><span class="p">()</span> <span class="k">override</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="n">callee</span><span class="p">();</span> <span class="c1">// This call is tail-call optimized.</span>
+ <span class="p">}</span>
+<span class="p">};</span>
+</pre></div>
+</div>
+</div></blockquote>
+</div>
+<div class="section" id="enable-if">
+<h3><a class="toc-backref" href="#id158">enable_if</a><a class="headerlink" href="#enable-if" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id24">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id24" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<div class="admonition note">
+<p class="first admonition-title">Note</p>
+<p class="last">Some features of this attribute are experimental. The meaning of
+multiple enable_if attributes on a single declaration is subject to change in
+a future version of clang. Also, the ABI is not standardized and the name
+mangling may change in future versions. To avoid that, use asm labels.</p>
+</div>
+<p>The <code class="docutils literal notranslate"><span class="pre">enable_if</span></code> attribute can be placed on function declarations to control
+which overload is selected based on the values of the functionâs arguments.
+When combined with the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute, this feature is also
+available in C.</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">isdigit</span><span class="p">(</span><span class="kt">int</span> <span class="n">c</span><span class="p">);</span>
+<span class="kt">int</span> <span class="nf">isdigit</span><span class="p">(</span><span class="kt">int</span> <span class="n">c</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">c</span> <span class="o"><=</span> <span class="o">-</span><span class="mi">1</span> <span class="o">||</span> <span class="n">c</span> <span class="o">></span> <span class="mi">255</span><span class="p">,</span> <span class="s">"chosen when 'c' is out of range"</span><span class="p">)))</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">unavailable</span><span class="p">(</span><span class="s">"'c' must have the value of an unsigned char or EOF"</span><span class="p">)));</span>
+
+<span class="kt">void</span> <span class="nf">foo</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="n">isdigit</span><span class="p">(</span><span class="n">c</span><span class="p">);</span>
+ <span class="n">isdigit</span><span class="p">(</span><span class="mi">10</span><span class="p">);</span>
+ <span class="n">isdigit</span><span class="p">(</span><span class="o">-</span><span class="mi">10</span><span class="p">);</span> <span class="c1">// results in a compile-time error.</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>The enable_if attribute takes two arguments, the first is an expression written
+in terms of the function parameters, the second is a string explaining why this
+overload candidate could not be selected to be displayed in diagnostics. The
+expression is part of the function signature for the purposes of determining
+whether it is a redeclaration (following the rules used when determining
+whether a C++ template specialization is ODR-equivalent), but is not part of
+the type.</p>
+<p>The enable_if expression is evaluated as if it were the body of a
+bool-returning constexpr function declared with the arguments of the function
+it is being applied to, then called with the parameters at the call site. If the
+result is false or could not be determined through constant expression
+evaluation, then this overload will not be chosen and the provided string may
+be used in a diagnostic if the compile fails as a result.</p>
+<p>Because the enable_if expression is an unevaluated context, there are no global
+state changes, nor the ability to pass information from the enable_if
+expression to the function body. For example, suppose we want calls to
+strnlen(strbuf, maxlen) to resolve to strnlen_chk(strbuf, maxlen, size of
+strbuf) only if the size of strbuf can be determined:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="n">__attribute__</span><span class="p">((</span><span class="n">always_inline</span><span class="p">))</span>
+<span class="k">static</span> <span class="kr">inline</span> <span class="kt">size_t</span> <span class="n">strnlen</span><span class="p">(</span><span class="k">const</span> <span class="kt">char</span> <span class="o">*</span><span class="n">s</span><span class="p">,</span> <span class="kt">size_t</span> <span class="n">maxlen</span><span class="p">)</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">))</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">__builtin_object_size</span><span class="p">(</span><span class="n">s</span><span class="p">,</span> <span class="mi">0</span><span class="p">)</span> <span class="o">!=</span> <span class="o">-</span><span class="mi">1</span><span class="p">))),</span>
+ <span class="s">"chosen when the buffer size is known but 'maxlen' is not"</span><span class="p">)))</span>
+<span class="p">{</span>
+ <span class="k">return</span> <span class="n">strnlen_chk</span><span class="p">(</span><span class="n">s</span><span class="p">,</span> <span class="n">maxlen</span><span class="p">,</span> <span class="n">__builtin_object_size</span><span class="p">(</span><span class="n">s</span><span class="p">,</span> <span class="mi">0</span><span class="p">));</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Multiple enable_if attributes may be applied to a single declaration. In this
+case, the enable_if expressions are evaluated from left to right in the
+following manner. First, the candidates whose enable_if expressions evaluate to
+false or cannot be evaluated are discarded. If the remaining candidates do not
+share ODR-equivalent enable_if expressions, the overload resolution is
+ambiguous. Otherwise, enable_if overload resolution continues with the next
+enable_if attribute on the candidates that have not been discarded and have
+remaining enable_if attributes. In this way, we pick the most specific
+overload out of a number of viable overloads using enable_if.</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">f</span><span class="p">()</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="nb">true</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span> <span class="c1">// #1</span>
+<span class="kt">void</span> <span class="nf">f</span><span class="p">()</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="nb">true</span><span class="p">,</span> <span class="s">""</span><span class="p">)))</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="nb">true</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span> <span class="c1">// #2</span>
+
+<span class="kt">void</span> <span class="nf">g</span><span class="p">(</span><span class="kt">int</span> <span class="n">i</span><span class="p">,</span> <span class="kt">int</span> <span class="n">j</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">i</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span> <span class="c1">// #1</span>
+<span class="kt">void</span> <span class="nf">g</span><span class="p">(</span><span class="kt">int</span> <span class="n">i</span><span class="p">,</span> <span class="kt">int</span> <span class="n">j</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">j</span><span class="p">,</span> <span class="s">""</span><span class="p">)))</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="nb">true</span><span class="p">)));</span> <span class="c1">// #2</span>
+</pre></div>
+</div>
+<p>In this example, a call to f() is always resolved to #2, as the first enable_if
+expression is ODR-equivalent for both declarations, but #1 does not have another
+enable_if expression to continue evaluating, so the next round of evaluation has
+only a single candidate. In a call to g(1, 1), the call is ambiguous even though
+#2 has more enable_if attributes, because the first enable_if expressions are
+not ODR-equivalent.</p>
+<p>Query for this feature with <code class="docutils literal notranslate"><span class="pre">__has_attribute(enable_if)</span></code>.</p>
+<p>Note that functions with one or more <code class="docutils literal notranslate"><span class="pre">enable_if</span></code> attributes may not have
+their address taken, unless all of the conditions specified by said
+<code class="docutils literal notranslate"><span class="pre">enable_if</span></code> are constants that evaluate to <code class="docutils literal notranslate"><span class="pre">true</span></code>. For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">const</span> <span class="kt">int</span> <span class="n">TrueConstant</span> <span class="o">=</span> <span class="mi">1</span><span class="p">;</span>
+<span class="k">const</span> <span class="kt">int</span> <span class="n">FalseConstant</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span>
+<span class="kt">int</span> <span class="nf">f</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">a</span> <span class="o">></span> <span class="mi">0</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span>
+<span class="kt">int</span> <span class="nf">g</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">a</span> <span class="o">==</span> <span class="mi">0</span> <span class="o">||</span> <span class="n">a</span> <span class="o">!=</span> <span class="mi">0</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span>
+<span class="kt">int</span> <span class="nf">h</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span>
+<span class="kt">int</span> <span class="nf">i</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">TrueConstant</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span>
+<span class="kt">int</span> <span class="nf">j</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">FalseConstant</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span>
+
+<span class="kt">void</span> <span class="nf">fn</span><span class="p">()</span> <span class="p">{</span>
+ <span class="kt">int</span> <span class="p">(</span><span class="o">*</span><span class="n">ptr</span><span class="p">)(</span><span class="kt">int</span><span class="p">);</span>
+ <span class="n">ptr</span> <span class="o">=</span> <span class="o">&</span><span class="n">f</span><span class="p">;</span> <span class="c1">// error: 'a > 0' is not always true</span>
+ <span class="n">ptr</span> <span class="o">=</span> <span class="o">&</span><span class="n">g</span><span class="p">;</span> <span class="c1">// error: 'a == 0 || a != 0' is not a truthy constant</span>
+ <span class="n">ptr</span> <span class="o">=</span> <span class="o">&</span><span class="n">h</span><span class="p">;</span> <span class="c1">// OK: 1 is a truthy constant</span>
+ <span class="n">ptr</span> <span class="o">=</span> <span class="o">&</span><span class="n">i</span><span class="p">;</span> <span class="c1">// OK: 'TrueConstant' is a truthy constant</span>
+ <span class="n">ptr</span> <span class="o">=</span> <span class="o">&</span><span class="n">j</span><span class="p">;</span> <span class="c1">// error: 'FalseConstant' is a constant, but not truthy</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Because <code class="docutils literal notranslate"><span class="pre">enable_if</span></code> evaluation happens during overload resolution,
+<code class="docutils literal notranslate"><span class="pre">enable_if</span></code> may give unintuitive results when used with templates, depending
+on when overloads are resolved. In the example below, clang will emit a
+diagnostic about no viable overloads for <code class="docutils literal notranslate"><span class="pre">foo</span></code> in <code class="docutils literal notranslate"><span class="pre">bar</span></code>, but not in <code class="docutils literal notranslate"><span class="pre">baz</span></code>:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="kt">double</span> <span class="nf">foo</span><span class="p">(</span><span class="kt">int</span> <span class="n">i</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">i</span> <span class="o">></span> <span class="mi">0</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span>
+<span class="kt">void</span> <span class="o">*</span><span class="nf">foo</span><span class="p">(</span><span class="kt">int</span> <span class="n">i</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">enable_if</span><span class="p">(</span><span class="n">i</span> <span class="o"><=</span> <span class="mi">0</span><span class="p">,</span> <span class="s">""</span><span class="p">)));</span>
+<span class="k">template</span> <span class="o"><</span><span class="kt">int</span> <span class="n">I</span><span class="o">></span>
+<span class="k">auto</span> <span class="n">bar</span><span class="p">()</span> <span class="p">{</span> <span class="k">return</span> <span class="n">foo</span><span class="p">(</span><span class="n">I</span><span class="p">);</span> <span class="p">}</span>
+
+<span class="k">template</span> <span class="o"><</span><span class="k">typename</span> <span class="n">T</span><span class="o">></span>
+<span class="k">auto</span> <span class="n">baz</span><span class="p">()</span> <span class="p">{</span> <span class="k">return</span> <span class="n">foo</span><span class="p">(</span><span class="n">T</span><span class="o">::</span><span class="n">number</span><span class="p">);</span> <span class="p">}</span>
+
+<span class="k">struct</span> <span class="n">WithNumber</span> <span class="p">{</span> <span class="k">constexpr</span> <span class="k">static</span> <span class="kt">int</span> <span class="n">number</span> <span class="o">=</span> <span class="mi">1</span><span class="p">;</span> <span class="p">};</span>
+<span class="kt">void</span> <span class="nf">callThem</span><span class="p">()</span> <span class="p">{</span>
+ <span class="n">bar</span><span class="o"><</span><span class="k">sizeof</span><span class="p">(</span><span class="n">WithNumber</span><span class="p">)</span><span class="o">></span><span class="p">();</span>
+ <span class="n">baz</span><span class="o"><</span><span class="n">WithNumber</span><span class="o">></span><span class="p">();</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>This is because, in <code class="docutils literal notranslate"><span class="pre">bar</span></code>, <code class="docutils literal notranslate"><span class="pre">foo</span></code> is resolved prior to template
+instantiation, so the value for <code class="docutils literal notranslate"><span class="pre">I</span></code> isnât known (thus, both <code class="docutils literal notranslate"><span class="pre">enable_if</span></code>
+conditions for <code class="docutils literal notranslate"><span class="pre">foo</span></code> fail). However, in <code class="docutils literal notranslate"><span class="pre">baz</span></code>, <code class="docutils literal notranslate"><span class="pre">foo</span></code> is resolved during
+template instantiation, so the value for <code class="docutils literal notranslate"><span class="pre">T::number</span></code> is known.</p>
+</div>
+<div class="section" id="external-source-symbol-clang-external-source-symbol-clang-external-source-symbol">
+<h3><a class="toc-backref" href="#id159">external_source_symbol (clang::external_source_symbol, clang::external_source_symbol)</a><a class="headerlink" href="#external-source-symbol-clang-external-source-symbol-clang-external-source-symbol" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id25">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id25" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">external_source_symbol</span></code> attribute specifies that a declaration originates
+from an external source and describes the nature of that source.</p>
+<p>The fact that Clang is capable of recognizing declarations that were defined
+externally can be used to provide better tooling support for mixed-language
+projects or projects that rely on auto-generated code. For instance, an IDE that
+uses Clang and that supports mixed-language projects can use this attribute to
+provide a correct âjump-to-definitionâ feature. For a concrete example,
+consider a protocol thatâs defined in a Swift file:</p>
+<div class="highlight-swift notranslate"><div class="highlight"><pre><span></span><span class="kr">@objc</span> <span class="kd">public</span> <span class="kd">protocol</span> <span class="nc">SwiftProtocol</span> <span class="p">{</span>
+ <span class="kd">func</span> <span class="nf">method</span><span class="p">()</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>This protocol can be used from Objective-C code by including a header file that
+was generated by the Swift compiler. The declarations in that header can use
+the <code class="docutils literal notranslate"><span class="pre">external_source_symbol</span></code> attribute to make Clang aware of the fact
+that <code class="docutils literal notranslate"><span class="pre">SwiftProtocol</span></code> actually originates from a Swift module:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="n">__attribute__</span><span class="p">((</span><span class="n">external_source_symbol</span><span class="p">(</span><span class="n">language</span><span class="o">=</span><span class="s">"Swift"</span><span class="p">,</span><span class="n">defined_in</span><span class="o">=</span><span class="s">"module"</span><span class="p">)))</span>
+<span class="k">@protocol</span> <span class="nc">SwiftProtocol</span>
+<span class="k">@required</span>
+<span class="o">-</span> <span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">method</span><span class="p">;</span>
+<span class="k">@end</span>
+</pre></div>
+</div>
+<p>Consequently, when âjump-to-definitionâ is performed at a location that
+references <code class="docutils literal notranslate"><span class="pre">SwiftProtocol</span></code>, the IDE can jump to the original definition in
+the Swift source file rather than jumping to the Objective-C declaration in the
+auto-generated header file.</p>
+<p>The <code class="docutils literal notranslate"><span class="pre">external_source_symbol</span></code> attribute is a comma-separated list that includes
+clauses that describe the origin and the nature of the particular declaration.
+Those clauses can be:</p>
+<dl class="docutils">
+<dt>language=<em>string-literal</em></dt>
+<dd>The name of the source language in which this declaration was defined.</dd>
+<dt>defined_in=<em>string-literal</em></dt>
+<dd>The name of the source container in which the declaration was defined. The
+exact definition of source container is language-specific, e.g. Swiftâs
+source containers are modules, so <code class="docutils literal notranslate"><span class="pre">defined_in</span></code> should specify the Swift
+module name.</dd>
+<dt>generated_declaration</dt>
+<dd>This declaration was automatically generated by some tool.</dd>
+</dl>
+<p>The clauses can be specified in any order. The clauses that are listed above are
+all optional, but the attribute has to have at least one clause.</p>
+</div>
+<div class="section" id="flatten-gnu-flatten">
+<h3><a class="toc-backref" href="#id160">flatten (gnu::flatten)</a><a class="headerlink" href="#flatten-gnu-flatten" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id26">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id26" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">flatten</span></code> attribute causes calls within the attributed function to
+be inlined unless it is impossible to do so, for example if the body of the
+callee is unavailable or if the callee has the <code class="docutils literal notranslate"><span class="pre">noinline</span></code> attribute.</p>
+</div>
+<div class="section" id="force-align-arg-pointer-gnu-force-align-arg-pointer">
+<h3><a class="toc-backref" href="#id161">force_align_arg_pointer (gnu::force_align_arg_pointer)</a><a class="headerlink" href="#force-align-arg-pointer-gnu-force-align-arg-pointer" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id27">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id27" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Use this attribute to force stack alignment.</p>
+<p>Legacy x86 code uses 4-byte stack alignment. Newer aligned SSE instructions
+(like âmovapsâ) that work with the stack require operands to be 16-byte aligned.
+This attribute realigns the stack in the function prologue to make sure the
+stack can be used with SSE instructions.</p>
+<p>Note that the x86_64 ABI forces 16-byte stack alignment at the call site.
+Because of this, âforce_align_arg_pointerâ is not needed on x86_64, except in
+rare cases where the caller does not align the stack properly (e.g. flow
+jumps from i386 arch code).</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="n">__attribute__</span> <span class="p">((</span><span class="n">force_align_arg_pointer</span><span class="p">))</span>
+<span class="kt">void</span> <span class="n">f</span> <span class="p">()</span> <span class="p">{</span>
+ <span class="p">...</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div></blockquote>
+</div>
+<div class="section" id="format-gnu-format">
+<h3><a class="toc-backref" href="#id162">format (gnu::format)</a><a class="headerlink" href="#format-gnu-format" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id28">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id28" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the <code class="docutils literal notranslate"><span class="pre">format</span></code> attribute, which indicates that the function
+accepts a <code class="docutils literal notranslate"><span class="pre">printf</span></code> or <code class="docutils literal notranslate"><span class="pre">scanf</span></code>-like format string and corresponding
+arguments or a <code class="docutils literal notranslate"><span class="pre">va_list</span></code> that contains these arguments.</p>
+<p>Please see <a class="reference external" href="http://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html">GCC documentation about format attribute</a> to find details
+about attribute syntax.</p>
+<p>Clang implements two kinds of checks with this attribute.</p>
+<ol class="arabic">
+<li><p class="first">Clang checks that the function with the <code class="docutils literal notranslate"><span class="pre">format</span></code> attribute is called with
+a format string that uses format specifiers that are allowed, and that
+arguments match the format string. This is the <code class="docutils literal notranslate"><span class="pre">-Wformat</span></code> warning, it is
+on by default.</p>
+</li>
+<li><p class="first">Clang checks that the format string argument is a literal string. This is
+the <code class="docutils literal notranslate"><span class="pre">-Wformat-nonliteral</span></code> warning, it is off by default.</p>
+<p>Clang implements this mostly the same way as GCC, but there is a difference
+for functions that accept a <code class="docutils literal notranslate"><span class="pre">va_list</span></code> argument (for example, <code class="docutils literal notranslate"><span class="pre">vprintf</span></code>).
+GCC does not emit <code class="docutils literal notranslate"><span class="pre">-Wformat-nonliteral</span></code> warning for calls to such
+functions. Clang does not warn if the format string comes from a function
+parameter, where the function is annotated with a compatible attribute,
+otherwise it warns. For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="n">__attribute__</span><span class="p">((</span><span class="n">__format__</span> <span class="p">(</span><span class="n">__scanf__</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">foo</span><span class="p">(</span><span class="k">const</span> <span class="kt">char</span><span class="o">*</span> <span class="n">s</span><span class="p">,</span> <span class="kt">char</span> <span class="o">*</span><span class="n">buf</span><span class="p">,</span> <span class="p">...)</span> <span class="p">{</span>
+ <span class="kt">va_list</span> <span class="n">ap</span><span class="p">;</span>
+ <span class="n">va_start</span><span class="p">(</span><span class="n">ap</span><span class="p">,</span> <span class="n">buf</span><span class="p">);</span>
+
+ <span class="n">vprintf</span><span class="p">(</span><span class="n">s</span><span class="p">,</span> <span class="n">ap</span><span class="p">);</span> <span class="c1">// warning: format string is not a string literal</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>In this case we warn because <code class="docutils literal notranslate"><span class="pre">s</span></code> contains a format string for a
+<code class="docutils literal notranslate"><span class="pre">scanf</span></code>-like function, but it is passed to a <code class="docutils literal notranslate"><span class="pre">printf</span></code>-like function.</p>
+<p>If the attribute is removed, clang still warns, because the format string is
+not a string literal.</p>
+<p>Another example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="n">__attribute__</span><span class="p">((</span><span class="n">__format__</span> <span class="p">(</span><span class="n">__printf__</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">foo</span><span class="p">(</span><span class="k">const</span> <span class="kt">char</span><span class="o">*</span> <span class="n">s</span><span class="p">,</span> <span class="kt">char</span> <span class="o">*</span><span class="n">buf</span><span class="p">,</span> <span class="p">...)</span> <span class="p">{</span>
+ <span class="kt">va_list</span> <span class="n">ap</span><span class="p">;</span>
+ <span class="n">va_start</span><span class="p">(</span><span class="n">ap</span><span class="p">,</span> <span class="n">buf</span><span class="p">);</span>
+
+ <span class="n">vprintf</span><span class="p">(</span><span class="n">s</span><span class="p">,</span> <span class="n">ap</span><span class="p">);</span> <span class="c1">// warning</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>In this case Clang does not warn because the format string <code class="docutils literal notranslate"><span class="pre">s</span></code> and
+the corresponding arguments are annotated. If the arguments are
+incorrect, the caller of <code class="docutils literal notranslate"><span class="pre">foo</span></code> will receive a warning.</p>
+</li>
+</ol>
+</div>
+<div class="section" id="ifunc-gnu-ifunc">
+<h3><a class="toc-backref" href="#id163">ifunc (gnu::ifunc)</a><a class="headerlink" href="#ifunc-gnu-ifunc" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id29">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id29" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p><code class="docutils literal notranslate"><span class="pre">__attribute__((ifunc("resolver")))</span></code> is used to mark that the address of a declaration should be resolved at runtime by calling a resolver function.</p>
+<p>The symbol name of the resolver function is given in quotes. A function with this name (after mangling) must be defined in the current translation unit; it may be <code class="docutils literal notranslate"><span class="pre">static</span></code>. The resolver function should take no arguments and return a pointer.</p>
+<p>The <code class="docutils literal notranslate"><span class="pre">ifunc</span></code> attribute may only be used on a function declaration. A function declaration with an <code class="docutils literal notranslate"><span class="pre">ifunc</span></code> attribute is considered to be a definition of the declared entity. The entity must not have weak linkage; for example, in C++, it cannot be applied to a declaration if a definition at that location would be considered inline.</p>
+<p>Not all targets support this attribute. ELF targets support this attribute when using binutils v2.20.1 or higher and glibc v2.11.1 or higher. Non-ELF targets currently do not support this attribute.</p>
+</div>
+<div class="section" id="internal-linkage-clang-internal-linkage-clang-internal-linkage">
+<h3><a class="toc-backref" href="#id164">internal_linkage (clang::internal_linkage, clang::internal_linkage)</a><a class="headerlink" href="#internal-linkage-clang-internal-linkage-clang-internal-linkage" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id30">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id30" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">internal_linkage</span></code> attribute changes the linkage type of the declaration to internal.
+This is similar to C-style <code class="docutils literal notranslate"><span class="pre">static</span></code>, but can be used on classes and class methods. When applied to a class definition,
+this attribute affects all methods and static data members of that class.
+This can be used to contain the ABI of a C++ library by excluding unwanted class methods from the export tables.</p>
+</div>
+<div class="section" id="interrupt-arm">
+<h3><a class="toc-backref" href="#id165">interrupt (ARM)</a><a class="headerlink" href="#interrupt-arm" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id31">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id31" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((interrupt("TYPE")))</span></code> attribute on
+ARM targets. This attribute may be attached to a function definition and
+instructs the backend to generate appropriate function entry/exit code so that
+it can be used directly as an interrupt service routine.</p>
+<p>The parameter passed to the interrupt attribute is optional, but if
+provided it must be a string literal with one of the following values: âIRQâ,
+âFIQâ, âSWIâ, âABORTâ, âUNDEFâ.</p>
+<p>The semantics are as follows:</p>
+<ul>
+<li><p class="first">If the function is AAPCS, Clang instructs the backend to realign the stack to
+8 bytes on entry. This is a general requirement of the AAPCS at public
+interfaces, but may not hold when an exception is taken. Doing this allows
+other AAPCS functions to be called.</p>
+</li>
+<li><p class="first">If the CPU is M-class this is all that needs to be done since the architecture
+itself is designed in such a way that functions obeying the normal AAPCS ABI
+constraints are valid exception handlers.</p>
+</li>
+<li><p class="first">If the CPU is not M-class, the prologue and epilogue are modified to save all
+non-banked registers that are used, so that upon return the user-mode state
+will not be corrupted. Note that to avoid unnecessary overhead, only
+general-purpose (integer) registers are saved in this way. If VFP operations
+are needed, that state must be saved manually.</p>
+<p>Specifically, interrupt kinds other than âFIQâ will save all core registers
+except âlrâ and âspâ. âFIQâ interrupts will save r0-r7.</p>
+</li>
+<li><p class="first">If the CPU is not M-class, the return instruction is changed to one of the
+canonical sequences permitted by the architecture for exception return. Where
+possible the function itself will make the necessary âlrâ adjustments so that
+the âpreferred return addressâ is selected.</p>
+<p>Unfortunately the compiler is unable to make this guarantee for an âUNDEFâ
+handler, where the offset from âlrâ to the preferred return address depends on
+the execution state of the code which generated the exception. In this case
+a sequence equivalent to âmovs pc, lrâ will be used.</p>
+</li>
+</ul>
+</div>
+<div class="section" id="interrupt-avr">
+<h3><a class="toc-backref" href="#id166">interrupt (AVR)</a><a class="headerlink" href="#interrupt-avr" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id32">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id32" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((interrupt))</span></code> attribute on
+AVR targets. This attribute may be attached to a function definition and instructs
+the backend to generate appropriate function entry/exit code so that it can be used
+directly as an interrupt service routine.</p>
+<p>On the AVR, the hardware globally disables interrupts when an interrupt is executed.
+The first instruction of an interrupt handler declared with this attribute is a SEI
+instruction to re-enable interrupts. See also the signal attribute that
+does not insert a SEI instruction.</p>
+</div>
+<div class="section" id="interrupt-mips">
+<h3><a class="toc-backref" href="#id167">interrupt (MIPS)</a><a class="headerlink" href="#interrupt-mips" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id33">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id33" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((interrupt("ARGUMENT")))</span></code> attribute on
+MIPS targets. This attribute may be attached to a function definition and instructs
+the backend to generate appropriate function entry/exit code so that it can be used
+directly as an interrupt service routine.</p>
+<p>By default, the compiler will produce a function prologue and epilogue suitable for
+an interrupt service routine that handles an External Interrupt Controller (eic)
+generated interrupt. This behaviour can be explicitly requested with the âeicâ
+argument.</p>
+<p>Otherwise, for use with vectored interrupt mode, the argument passed should be
+of the form âvector=LEVELâ where LEVEL is one of the following values:
+âsw0â, âsw1â, âhw0â, âhw1â, âhw2â, âhw3â, âhw4â, âhw5â. The compiler will
+then set the interrupt mask to the corresponding level which will mask all
+interrupts up to and including the argument.</p>
+<p>The semantics are as follows:</p>
+<ul class="simple">
+<li>The prologue is modified so that the Exception Program Counter (EPC) and
+Status coprocessor registers are saved to the stack. The interrupt mask is
+set so that the function can only be interrupted by a higher priority
+interrupt. The epilogue will restore the previous values of EPC and Status.</li>
+<li>The prologue and epilogue are modified to save and restore all non-kernel
+registers as necessary.</li>
+<li>The FPU is disabled in the prologue, as the floating pointer registers are not
+spilled to the stack.</li>
+<li>The function return sequence is changed to use an exception return instruction.</li>
+<li>The parameter sets the interrupt mask for the function corresponding to the
+interrupt level specified. If no mask is specified the interrupt mask
+defaults to âeicâ.</li>
+</ul>
+</div>
+<div class="section" id="interrupt-riscv">
+<h3><a class="toc-backref" href="#id168">interrupt (RISCV)</a><a class="headerlink" href="#interrupt-riscv" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id34">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id34" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((interrupt))</span></code> attribute on RISCV
+targets. This attribute may be attached to a function definition and instructs
+the backend to generate appropriate function entry/exit code so that it can be
+used directly as an interrupt service routine.</p>
+<p>Permissible values for this parameter are <code class="docutils literal notranslate"><span class="pre">user</span></code>, <code class="docutils literal notranslate"><span class="pre">supervisor</span></code>,
+and <code class="docutils literal notranslate"><span class="pre">machine</span></code>. If there is no parameter, then it defaults to machine.</p>
+<p>Repeated interrupt attribute on the same declaration will cause a warning
+to be emitted. In case of repeated declarations, the last one prevails.</p>
+<p>Refer to:
+<a class="reference external" href="https://gcc.gnu.org/onlinedocs/gcc/RISC-V-Function-Attributes.html">https://gcc.gnu.org/onlinedocs/gcc/RISC-V-Function-Attributes.html</a>
+<a class="reference external" href="https://riscv.org/specifications/privileged-isa/">https://riscv.org/specifications/privileged-isa/</a>
+The RISC-V Instruction Set Manual Volume II: Privileged Architecture
+Version 1.10.</p>
+</div>
+<div class="section" id="kernel">
+<h3><a class="toc-backref" href="#id169">kernel</a><a class="headerlink" href="#kernel" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id35">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id35" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p><code class="docutils literal notranslate"><span class="pre">__attribute__((kernel))</span></code> is used to mark a <code class="docutils literal notranslate"><span class="pre">kernel</span></code> function in
+RenderScript.</p>
+<p>In RenderScript, <code class="docutils literal notranslate"><span class="pre">kernel</span></code> functions are used to express data-parallel
+computations. The RenderScript runtime efficiently parallelizes <code class="docutils literal notranslate"><span class="pre">kernel</span></code>
+functions to run on computational resources such as multi-core CPUs and GPUs.
+See the <a class="reference external" href="https://developer.android.com/guide/topics/renderscript/compute.html">RenderScript</a> documentation for more information.</p>
+</div>
+<div class="section" id="lifetimebound-clang-lifetimebound">
+<h3><a class="toc-backref" href="#id170">lifetimebound (clang::lifetimebound)</a><a class="headerlink" href="#lifetimebound-clang-lifetimebound" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id36">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id36" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">lifetimebound</span></code> attribute indicates that a resource owned by
+a function parameter or implicit object parameter
+is retained by the return value of the annotated function
+(or, for a parameter of a constructor, in the value of the constructed object).
+It is only supported in C++.</p>
+<p>This attribute provides an experimental implementation of the facility
+described in the C++ committee paper [<a class="reference external" href="http://wg21.link/p0936r0](P0936R0">http://wg21.link/p0936r0](P0936R0</a>),
+and is subject to change as the design of the corresponding functionality
+changes.</p>
+</div>
+<div class="section" id="long-call-gnu-long-call-gnu-far">
+<h3><a class="toc-backref" href="#id171">long_call (gnu::long_call, gnu::far)</a><a class="headerlink" href="#long-call-gnu-long-call-gnu-far" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id37">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id37" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the <code class="docutils literal notranslate"><span class="pre">__attribute__((long_call))</span></code>, <code class="docutils literal notranslate"><span class="pre">__attribute__((far))</span></code>,
+and <code class="docutils literal notranslate"><span class="pre">__attribute__((near))</span></code> attributes on MIPS targets. These attributes may
+only be added to function declarations and change the code generated
+by the compiler when directly calling the function. The <code class="docutils literal notranslate"><span class="pre">near</span></code> attribute
+allows calls to the function to be made using the <code class="docutils literal notranslate"><span class="pre">jal</span></code> instruction, which
+requires the function to be located in the same naturally aligned 256MB
+segment as the caller. The <code class="docutils literal notranslate"><span class="pre">long_call</span></code> and <code class="docutils literal notranslate"><span class="pre">far</span></code> attributes are synonyms
+and require the use of a different call sequence that works regardless
+of the distance between the functions.</p>
+<p>These attributes have no effect for position-independent code.</p>
+<p>These attributes take priority over command line switches such
+as <code class="docutils literal notranslate"><span class="pre">-mlong-calls</span></code> and <code class="docutils literal notranslate"><span class="pre">-mno-long-calls</span></code>.</p>
+</div>
+<div class="section" id="micromips-gnu-micromips">
+<h3><a class="toc-backref" href="#id172">micromips (gnu::micromips)</a><a class="headerlink" href="#micromips-gnu-micromips" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id38">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id38" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((micromips))</span></code> and
+<code class="docutils literal notranslate"><span class="pre">__attribute__((nomicromips))</span></code> attributes on MIPS targets. These attributes
+may be attached to a function definition and instructs the backend to generate
+or not to generate microMIPS code for that function.</p>
+<p>These attributes override the <cite>-mmicromips</cite> and <cite>-mno-micromips</cite> options
+on the command line.</p>
+</div>
+<div class="section" id="min-vector-width-clang-min-vector-width-clang-min-vector-width">
+<h3><a class="toc-backref" href="#id173">min_vector_width (clang::min_vector_width, clang::min_vector_width)</a><a class="headerlink" href="#min-vector-width-clang-min-vector-width-clang-min-vector-width" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id39">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id39" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the <code class="docutils literal notranslate"><span class="pre">__attribute__((min_vector_width(width)))</span></code> attribute. This
+attribute may be attached to a function and informs the backend that this
+function desires vectors of at least this width to be generated. Target-specific
+maximum vector widths still apply. This means even if you ask for something
+larger than the target supports, you will only get what the target supports.
+This attribute is meant to be a hint to control target heuristics that may
+generate narrower vectors than what the target hardware supports.</p>
+<p>This is currently used by the X86 target to allow some CPUs that support 512-bit
+vectors to be limited to using 256-bit vectors to avoid frequency penalties.
+This is currently enabled with the <code class="docutils literal notranslate"><span class="pre">-prefer-vector-width=256</span></code> command line
+option. The <code class="docutils literal notranslate"><span class="pre">min_vector_width</span></code> attribute can be used to prevent the backend
+from trying to split vector operations to match the <code class="docutils literal notranslate"><span class="pre">prefer-vector-width</span></code>. All
+X86 vector intrinsics from x86intrin.h already set this attribute. Additionally,
+use of any of the X86-specific vector builtins will implicitly set this
+attribute on the calling function. The intent is that explicitly writing vector
+code using the X86 intrinsics will prevent <code class="docutils literal notranslate"><span class="pre">prefer-vector-width</span></code> from
+affecting the code.</p>
+</div>
+<div class="section" id="no-caller-saved-registers-gnu-no-caller-saved-registers">
+<h3><a class="toc-backref" href="#id174">no_caller_saved_registers (gnu::no_caller_saved_registers)</a><a class="headerlink" href="#no-caller-saved-registers-gnu-no-caller-saved-registers" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id40">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id40" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Use this attribute to indicate that the specified function has no
+caller-saved registers. That is, all registers are callee-saved except for
+registers used for passing parameters to the function or returning parameters
+from the function.
+The compiler saves and restores any modified registers that were not used for
+passing or returning arguments to the function.</p>
+<p>The user can call functions specified with the âno_caller_saved_registersâ
+attribute from an interrupt handler without saving and restoring all
+call-clobbered registers.</p>
+<p>Note that âno_caller_saved_registersâ attribute is not a calling convention.
+In fact, it only overrides the decision of which registers should be saved by
+the caller, but not how the parameters are passed from the caller to the callee.</p>
+<p>For example:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="n">__attribute__</span> <span class="p">((</span><span class="n">no_caller_saved_registers</span><span class="p">,</span> <span class="n">fastcall</span><span class="p">))</span>
+<span class="kt">void</span> <span class="n">f</span> <span class="p">(</span><span class="kt">int</span> <span class="n">arg1</span><span class="p">,</span> <span class="kt">int</span> <span class="n">arg2</span><span class="p">)</span> <span class="p">{</span>
+ <span class="p">...</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>In this case parameters âarg1â and âarg2â will be passed in registers.
+In this case, on 32-bit x86 targets, the function âfâ will use ECX and EDX as
+register parameters. However, it will not assume any scratch registers and
+should save and restore any modified registers except for ECX and EDX.</p>
+</div></blockquote>
+</div>
+<div class="section" id="no-sanitize-clang-no-sanitize-clang-no-sanitize">
+<h3><a class="toc-backref" href="#id175">no_sanitize (clang::no_sanitize, clang::no_sanitize)</a><a class="headerlink" href="#no-sanitize-clang-no-sanitize-clang-no-sanitize" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id41">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id41" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Use the <code class="docutils literal notranslate"><span class="pre">no_sanitize</span></code> attribute on a function or a global variable
+declaration to specify that a particular instrumentation or set of
+instrumentations should not be applied. The attribute takes a list of
+string literals, which have the same meaning as values accepted by the
+<code class="docutils literal notranslate"><span class="pre">-fno-sanitize=</span></code> flag. For example,
+<code class="docutils literal notranslate"><span class="pre">__attribute__((no_sanitize("address",</span> <span class="pre">"thread")))</span></code> specifies that
+AddressSanitizer and ThreadSanitizer should not be applied to the
+function or variable.</p>
+<p>See <a class="reference internal" href="UsersManual.html#controlling-code-generation"><span class="std std-ref">Controlling Code Generation</span></a> for a
+full list of supported sanitizer flags.</p>
+</div>
+<div class="section" id="no-sanitize-address-no-address-safety-analysis-gnu-no-address-safety-analysis-gnu-no-sanitize-address">
+<h3><a class="toc-backref" href="#id176">no_sanitize_address (no_address_safety_analysis, gnu::no_address_safety_analysis, gnu::no_sanitize_address)</a><a class="headerlink" href="#no-sanitize-address-no-address-safety-analysis-gnu-no-address-safety-analysis-gnu-no-sanitize-address" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id42">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id42" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p id="langext-address-sanitizer">Use <code class="docutils literal notranslate"><span class="pre">__attribute__((no_sanitize_address))</span></code> on a function or a global
+variable declaration to specify that address safety instrumentation
+(e.g. AddressSanitizer) should not be applied.</p>
+</div>
+<div class="section" id="no-sanitize-memory">
+<h3><a class="toc-backref" href="#id177">no_sanitize_memory</a><a class="headerlink" href="#no-sanitize-memory" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id43">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id43" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p id="langext-memory-sanitizer">Use <code class="docutils literal notranslate"><span class="pre">__attribute__((no_sanitize_memory))</span></code> on a function declaration to
+specify that checks for uninitialized memory should not be inserted
+(e.g. by MemorySanitizer). The function may still be instrumented by the tool
+to avoid false positives in other places.</p>
+</div>
+<div class="section" id="no-sanitize-thread">
+<h3><a class="toc-backref" href="#id178">no_sanitize_thread</a><a class="headerlink" href="#no-sanitize-thread" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id44">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id44" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p id="langext-thread-sanitizer">Use <code class="docutils literal notranslate"><span class="pre">__attribute__((no_sanitize_thread))</span></code> on a function declaration to
+specify that checks for data races on plain (non-atomic) memory accesses should
+not be inserted by ThreadSanitizer. The function is still instrumented by the
+tool to avoid false positives and provide meaningful stack traces.</p>
+</div>
+<div class="section" id="no-split-stack-gnu-no-split-stack">
+<h3><a class="toc-backref" href="#id179">no_split_stack (gnu::no_split_stack)</a><a class="headerlink" href="#no-split-stack-gnu-no-split-stack" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id45">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id45" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">no_split_stack</span></code> attribute disables the emission of the split stack
+preamble for a particular function. It has no effect if <code class="docutils literal notranslate"><span class="pre">-fsplit-stack</span></code>
+is not specified.</p>
+</div>
+<div class="section" id="no-stack-protector-clang-no-stack-protector-clang-no-stack-protector">
+<h3><a class="toc-backref" href="#id180">no_stack_protector (clang::no_stack_protector, clang::no_stack_protector)</a><a class="headerlink" href="#no-stack-protector-clang-no-stack-protector-clang-no-stack-protector" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id46">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id46" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the <code class="docutils literal notranslate"><span class="pre">__attribute__((no_stack_protector))</span></code> attribute which disables
+the stack protector on the specified function. This attribute is useful for
+selectively disabling the stack protector on some functions when building with
+<code class="docutils literal notranslate"><span class="pre">-fstack-protector</span></code> compiler option.</p>
+<p>For example, it disables the stack protector for the function <code class="docutils literal notranslate"><span class="pre">foo</span></code> but function
+<code class="docutils literal notranslate"><span class="pre">bar</span></code> will still be built with the stack protector with the <code class="docutils literal notranslate"><span class="pre">-fstack-protector</span></code>
+option.</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">no_stack_protector</span><span class="p">))</span>
+<span class="n">foo</span> <span class="p">(</span><span class="kt">int</span> <span class="n">x</span><span class="p">);</span> <span class="c1">// stack protection will be disabled for foo.</span>
+
+<span class="kt">int</span> <span class="nf">bar</span><span class="p">(</span><span class="kt">int</span> <span class="n">y</span><span class="p">);</span> <span class="c1">// bar can be built with the stack protector.</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="noalias">
+<h3><a class="toc-backref" href="#id181">noalias</a><a class="headerlink" href="#noalias" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id47">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id47" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">noalias</span></code> 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.</p>
+</div>
+<div class="section" id="nocf-check-gnu-nocf-check">
+<h3><a class="toc-backref" href="#id182">nocf_check (gnu::nocf_check)</a><a class="headerlink" href="#nocf-check-gnu-nocf-check" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id48">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id48" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Jump Oriented Programming attacks rely on tampering with addresses used by
+indirect call / jmp, e.g. redirect control-flow to non-programmer
+intended bytes in the binary.
+X86 Supports Indirect Branch Tracking (IBT) as part of Control-Flow
+Enforcement Technology (CET). IBT instruments ENDBR instructions used to
+specify valid targets of indirect call / jmp.
+The <code class="docutils literal notranslate"><span class="pre">nocf_check</span></code> attribute has two roles:
+1. Appertains to a function - do not add ENDBR instruction at the beginning of
+the function.
+2. Appertains to a function pointer - do not track the target function of this
+pointer (by adding nocf_check prefix to the indirect-call instruction).</p>
+</div>
+<div class="section" id="nodiscard-warn-unused-result-clang-warn-unused-result-gnu-warn-unused-result">
+<h3><a class="toc-backref" href="#id183">nodiscard, warn_unused_result, clang::warn_unused_result, gnu::warn_unused_result</a><a class="headerlink" href="#nodiscard-warn-unused-result-clang-warn-unused-result-gnu-warn-unused-result" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id49">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id49" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the ability to diagnose when the results of a function call
+expression are discarded under suspicious circumstances. A diagnostic is
+generated when a function or its return type is marked with <code class="docutils literal notranslate"><span class="pre">[[nodiscard]]</span></code>
+(or <code class="docutils literal notranslate"><span class="pre">__attribute__((warn_unused_result))</span></code>) and the function call appears as a
+potentially-evaluated discarded-value expression that is not explicitly cast to
+<cite>void</cite>.</p>
+</div>
+<div class="section" id="noduplicate-clang-noduplicate-clang-noduplicate">
+<h3><a class="toc-backref" href="#id184">noduplicate (clang::noduplicate, clang::noduplicate)</a><a class="headerlink" href="#noduplicate-clang-noduplicate-clang-noduplicate" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id50">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id50" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">noduplicate</span></code> attribute can be placed on function declarations to control
+whether function calls to this function can be duplicated or not as a result of
+optimizations. This is required for the implementation of functions with
+certain special requirements, like the OpenCL âbarrierâ function, that might
+need to be run concurrently by all the threads that are executing in lockstep
+on the hardware. For example this attribute applied on the function
+ânodupfuncâ in the code below avoids that:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">nodupfunc</span><span class="p">()</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">noduplicate</span><span class="p">));</span>
+<span class="c1">// Setting it as a C++11 attribute is also valid</span>
+<span class="c1">// void nodupfunc() [[clang::noduplicate]];</span>
+<span class="kt">void</span> <span class="nf">foo</span><span class="p">();</span>
+<span class="kt">void</span> <span class="nf">bar</span><span class="p">();</span>
+
+<span class="n">nodupfunc</span><span class="p">();</span>
+<span class="k">if</span> <span class="p">(</span><span class="n">a</span> <span class="o">></span> <span class="n">n</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">foo</span><span class="p">();</span>
+<span class="p">}</span> <span class="k">else</span> <span class="p">{</span>
+ <span class="n">bar</span><span class="p">();</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>gets possibly modified by some optimizations into code similar to this:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">if</span> <span class="p">(</span><span class="n">a</span> <span class="o">></span> <span class="n">n</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">nodupfunc</span><span class="p">();</span>
+ <span class="n">foo</span><span class="p">();</span>
+<span class="p">}</span> <span class="k">else</span> <span class="p">{</span>
+ <span class="n">nodupfunc</span><span class="p">();</span>
+ <span class="n">bar</span><span class="p">();</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>where the call to ânodupfuncâ is duplicated and sunk into the two branches
+of the condition.</p>
+</div>
+<div class="section" id="nomicromips-gnu-nomicromips">
+<h3><a class="toc-backref" href="#id185">nomicromips (gnu::nomicromips)</a><a class="headerlink" href="#nomicromips-gnu-nomicromips" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id51">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id51" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((micromips))</span></code> and
+<code class="docutils literal notranslate"><span class="pre">__attribute__((nomicromips))</span></code> attributes on MIPS targets. These attributes
+may be attached to a function definition and instructs the backend to generate
+or not to generate microMIPS code for that function.</p>
+<p>These attributes override the <cite>-mmicromips</cite> and <cite>-mno-micromips</cite> options
+on the command line.</p>
+</div>
+<div class="section" id="id1">
+<h3><a class="toc-backref" href="#id186">noreturn</a><a class="headerlink" href="#id1" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id52">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id52" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>A function declared as <code class="docutils literal notranslate"><span class="pre">[[noreturn]]</span></code> shall not return to its caller. The
+compiler will generate a diagnostic for a function declared as <code class="docutils literal notranslate"><span class="pre">[[noreturn]]</span></code>
+that appears to be capable of returning to its caller.</p>
+</div>
+<div class="section" id="not-tail-called-clang-not-tail-called-clang-not-tail-called">
+<h3><a class="toc-backref" href="#id187">not_tail_called (clang::not_tail_called, clang::not_tail_called)</a><a class="headerlink" href="#not-tail-called-clang-not-tail-called-clang-not-tail-called" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id53">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id53" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">not_tail_called</span></code> attribute prevents tail-call optimization on statically bound calls. It has no effect on indirect calls. Virtual functions, objective-c methods, and functions marked as <code class="docutils literal notranslate"><span class="pre">always_inline</span></code> cannot be marked as <code class="docutils literal notranslate"><span class="pre">not_tail_called</span></code>.</p>
+<p>For example, it prevents tail-call optimization in the following case:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">not_tail_called</span><span class="p">))</span> <span class="n">foo1</span><span class="p">(</span><span class="kt">int</span><span class="p">);</span>
+
+<span class="kt">int</span> <span class="nf">foo2</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="n">foo1</span><span class="p">(</span><span class="n">a</span><span class="p">);</span> <span class="c1">// No tail-call optimization on direct calls.</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>However, it doesnât prevent tail-call optimization in this case:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">not_tail_called</span><span class="p">))</span> <span class="n">foo1</span><span class="p">(</span><span class="kt">int</span><span class="p">);</span>
+
+<span class="kt">int</span> <span class="nf">foo2</span><span class="p">(</span><span class="kt">int</span> <span class="n">a</span><span class="p">)</span> <span class="p">{</span>
+ <span class="kt">int</span> <span class="p">(</span><span class="o">*</span><span class="n">fn</span><span class="p">)(</span><span class="kt">int</span><span class="p">)</span> <span class="o">=</span> <span class="o">&</span><span class="n">foo1</span><span class="p">;</span>
+
+ <span class="c1">// not_tail_called has no effect on an indirect call even if the call can be</span>
+ <span class="c1">// resolved at compile time.</span>
+ <span class="k">return</span> <span class="p">(</span><span class="o">*</span><span class="n">fn</span><span class="p">)(</span><span class="n">a</span><span class="p">);</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>Marking virtual functions as <code class="docutils literal notranslate"><span class="pre">not_tail_called</span></code> is an error:</p>
+<blockquote>
+<div><div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">Base</span> <span class="p">{</span>
+<span class="k">public</span><span class="o">:</span>
+ <span class="c1">// not_tail_called on a virtual function is an error.</span>
+ <span class="p">[[</span><span class="n">clang</span><span class="o">::</span><span class="n">not_tail_called</span><span class="p">]]</span> <span class="k">virtual</span> <span class="kt">int</span> <span class="n">foo1</span><span class="p">();</span>
+
+ <span class="k">virtual</span> <span class="kt">int</span> <span class="nf">foo2</span><span class="p">();</span>
+
+ <span class="c1">// Non-virtual functions can be marked ``not_tail_called``.</span>
+ <span class="p">[[</span><span class="n">clang</span><span class="o">::</span><span class="n">not_tail_called</span><span class="p">]]</span> <span class="kt">int</span> <span class="n">foo3</span><span class="p">();</span>
+<span class="p">};</span>
+
+<span class="k">class</span> <span class="nc">Derived1</span> <span class="o">:</span> <span class="k">public</span> <span class="n">Base</span> <span class="p">{</span>
+<span class="k">public</span><span class="o">:</span>
+ <span class="kt">int</span> <span class="n">foo1</span><span class="p">()</span> <span class="k">override</span><span class="p">;</span>
+
+ <span class="c1">// not_tail_called on a virtual function is an error.</span>
+ <span class="p">[[</span><span class="n">clang</span><span class="o">::</span><span class="n">not_tail_called</span><span class="p">]]</span> <span class="kt">int</span> <span class="n">foo2</span><span class="p">()</span> <span class="k">override</span><span class="p">;</span>
+<span class="p">};</span>
+</pre></div>
+</div>
+</div></blockquote>
+</div>
+<div class="section" id="nothrow-gnu-nothrow">
+<h3><a class="toc-backref" href="#id188">nothrow (gnu::nothrow)</a><a class="headerlink" href="#nothrow-gnu-nothrow" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id54">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id54" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((nothrow))</span></code> and Microsoft style
+<code class="docutils literal notranslate"><span class="pre">__declspec(nothrow)</span></code> attribute as an equivalent of <cite>noexcept</cite> on function
+declarations. This attribute informs the compiler that the annotated function
+does not throw an exception. This prevents exception-unwinding. This attribute
+is particularly useful on functions in the C Standard Library that are
+guaranteed to not throw an exception.</p>
+</div>
+<div class="section" id="objc-boxable-clang-objc-boxable-clang-objc-boxable">
+<h3><a class="toc-backref" href="#id189">objc_boxable (clang::objc_boxable, clang::objc_boxable)</a><a class="headerlink" href="#objc-boxable-clang-objc-boxable-clang-objc-boxable" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id55">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id55" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Structs and unions marked with the <code class="docutils literal notranslate"><span class="pre">objc_boxable</span></code> attribute can be used
+with the Objective-C boxed expression syntax, <code class="docutils literal notranslate"><span class="pre">@(...)</span></code>.</p>
+<p><strong>Usage</strong>: <code class="docutils literal notranslate"><span class="pre">__attribute__((objc_boxable))</span></code>. This attribute
+can only be placed on a declaration of a trivially-copyable struct or union:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="k">struct</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">objc_boxable</span><span class="p">))</span> <span class="n">some_struct</span> <span class="p">{</span>
+ <span class="kt">int</span> <span class="n">i</span><span class="p">;</span>
+<span class="p">};</span>
+<span class="k">union</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">objc_boxable</span><span class="p">))</span> <span class="n">some_union</span> <span class="p">{</span>
+ <span class="kt">int</span> <span class="n">i</span><span class="p">;</span>
+ <span class="kt">float</span> <span class="n">f</span><span class="p">;</span>
+<span class="p">};</span>
+<span class="k">typedef</span> <span class="k">struct</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">objc_boxable</span><span class="p">))</span> <span class="n">_some_struct</span> <span class="n">some_struct</span><span class="p">;</span>
+
+<span class="c1">// ...</span>
+
+<span class="n">some_struct</span> <span class="n">ss</span><span class="p">;</span>
+<span class="bp">NSValue</span> <span class="o">*</span><span class="n">boxed</span> <span class="o">=</span> <span class="l">@(</span><span class="n">ss</span><span class="l">)</span><span class="p">;</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="objc-method-family-clang-objc-method-family-clang-objc-method-family">
+<h3><a class="toc-backref" href="#id190">objc_method_family (clang::objc_method_family, clang::objc_method_family)</a><a class="headerlink" href="#objc-method-family-clang-objc-method-family-clang-objc-method-family" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id56">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id56" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Many methods in Objective-C have conventional meanings determined by their
+selectors. It is sometimes useful to be able to mark a method as having a
+particular conventional meaning despite not having the right selector, or as
+not having the conventional meaning that its selector would suggest. For these
+use cases, we provide an attribute to specifically describe the âmethod familyâ
+that a method belongs to.</p>
+<p><strong>Usage</strong>: <code class="docutils literal notranslate"><span class="pre">__attribute__((objc_method_family(X)))</span></code>, where <code class="docutils literal notranslate"><span class="pre">X</span></code> is one of
+<code class="docutils literal notranslate"><span class="pre">none</span></code>, <code class="docutils literal notranslate"><span class="pre">alloc</span></code>, <code class="docutils literal notranslate"><span class="pre">copy</span></code>, <code class="docutils literal notranslate"><span class="pre">init</span></code>, <code class="docutils literal notranslate"><span class="pre">mutableCopy</span></code>, or <code class="docutils literal notranslate"><span class="pre">new</span></code>. This
+attribute can only be placed at the end of a method declaration:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="p">-</span> <span class="p">(</span><span class="bp">NSString</span> <span class="o">*</span><span class="p">)</span><span class="nf">initMyStringValue</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">objc_method_family</span><span class="p">(</span><span class="n">none</span><span class="p">)));</span>
+</pre></div>
+</div>
+<p>Users who do not wish to change the conventional meaning of a method, and who
+merely want to document its non-standard retain and release semantics, should
+use the retaining behavior attributes (<code class="docutils literal notranslate"><span class="pre">ns_returns_retained</span></code>,
+<code class="docutils literal notranslate"><span class="pre">ns_returns_not_retained</span></code>, etc).</p>
+<p>Query for this feature with <code class="docutils literal notranslate"><span class="pre">__has_attribute(objc_method_family)</span></code>.</p>
+</div>
+<div class="section" id="objc-requires-super-clang-objc-requires-super-clang-objc-requires-super">
+<h3><a class="toc-backref" href="#id191">objc_requires_super (clang::objc_requires_super, clang::objc_requires_super)</a><a class="headerlink" href="#objc-requires-super-clang-objc-requires-super-clang-objc-requires-super" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id57">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id57" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Some Objective-C classes allow a subclass to override a particular method in a
+parent class but expect that the overriding method also calls the overridden
+method in the parent class. For these cases, we provide an attribute to
+designate that a method requires a âcall to <code class="docutils literal notranslate"><span class="pre">super</span></code>â in the overriding
+method in the subclass.</p>
+<p><strong>Usage</strong>: <code class="docutils literal notranslate"><span class="pre">__attribute__((objc_requires_super))</span></code>. This attribute can only
+be placed at the end of a method declaration:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="p">-</span> <span class="p">(</span><span class="kt">void</span><span class="p">)</span><span class="nf">foo</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">objc_requires_super</span><span class="p">));</span>
+</pre></div>
+</div>
+<p>This attribute can only be applied the method declarations within a class, and
+not a protocol. Currently this attribute does not enforce any placement of
+where the call occurs in the overriding method (such as in the case of
+<code class="docutils literal notranslate"><span class="pre">-dealloc</span></code> where the call must appear at the end). It checks only that it
+exists.</p>
+<p>Note that on both OS X and iOS that the Foundation framework provides a
+convenience macro <code class="docutils literal notranslate"><span class="pre">NS_REQUIRES_SUPER</span></code> that provides syntactic sugar for this
+attribute:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="p">-</span> <span class="p">(</span><span class="kt">void</span><span class="p">)</span><span class="nf">foo</span> <span class="n">NS_REQUIRES_SUPER</span><span class="p">;</span>
+</pre></div>
+</div>
+<p>This macro is conditionally defined depending on the compilerâs support for
+this attribute. If the compiler does not support the attribute the macro
+expands to nothing.</p>
+<p>Operationally, when a method has this annotation the compiler will warn if the
+implementation of an override in a subclass does not call super. For example:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="nl">warning</span><span class="p">:</span> <span class="n">method</span> <span class="n">possibly</span> <span class="n">missing</span> <span class="n">a</span> <span class="p">[</span><span class="nb">super</span> <span class="n">AnnotMeth</span><span class="p">]</span> <span class="n">call</span>
+<span class="o">-</span> <span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">AnnotMeth</span><span class="p">{};</span>
+ <span class="o">^</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="objc-runtime-name-clang-objc-runtime-name-clang-objc-runtime-name">
+<h3><a class="toc-backref" href="#id192">objc_runtime_name (clang::objc_runtime_name, clang::objc_runtime_name)</a><a class="headerlink" href="#objc-runtime-name-clang-objc-runtime-name-clang-objc-runtime-name" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id58">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id58" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>By default, the Objective-C interface or protocol identifier is used
+in the metadata name for that object. The <cite>objc_runtime_name</cite>
+attribute allows annotated interfaces or protocols to use the
+specified string argument in the objectâs metadata name instead of the
+default name.</p>
+<p><strong>Usage</strong>: <code class="docutils literal notranslate"><span class="pre">__attribute__((objc_runtime_name("MyLocalName")))</span></code>. This attribute
+can only be placed before an @protocol or @interface declaration:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="n">__attribute__</span><span class="p">((</span><span class="n">objc_runtime_name</span><span class="p">(</span><span class="s">"MyLocalName"</span><span class="p">)))</span>
+<span class="k">@interface</span> <span class="nc">Message</span>
+<span class="k">@end</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="objc-runtime-visible-clang-objc-runtime-visible-clang-objc-runtime-visible">
+<h3><a class="toc-backref" href="#id193">objc_runtime_visible (clang::objc_runtime_visible, clang::objc_runtime_visible)</a><a class="headerlink" href="#objc-runtime-visible-clang-objc-runtime-visible-clang-objc-runtime-visible" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id59">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id59" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>This attribute specifies that the Objective-C class to which it applies is visible to the Objective-C runtime but not to the linker. Classes annotated with this attribute cannot be subclassed and cannot have categories defined for them.</p>
+</div>
+<div class="section" id="optnone-clang-optnone-clang-optnone">
+<h3><a class="toc-backref" href="#id194">optnone (clang::optnone, clang::optnone)</a><a class="headerlink" href="#optnone-clang-optnone-clang-optnone" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id60">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id60" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">optnone</span></code> attribute suppresses essentially all optimizations
+on a function or method, regardless of the optimization level applied to
+the compilation unit as a whole. This is particularly useful when you
+need to debug a particular function, but it is infeasible to build the
+entire application without optimization. Avoiding optimization on the
+specified function can improve the quality of the debugging information
+for that function.</p>
+<p>This attribute is incompatible with the <code class="docutils literal notranslate"><span class="pre">always_inline</span></code> and <code class="docutils literal notranslate"><span class="pre">minsize</span></code>
+attributes.</p>
+</div>
+<div class="section" id="overloadable-clang-overloadable-clang-overloadable">
+<h3><a class="toc-backref" href="#id195">overloadable (clang::overloadable, clang::overloadable)</a><a class="headerlink" href="#overloadable-clang-overloadable-clang-overloadable" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id61">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id61" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang provides support for C++ function overloading in C. Function overloading
+in C is introduced using the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute. For example, one
+might provide several overloaded versions of a <code class="docutils literal notranslate"><span class="pre">tgsin</span></code> function that invokes
+the appropriate standard function computing the sine of a value with <code class="docutils literal notranslate"><span class="pre">float</span></code>,
+<code class="docutils literal notranslate"><span class="pre">double</span></code>, or <code class="docutils literal notranslate"><span class="pre">long</span> <span class="pre">double</span></code> precision:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="cp">#include</span> <span class="cpf"><math.h></span><span class="cp"></span>
+<span class="kt">float</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">))</span> <span class="n">tgsin</span><span class="p">(</span><span class="kt">float</span> <span class="n">x</span><span class="p">)</span> <span class="p">{</span> <span class="k">return</span> <span class="n">sinf</span><span class="p">(</span><span class="n">x</span><span class="p">);</span> <span class="p">}</span>
+<span class="kt">double</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">))</span> <span class="n">tgsin</span><span class="p">(</span><span class="kt">double</span> <span class="n">x</span><span class="p">)</span> <span class="p">{</span> <span class="k">return</span> <span class="n">sin</span><span class="p">(</span><span class="n">x</span><span class="p">);</span> <span class="p">}</span>
+<span class="kt">long</span> <span class="kt">double</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">))</span> <span class="n">tgsin</span><span class="p">(</span><span class="kt">long</span> <span class="kt">double</span> <span class="n">x</span><span class="p">)</span> <span class="p">{</span> <span class="k">return</span> <span class="n">sinl</span><span class="p">(</span><span class="n">x</span><span class="p">);</span> <span class="p">}</span>
+</pre></div>
+</div>
+<p>Given these declarations, one can call <code class="docutils literal notranslate"><span class="pre">tgsin</span></code> with a <code class="docutils literal notranslate"><span class="pre">float</span></code> value to
+receive a <code class="docutils literal notranslate"><span class="pre">float</span></code> result, with a <code class="docutils literal notranslate"><span class="pre">double</span></code> to receive a <code class="docutils literal notranslate"><span class="pre">double</span></code> result,
+etc. Function overloading in C follows the rules of C++ function overloading
+to pick the best overload given the call arguments, with a few C-specific
+semantics:</p>
+<ul class="simple">
+<li>Conversion from <code class="docutils literal notranslate"><span class="pre">float</span></code> or <code class="docutils literal notranslate"><span class="pre">double</span></code> to <code class="docutils literal notranslate"><span class="pre">long</span> <span class="pre">double</span></code> is ranked as a
+floating-point promotion (per C99) rather than as a floating-point conversion
+(as in C++).</li>
+<li>A conversion from a pointer of type <code class="docutils literal notranslate"><span class="pre">T*</span></code> to a pointer of type <code class="docutils literal notranslate"><span class="pre">U*</span></code> is
+considered a pointer conversion (with conversion rank) if <code class="docutils literal notranslate"><span class="pre">T</span></code> and <code class="docutils literal notranslate"><span class="pre">U</span></code> are
+compatible types.</li>
+<li>A conversion from type <code class="docutils literal notranslate"><span class="pre">T</span></code> to a value of type <code class="docutils literal notranslate"><span class="pre">U</span></code> is permitted if <code class="docutils literal notranslate"><span class="pre">T</span></code>
+and <code class="docutils literal notranslate"><span class="pre">U</span></code> are compatible types. This conversion is given âconversionâ rank.</li>
+<li>If no viable candidates are otherwise available, we allow a conversion from a
+pointer of type <code class="docutils literal notranslate"><span class="pre">T*</span></code> to a pointer of type <code class="docutils literal notranslate"><span class="pre">U*</span></code>, where <code class="docutils literal notranslate"><span class="pre">T</span></code> and <code class="docutils literal notranslate"><span class="pre">U</span></code> are
+incompatible. This conversion is ranked below all other types of conversions.
+Please note: <code class="docutils literal notranslate"><span class="pre">U</span></code> lacking qualifiers that are present on <code class="docutils literal notranslate"><span class="pre">T</span></code> is sufficient
+for <code class="docutils literal notranslate"><span class="pre">T</span></code> and <code class="docutils literal notranslate"><span class="pre">U</span></code> to be incompatible.</li>
+</ul>
+<p>The declaration of <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> functions is restricted to function
+declarations and definitions. If a function is marked with the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code>
+attribute, then all declarations and definitions of functions with that name,
+except for at most one (see the note below about unmarked overloads), must have
+the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute. In addition, redeclarations of a function with
+the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute must have the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute, and
+redeclarations of a function without the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute must <em>not</em>
+have the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute. e.g.,</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">f</span><span class="p">(</span><span class="kt">int</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">));</span>
+<span class="kt">float</span> <span class="nf">f</span><span class="p">(</span><span class="kt">float</span><span class="p">);</span> <span class="c1">// error: declaration of "f" must have the "overloadable" attribute</span>
+<span class="kt">int</span> <span class="nf">f</span><span class="p">(</span><span class="kt">int</span><span class="p">);</span> <span class="c1">// error: redeclaration of "f" must have the "overloadable" attribute</span>
+
+<span class="kt">int</span> <span class="nf">g</span><span class="p">(</span><span class="kt">int</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">));</span>
+<span class="kt">int</span> <span class="nf">g</span><span class="p">(</span><span class="kt">int</span><span class="p">)</span> <span class="p">{</span> <span class="p">}</span> <span class="c1">// error: redeclaration of "g" must also have the "overloadable" attribute</span>
+
+<span class="kt">int</span> <span class="nf">h</span><span class="p">(</span><span class="kt">int</span><span class="p">);</span>
+<span class="kt">int</span> <span class="nf">h</span><span class="p">(</span><span class="kt">int</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">));</span> <span class="c1">// error: declaration of "h" must not</span>
+ <span class="c1">// have the "overloadable" attribute</span>
+</pre></div>
+</div>
+<p>Functions marked <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> must have prototypes. Therefore, the
+following code is ill-formed:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">h</span><span class="p">()</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">));</span> <span class="c1">// error: h does not have a prototype</span>
+</pre></div>
+</div>
+<p>However, <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> functions are allowed to use a ellipsis even if there
+are no named parameters (as is permitted in C++). This feature is particularly
+useful when combined with the <code class="docutils literal notranslate"><span class="pre">unavailable</span></code> attribute:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">honeypot</span><span class="p">(...)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">,</span> <span class="n">unavailable</span><span class="p">));</span> <span class="c1">// calling me is an error</span>
+</pre></div>
+</div>
+<p>Functions declared with the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute have their names mangled
+according to the same rules as C++ function names. For example, the three
+<code class="docutils literal notranslate"><span class="pre">tgsin</span></code> functions in our motivating example get the mangled names
+<code class="docutils literal notranslate"><span class="pre">_Z5tgsinf</span></code>, <code class="docutils literal notranslate"><span class="pre">_Z5tgsind</span></code>, and <code class="docutils literal notranslate"><span class="pre">_Z5tgsine</span></code>, respectively. There are two
+caveats to this use of name mangling:</p>
+<ul class="simple">
+<li>Future versions of Clang may change the name mangling of functions overloaded
+in C, so you should not depend on an specific mangling. To be completely
+safe, we strongly urge the use of <code class="docutils literal notranslate"><span class="pre">static</span> <span class="pre">inline</span></code> with <code class="docutils literal notranslate"><span class="pre">overloadable</span></code>
+functions.</li>
+<li>The <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute has almost no meaning when used in C++,
+because names will already be mangled and functions are already overloadable.
+However, when an <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> function occurs within an <code class="docutils literal notranslate"><span class="pre">extern</span> <span class="pre">"C"</span></code>
+linkage specification, itâs name <em>will</em> be mangled in the same way as it
+would in C.</li>
+</ul>
+<p>For the purpose of backwards compatibility, at most one function with the same
+name as other <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> functions may omit the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code>
+attribute. In this case, the function without the <code class="docutils literal notranslate"><span class="pre">overloadable</span></code> attribute
+will not have its name mangled.</p>
+<p>For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="c1">// Notes with mangled names assume Itanium mangling.</span>
+<span class="kt">int</span> <span class="nf">f</span><span class="p">(</span><span class="kt">int</span><span class="p">);</span>
+<span class="kt">int</span> <span class="nf">f</span><span class="p">(</span><span class="kt">double</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">overloadable</span><span class="p">));</span>
+<span class="kt">void</span> <span class="nf">foo</span><span class="p">()</span> <span class="p">{</span>
+ <span class="n">f</span><span class="p">(</span><span class="mi">5</span><span class="p">);</span> <span class="c1">// Emits a call to f (not _Z1fi, as it would with an overload that</span>
+ <span class="c1">// was marked with overloadable).</span>
+ <span class="n">f</span><span class="p">(</span><span class="mf">1.0</span><span class="p">);</span> <span class="c1">// Emits a call to _Z1fd.</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Support for unmarked overloads is not present in some versions of clang. You may
+query for it using <code class="docutils literal notranslate"><span class="pre">__has_extension(overloadable_unmarked)</span></code>.</p>
+<p>Query for this attribute with <code class="docutils literal notranslate"><span class="pre">__has_attribute(overloadable)</span></code>.</p>
+</div>
+<div class="section" id="release-capability-release-shared-capability-clang-release-capability-clang-release-shared-capability">
+<h3><a class="toc-backref" href="#id196">release_capability (release_shared_capability, clang::release_capability, clang::release_shared_capability)</a><a class="headerlink" href="#release-capability-release-shared-capability-clang-release-capability-clang-release-shared-capability" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id62">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id62" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Marks a function as releasing a capability.</p>
+</div>
+<div class="section" id="short-call-gnu-short-call-gnu-near">
+<h3><a class="toc-backref" href="#id197">short_call (gnu::short_call, gnu::near)</a><a class="headerlink" href="#short-call-gnu-short-call-gnu-near" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id63">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id63" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the <code class="docutils literal notranslate"><span class="pre">__attribute__((long_call))</span></code>, <code class="docutils literal notranslate"><span class="pre">__attribute__((far))</span></code>,
+<code class="docutils literal notranslate"><span class="pre">__attribute__((short__call))</span></code>, and <code class="docutils literal notranslate"><span class="pre">__attribute__((near))</span></code> attributes
+on MIPS targets. These attributes may only be added to function declarations
+and change the code generated by the compiler when directly calling
+the function. The <code class="docutils literal notranslate"><span class="pre">short_call</span></code> and <code class="docutils literal notranslate"><span class="pre">near</span></code> attributes are synonyms and
+allow calls to the function to be made using the <code class="docutils literal notranslate"><span class="pre">jal</span></code> instruction, which
+requires the function to be located in the same naturally aligned 256MB segment
+as the caller. The <code class="docutils literal notranslate"><span class="pre">long_call</span></code> and <code class="docutils literal notranslate"><span class="pre">far</span></code> attributes are synonyms and
+require the use of a different call sequence that works regardless
+of the distance between the functions.</p>
+<p>These attributes have no effect for position-independent code.</p>
+<p>These attributes take priority over command line switches such
+as <code class="docutils literal notranslate"><span class="pre">-mlong-calls</span></code> and <code class="docutils literal notranslate"><span class="pre">-mno-long-calls</span></code>.</p>
+</div>
+<div class="section" id="signal-gnu-signal">
+<h3><a class="toc-backref" href="#id198">signal (gnu::signal)</a><a class="headerlink" href="#signal-gnu-signal" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id64">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id64" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((signal))</span></code> attribute on
+AVR targets. This attribute may be attached to a function definition and instructs
+the backend to generate appropriate function entry/exit code so that it can be used
+directly as an interrupt service routine.</p>
+<p>Interrupt handler functions defined with the signal attribute do not re-enable interrupts.</p>
+</div>
+<div class="section" id="target-gnu-target">
+<h3><a class="toc-backref" href="#id199">target (gnu::target)</a><a class="headerlink" href="#target-gnu-target" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id65">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id65" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the GNU style <code class="docutils literal notranslate"><span class="pre">__attribute__((target("OPTIONS")))</span></code> attribute.
+This attribute may be attached to a function definition and instructs
+the backend to use different code generation options than were passed on the
+command line.</p>
+<p>The current set of options correspond to the existing âsubtarget featuresâ for
+the target with or without a â-mno-â in front corresponding to the absence
+of the feature, as well as <code class="docutils literal notranslate"><span class="pre">arch="CPU"</span></code> which will change the default âCPUâ
+for the function.</p>
+<p>Example âsubtarget featuresâ from the x86 backend include: âmmxâ, âsseâ, âsse4.2â,
+âavxâ, âxopâ and largely correspond to the machine specific options handled by
+the front end.</p>
+<p>Additionally, this attribute supports function multiversioning for ELF based
+x86/x86-64 targets, which can be used to create multiple implementations of the
+same function that will be resolved at runtime based on the priority of their
+<code class="docutils literal notranslate"><span class="pre">target</span></code> attribute strings. A function is considered a multiversioned function
+if either two declarations of the function have different <code class="docutils literal notranslate"><span class="pre">target</span></code> attribute
+strings, or if it has a <code class="docutils literal notranslate"><span class="pre">target</span></code> attribute string of <code class="docutils literal notranslate"><span class="pre">default</span></code>. For
+example:</p>
+<blockquote>
+<div><div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="n">__attribute__</span><span class="p">((</span><span class="n">target</span><span class="p">(</span><span class="s">"arch=atom"</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">foo</span><span class="p">()</span> <span class="p">{}</span> <span class="c1">// will be called on 'atom' processors.</span>
+<span class="n">__attribute__</span><span class="p">((</span><span class="n">target</span><span class="p">(</span><span class="s">"default"</span><span class="p">)))</span>
+<span class="kt">void</span> <span class="n">foo</span><span class="p">()</span> <span class="p">{}</span> <span class="c1">// will be called on any other processors.</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>All multiversioned functions must contain a <code class="docutils literal notranslate"><span class="pre">default</span></code> (fallback)
+implementation, otherwise usages of the function are considered invalid.
+Additionally, a function may not become multiversioned after its first use.</p>
+</div>
+<div class="section" id="try-acquire-capability-try-acquire-shared-capability-clang-try-acquire-capability-clang-try-acquire-shared-capability">
+<h3><a class="toc-backref" href="#id200">try_acquire_capability (try_acquire_shared_capability, clang::try_acquire_capability, clang::try_acquire_shared_capability)</a><a class="headerlink" href="#try-acquire-capability-try-acquire-shared-capability-clang-try-acquire-capability-clang-try-acquire-shared-capability" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id66">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id66" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Marks a function that attempts to acquire a capability. This function may fail to
+actually acquire the capability; they accept a Boolean value determining
+whether acquiring the capability means success (true), or failing to acquire
+the capability means success (false).</p>
+</div>
+<div class="section" id="xray-always-instrument-clang-xray-always-instrument-xray-never-instrument-clang-xray-never-instrument-xray-log-args-clang-xray-log-args">
+<h3><a class="toc-backref" href="#id201">xray_always_instrument (clang::xray_always_instrument), xray_never_instrument (clang::xray_never_instrument), xray_log_args (clang::xray_log_args)</a><a class="headerlink" href="#xray-always-instrument-clang-xray-always-instrument-xray-never-instrument-clang-xray-never-instrument-xray-log-args-clang-xray-log-args" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id67">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id67" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p><code class="docutils literal notranslate"><span class="pre">__attribute__((xray_always_instrument))</span></code> or <code class="docutils literal notranslate"><span class="pre">[[clang::xray_always_instrument]]</span></code> is used to mark member functions (in C++), methods (in Objective C), and free functions (in C, C++, and Objective C) to be instrumented with XRay. This will cause the function to always have space at the beginning and exit points to allow for runtime patching.</p>
+<p>Conversely, <code class="docutils literal notranslate"><span class="pre">__attribute__((xray_never_instrument))</span></code> or <code class="docutils literal notranslate"><span class="pre">[[clang::xray_never_instrument]]</span></code> will inhibit the insertion of these instrumentation points.</p>
+<p>If a function has neither of these attributes, they become subject to the XRay heuristics used to determine whether a function should be instrumented or otherwise.</p>
+<p><code class="docutils literal notranslate"><span class="pre">__attribute__((xray_log_args(N)))</span></code> or <code class="docutils literal notranslate"><span class="pre">[[clang::xray_log_args(N)]]</span></code> is used to preserve N function arguments for the logging function. Currently, only N==1 is supported.</p>
+</div>
+<div class="section" id="id2">
+<h3><a class="toc-backref" href="#id202">xray_always_instrument (clang::xray_always_instrument), xray_never_instrument (clang::xray_never_instrument), xray_log_args (clang::xray_log_args)</a><a class="headerlink" href="#id2" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id68">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id68" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p><code class="docutils literal notranslate"><span class="pre">__attribute__((xray_always_instrument))</span></code> or <code class="docutils literal notranslate"><span class="pre">[[clang::xray_always_instrument]]</span></code> is used to mark member functions (in C++), methods (in Objective C), and free functions (in C, C++, and Objective C) to be instrumented with XRay. This will cause the function to always have space at the beginning and exit points to allow for runtime patching.</p>
+<p>Conversely, <code class="docutils literal notranslate"><span class="pre">__attribute__((xray_never_instrument))</span></code> or <code class="docutils literal notranslate"><span class="pre">[[clang::xray_never_instrument]]</span></code> will inhibit the insertion of these instrumentation points.</p>
+<p>If a function has neither of these attributes, they become subject to the XRay heuristics used to determine whether a function should be instrumented or otherwise.</p>
+<p><code class="docutils literal notranslate"><span class="pre">__attribute__((xray_log_args(N)))</span></code> or <code class="docutils literal notranslate"><span class="pre">[[clang::xray_log_args(N)]]</span></code> is used to preserve N function arguments for the logging function. Currently, only N==1 is supported.</p>
+</div>
+</div>
+<div class="section" id="variable-attributes">
+<h2><a class="toc-backref" href="#id203">Variable Attributes</a><a class="headerlink" href="#variable-attributes" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="dllexport-gnu-dllexport">
+<h3><a class="toc-backref" href="#id204">dllexport (gnu::dllexport)</a><a class="headerlink" href="#dllexport-gnu-dllexport" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id69">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id69" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">__declspec(dllexport)</span></code> attribute declares a variable, function, or
+Objective-C interface to be exported from the module. It is available under the
+<code class="docutils literal notranslate"><span class="pre">-fdeclspec</span></code> flag for compatibility with various compilers. The primary use
+is for COFF object files which explicitly specify what interfaces are available
+for external use. See the <a class="reference external" href="https://msdn.microsoft.com/en-us/library/3y1sfaz2.aspx">dllexport</a> documentation on MSDN for more
+information.</p>
+</div>
+<div class="section" id="dllimport-gnu-dllimport">
+<h3><a class="toc-backref" href="#id205">dllimport (gnu::dllimport)</a><a class="headerlink" href="#dllimport-gnu-dllimport" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id70">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id70" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">__declspec(dllimport)</span></code> attribute declares a variable, function, or
+Objective-C interface to be imported from an external module. It is available
+under the <code class="docutils literal notranslate"><span class="pre">-fdeclspec</span></code> flag for compatibility with various compilers. The
+primary use is for COFF object files which explicitly specify what interfaces
+are imported from external modules. See the <a class="reference external" href="https://msdn.microsoft.com/en-us/library/3y1sfaz2.aspx">dllimport</a> documentation on MSDN
+for more information.</p>
+</div>
+<div class="section" id="init-seg">
+<h3><a class="toc-backref" href="#id206">init_seg</a><a class="headerlink" href="#init-seg" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id71">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id71" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The attribute applied by <code class="docutils literal notranslate"><span class="pre">pragma</span> <span class="pre">init_seg()</span></code> controls the section into
+which global initialization function pointers are emitted. It is only
+available with <code class="docutils literal notranslate"><span class="pre">-fms-extensions</span></code>. Typically, this function pointer is
+emitted into <code class="docutils literal notranslate"><span class="pre">.CRT$XCU</span></code> on Windows. The user can change the order of
+initialization by using a different section name with the same
+<code class="docutils literal notranslate"><span class="pre">.CRT$XC</span></code> prefix and a suffix that sorts lexicographically before or
+after the standard <code class="docutils literal notranslate"><span class="pre">.CRT$XCU</span></code> sections. See the <a class="reference external" href="http://msdn.microsoft.com/en-us/library/7977wcck(v=vs.110).aspx">init_seg</a>
+documentation on MSDN for more information.</p>
+</div>
+<div class="section" id="maybe-unused-unused-gnu-unused">
+<h3><a class="toc-backref" href="#id207">maybe_unused, unused, gnu::unused</a><a class="headerlink" href="#maybe-unused-unused-gnu-unused" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id72">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id72" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>When passing the <code class="docutils literal notranslate"><span class="pre">-Wunused</span></code> flag to Clang, entities that are unused by the
+program may be diagnosed. The <code class="docutils literal notranslate"><span class="pre">[[maybe_unused]]</span></code> (or
+<code class="docutils literal notranslate"><span class="pre">__attribute__((unused))</span></code>) attribute can be used to silence such diagnostics
+when the entity cannot be removed. For instance, a local variable may exist
+solely for use in an <code class="docutils literal notranslate"><span class="pre">assert()</span></code> statement, which makes the local variable
+unused when <code class="docutils literal notranslate"><span class="pre">NDEBUG</span></code> is defined.</p>
+<p>The attribute may be applied to the declaration of a class, a typedef, a
+variable, a function or method, a function parameter, an enumeration, an
+enumerator, a non-static data member, or a label.</p>
+</div>
+<div class="section" id="nodebug-gnu-nodebug">
+<h3><a class="toc-backref" href="#id208">nodebug (gnu::nodebug)</a><a class="headerlink" href="#nodebug-gnu-nodebug" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id73">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id73" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">nodebug</span></code> attribute allows you to suppress debugging information for a
+function or method, or for a variable that is not a parameter or a non-static
+data member.</p>
+</div>
+<div class="section" id="noescape-clang-noescape-clang-noescape">
+<h3><a class="toc-backref" href="#id209">noescape (clang::noescape, clang::noescape)</a><a class="headerlink" href="#noescape-clang-noescape-clang-noescape" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id74">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id74" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p><code class="docutils literal notranslate"><span class="pre">noescape</span></code> placed on a function parameter of a pointer type is used to inform
+the compiler that the pointer cannot escape: that is, no reference to the object
+the pointer points to that is derived from the parameter value will survive
+after the function returns. Users are responsible for making sure parameters
+annotated with <code class="docutils literal notranslate"><span class="pre">noescape</span></code> do not actuallly escape.</p>
+<p>For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="o">*</span><span class="n">gp</span><span class="p">;</span>
+
+<span class="kt">void</span> <span class="nf">nonescapingFunc</span><span class="p">(</span><span class="n">__attribute__</span><span class="p">((</span><span class="n">noescape</span><span class="p">))</span> <span class="kt">int</span> <span class="o">*</span><span class="n">p</span><span class="p">)</span> <span class="p">{</span>
+ <span class="o">*</span><span class="n">p</span> <span class="o">+=</span> <span class="mi">100</span><span class="p">;</span> <span class="c1">// OK.</span>
+<span class="p">}</span>
+
+<span class="kt">void</span> <span class="nf">escapingFunc</span><span class="p">(</span><span class="n">__attribute__</span><span class="p">((</span><span class="n">noescape</span><span class="p">))</span> <span class="kt">int</span> <span class="o">*</span><span class="n">p</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">gp</span> <span class="o">=</span> <span class="n">p</span><span class="p">;</span> <span class="c1">// Not OK.</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Additionally, when the parameter is a <cite>block pointer
+<https://clang.llvm.org/docs/BlockLanguageSpec.html></cite>, the same restriction
+applies to copies of the block. For example:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">typedef</span> <span class="nf">void</span> <span class="p">(</span><span class="o">^</span><span class="n">BlockTy</span><span class="p">)();</span>
+<span class="n">BlockTy</span> <span class="n">g0</span><span class="p">,</span> <span class="n">g1</span><span class="p">;</span>
+
+<span class="kt">void</span> <span class="nf">nonescapingFunc</span><span class="p">(</span><span class="n">__attribute__</span><span class="p">((</span><span class="n">noescape</span><span class="p">))</span> <span class="n">BlockTy</span> <span class="n">block</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">block</span><span class="p">();</span> <span class="c1">// OK.</span>
+<span class="p">}</span>
+
+<span class="kt">void</span> <span class="nf">escapingFunc</span><span class="p">(</span><span class="n">__attribute__</span><span class="p">((</span><span class="n">noescape</span><span class="p">))</span> <span class="n">BlockTy</span> <span class="n">block</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">g0</span> <span class="o">=</span> <span class="n">block</span><span class="p">;</span> <span class="c1">// Not OK.</span>
+ <span class="n">g1</span> <span class="o">=</span> <span class="n">Block_copy</span><span class="p">(</span><span class="n">block</span><span class="p">);</span> <span class="c1">// Not OK either.</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="nosvm">
+<h3><a class="toc-backref" href="#id210">nosvm</a><a class="headerlink" href="#nosvm" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id75">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id75" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>OpenCL 2.0 supports the optional <code class="docutils literal notranslate"><span class="pre">__attribute__((nosvm))</span></code> qualifier for
+pointer variable. It informs the compiler that the pointer does not refer
+to a shared virtual memory region. See OpenCL v2.0 s6.7.2 for details.</p>
+<p>Since it is not widely used and has been removed from OpenCL 2.1, it is ignored
+by Clang.</p>
+</div>
+<div class="section" id="pass-object-size-clang-pass-object-size-clang-pass-object-size">
+<h3><a class="toc-backref" href="#id211">pass_object_size (clang::pass_object_size, clang::pass_object_size)</a><a class="headerlink" href="#pass-object-size-clang-pass-object-size-clang-pass-object-size" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id76">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id76" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<div class="admonition note">
+<p class="first admonition-title">Note</p>
+<p class="last">The mangling of functions with parameters that are annotated with
+<code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> is subject to change. You can get around this by
+using <code class="docutils literal notranslate"><span class="pre">__asm__("foo")</span></code> to explicitly name your functions, thus preserving
+your ABI; also, non-overloadable C functions with <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> are
+not mangled.</p>
+</div>
+<p>The <code class="docutils literal notranslate"><span class="pre">pass_object_size(Type)</span></code> attribute can be placed on function parameters to
+instruct clang to call <code class="docutils literal notranslate"><span class="pre">__builtin_object_size(param,</span> <span class="pre">Type)</span></code> at each callsite
+of said function, and implicitly pass the result of this call in as an invisible
+argument of type <code class="docutils literal notranslate"><span class="pre">size_t</span></code> directly after the parameter annotated with
+<code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code>. Clang will also replace any calls to
+<code class="docutils literal notranslate"><span class="pre">__builtin_object_size(param,</span> <span class="pre">Type)</span></code> in the function by said implicit
+parameter.</p>
+<p>Example usage:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">bzero1</span><span class="p">(</span><span class="kt">char</span> <span class="o">*</span><span class="k">const</span> <span class="n">p</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">pass_object_size</span><span class="p">(</span><span class="mi">0</span><span class="p">))))</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">noinline</span><span class="p">))</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="k">for</span> <span class="p">(</span><span class="cm">/**/</span><span class="p">;</span> <span class="n">i</span> <span class="o"><</span> <span class="p">(</span><span class="kt">int</span><span class="p">)</span><span class="n">__builtin_object_size</span><span class="p">(</span><span class="n">p</span><span class="p">,</span> <span class="mi">0</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">p</span><span class="p">[</span><span class="n">i</span><span class="p">]</span> <span class="o">=</span> <span class="mi">0</span><span class="p">;</span>
+ <span class="p">}</span>
+ <span class="k">return</span> <span class="n">i</span><span class="p">;</span>
+<span class="p">}</span>
+
+<span class="kt">int</span> <span class="nf">main</span><span class="p">()</span> <span class="p">{</span>
+ <span class="kt">char</span> <span class="n">chars</span><span class="p">[</span><span class="mi">100</span><span class="p">];</span>
+ <span class="kt">int</span> <span class="n">n</span> <span class="o">=</span> <span class="n">bzero1</span><span class="p">(</span><span class="o">&</span><span class="n">chars</span><span class="p">[</span><span class="mi">0</span><span class="p">]);</span>
+ <span class="n">assert</span><span class="p">(</span><span class="n">n</span> <span class="o">==</span> <span class="k">sizeof</span><span class="p">(</span><span class="n">chars</span><span class="p">));</span>
+ <span class="k">return</span> <span class="mi">0</span><span class="p">;</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>If successfully evaluating <code class="docutils literal notranslate"><span class="pre">__builtin_object_size(param,</span> <span class="pre">Type)</span></code> at the
+callsite is not possible, then the âfailedâ value is passed in. So, using the
+definition of <code class="docutils literal notranslate"><span class="pre">bzero1</span></code> from above, the following code would exit cleanly:</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">main2</span><span class="p">(</span><span class="kt">int</span> <span class="n">argc</span><span class="p">,</span> <span class="kt">char</span> <span class="o">*</span><span class="n">argv</span><span class="p">[])</span> <span class="p">{</span>
+ <span class="kt">int</span> <span class="n">n</span> <span class="o">=</span> <span class="n">bzero1</span><span class="p">(</span><span class="n">argv</span><span class="p">);</span>
+ <span class="n">assert</span><span class="p">(</span><span class="n">n</span> <span class="o">==</span> <span class="o">-</span><span class="mi">1</span><span class="p">);</span>
+ <span class="k">return</span> <span class="mi">0</span><span class="p">;</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p><code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> plays a part in overload resolution. If two overload
+candidates are otherwise equally good, then the overload with one or more
+parameters with <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> is preferred. This implies that the choice
+between two identical overloads both with <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> on one or more
+parameters will always be ambiguous; for this reason, having two such overloads
+is illegal. For example:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#define PS(N) __attribute__((pass_object_size(N)))</span>
+<span class="c1">// OK</span>
+<span class="kt">void</span> <span class="nf">Foo</span><span class="p">(</span><span class="kt">char</span> <span class="o">*</span><span class="n">a</span><span class="p">,</span> <span class="kt">char</span> <span class="o">*</span><span class="n">b</span><span class="p">);</span> <span class="c1">// Overload A</span>
+<span class="c1">// OK -- overload A has no parameters with pass_object_size.</span>
+<span class="kt">void</span> <span class="nf">Foo</span><span class="p">(</span><span class="kt">char</span> <span class="o">*</span><span class="n">a</span> <span class="n">PS</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span> <span class="kt">char</span> <span class="o">*</span><span class="n">b</span> <span class="n">PS</span><span class="p">(</span><span class="mi">0</span><span class="p">));</span> <span class="c1">// Overload B</span>
+<span class="c1">// Error -- Same signature (sans pass_object_size) as overload B, and both</span>
+<span class="c1">// overloads have one or more parameters with the pass_object_size attribute.</span>
+<span class="kt">void</span> <span class="nf">Foo</span><span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">a</span> <span class="n">PS</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span> <span class="kt">void</span> <span class="o">*</span><span class="n">b</span><span class="p">);</span>
+
+<span class="c1">// OK</span>
+<span class="kt">void</span> <span class="nf">Bar</span><span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">a</span> <span class="n">PS</span><span class="p">(</span><span class="mi">0</span><span class="p">));</span> <span class="c1">// Overload C</span>
+<span class="c1">// OK</span>
+<span class="kt">void</span> <span class="nf">Bar</span><span class="p">(</span><span class="kt">char</span> <span class="o">*</span><span class="n">c</span> <span class="n">PS</span><span class="p">(</span><span class="mi">1</span><span class="p">));</span> <span class="c1">// Overload D</span>
+
+<span class="kt">void</span> <span class="nf">main</span><span class="p">()</span> <span class="p">{</span>
+ <span class="kt">char</span> <span class="n">known</span><span class="p">[</span><span class="mi">10</span><span class="p">],</span> <span class="o">*</span><span class="n">unknown</span><span class="p">;</span>
+ <span class="n">Foo</span><span class="p">(</span><span class="n">unknown</span><span class="p">,</span> <span class="n">unknown</span><span class="p">);</span> <span class="c1">// Calls overload B</span>
+ <span class="n">Foo</span><span class="p">(</span><span class="n">known</span><span class="p">,</span> <span class="n">unknown</span><span class="p">);</span> <span class="c1">// Calls overload B</span>
+ <span class="n">Foo</span><span class="p">(</span><span class="n">unknown</span><span class="p">,</span> <span class="n">known</span><span class="p">);</span> <span class="c1">// Calls overload B</span>
+ <span class="n">Foo</span><span class="p">(</span><span class="n">known</span><span class="p">,</span> <span class="n">known</span><span class="p">);</span> <span class="c1">// Calls overload B</span>
+
+ <span class="n">Bar</span><span class="p">(</span><span class="n">known</span><span class="p">);</span> <span class="c1">// Calls overload D</span>
+ <span class="n">Bar</span><span class="p">(</span><span class="n">unknown</span><span class="p">);</span> <span class="c1">// Calls overload D</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Currently, <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> is a bit restricted in terms of its usage:</p>
+<ul class="simple">
+<li>Only one use of <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> is allowed per parameter.</li>
+<li>It is an error to take the address of a function with <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> on
+any of its parameters. If you wish to do this, you can create an overload
+without <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> on any parameters.</li>
+<li>It is an error to apply the <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> attribute to parameters that
+are not pointers. Additionally, any parameter that <code class="docutils literal notranslate"><span class="pre">pass_object_size</span></code> is
+applied to must be marked <code class="docutils literal notranslate"><span class="pre">const</span></code> at its functionâs definition.</li>
+</ul>
+</div>
+<div class="section" id="require-constant-initialization-clang-require-constant-initialization">
+<h3><a class="toc-backref" href="#id212">require_constant_initialization (clang::require_constant_initialization)</a><a class="headerlink" href="#require-constant-initialization-clang-require-constant-initialization" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id77">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id77" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>This attribute specifies that the variable to which it is attached is intended
+to have a <a class="reference external" href="http://en.cppreference.com/w/cpp/language/constant_initialization">constant initializer</a>
+according to the rules of [basic.start.static]. The variable is required to
+have static or thread storage duration. If the initialization of the variable
+is not a constant initializer an error will be produced. This attribute may
+only be used in C++.</p>
+<p>Note that in C++03 strict constant expression checking is not done. Instead
+the attribute reports if Clang can emit the variable as a constant, even if itâs
+not technically a âconstant initializerâ. This behavior is non-portable.</p>
+<p>Static storage duration variables with constant initializers avoid hard-to-find
+bugs caused by the indeterminate order of dynamic initialization. They can also
+be safely used during dynamic initialization across translation units.</p>
+<p>This attribute acts as a compile time assertion that the requirements
+for constant initialization have been met. Since these requirements change
+between dialects and have subtle pitfalls itâs important to fail fast instead
+of silently falling back on dynamic initialization.</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="c1">// -std=c++14</span>
+<span class="cp">#define SAFE_STATIC [[clang::require_constant_initialization]]</span>
+<span class="k">struct</span> <span class="n">T</span> <span class="p">{</span>
+ <span class="k">constexpr</span> <span class="n">T</span><span class="p">(</span><span class="kt">int</span><span class="p">)</span> <span class="p">{}</span>
+ <span class="o">~</span><span class="n">T</span><span class="p">();</span> <span class="c1">// non-trivial</span>
+<span class="p">};</span>
+<span class="n">SAFE_STATIC</span> <span class="n">T</span> <span class="n">x</span> <span class="o">=</span> <span class="p">{</span><span class="mi">42</span><span class="p">};</span> <span class="c1">// Initialization OK. Doesn't check destructor.</span>
+<span class="n">SAFE_STATIC</span> <span class="n">T</span> <span class="n">y</span> <span class="o">=</span> <span class="mi">42</span><span class="p">;</span> <span class="c1">// error: variable does not have a constant initializer</span>
+<span class="c1">// copy initialization is not a constant expression on a non-literal type.</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="section-gnu-section-declspec-allocate">
+<h3><a class="toc-backref" href="#id213">section (gnu::section, __declspec(allocate))</a><a class="headerlink" href="#section-gnu-section-declspec-allocate" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id78">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id78" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">section</span></code> attribute allows you to specify a specific section a
+global variable or function should be in after translation.</p>
+</div>
+<div class="section" id="swift-context-clang-swift-context-clang-swift-context">
+<h3><a class="toc-backref" href="#id214">swift_context (clang::swift_context, clang::swift_context)</a><a class="headerlink" href="#swift-context-clang-swift-context-clang-swift-context" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id79">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id79" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">swift_context</span></code> attribute marks a parameter of a <code class="docutils literal notranslate"><span class="pre">swiftcall</span></code>
+function as having the special context-parameter ABI treatment.</p>
+<p>This treatment generally passes the context value in a special register
+which is normally callee-preserved.</p>
+<p>A <code class="docutils literal notranslate"><span class="pre">swift_context</span></code> parameter must either be the last parameter or must be
+followed by a <code class="docutils literal notranslate"><span class="pre">swift_error_result</span></code> parameter (which itself must always be
+the last parameter).</p>
+<p>A context parameter must have pointer or reference type.</p>
+</div>
+<div class="section" id="swift-error-result-clang-swift-error-result-clang-swift-error-result">
+<h3><a class="toc-backref" href="#id215">swift_error_result (clang::swift_error_result, clang::swift_error_result)</a><a class="headerlink" href="#swift-error-result-clang-swift-error-result-clang-swift-error-result" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id80">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id80" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">swift_error_result</span></code> attribute marks a parameter of a <code class="docutils literal notranslate"><span class="pre">swiftcall</span></code>
+function as having the special error-result ABI treatment.</p>
+<p>This treatment generally passes the underlying error value in and out of
+the function through a special register which is normally callee-preserved.
+This is modeled in C by pretending that the register is addressable memory:</p>
+<ul class="simple">
+<li>The caller appears to pass the address of a variable of pointer type.
+The current value of this variable is copied into the register before
+the call; if the call returns normally, the value is copied back into the
+variable.</li>
+<li>The callee appears to receive the address of a variable. This address
+is actually a hidden location in its own stack, initialized with the
+value of the register upon entry. When the function returns normally,
+the value in that hidden location is written back to the register.</li>
+</ul>
+<p>A <code class="docutils literal notranslate"><span class="pre">swift_error_result</span></code> parameter must be the last parameter, and it must be
+preceded by a <code class="docutils literal notranslate"><span class="pre">swift_context</span></code> parameter.</p>
+<p>A <code class="docutils literal notranslate"><span class="pre">swift_error_result</span></code> parameter must have type <code class="docutils literal notranslate"><span class="pre">T**</span></code> or <code class="docutils literal notranslate"><span class="pre">T*&</span></code> for some
+type T. Note that no qualifiers are permitted on the intermediate level.</p>
+<p>It is undefined behavior if the caller does not pass a pointer or
+reference to a valid object.</p>
+<p>The standard convention is that the error value itself (that is, the
+value stored in the apparent argument) will be null upon function entry,
+but this is not enforced by the ABI.</p>
+</div>
+<div class="section" id="swift-indirect-result-clang-swift-indirect-result-clang-swift-indirect-result">
+<h3><a class="toc-backref" href="#id216">swift_indirect_result (clang::swift_indirect_result, clang::swift_indirect_result)</a><a class="headerlink" href="#swift-indirect-result-clang-swift-indirect-result-clang-swift-indirect-result" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id81">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id81" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">swift_indirect_result</span></code> attribute marks a parameter of a <code class="docutils literal notranslate"><span class="pre">swiftcall</span></code>
+function as having the special indirect-result ABI treatment.</p>
+<p>This treatment gives the parameter the targetâs normal indirect-result
+ABI treatment, which may involve passing it differently from an ordinary
+parameter. However, only the first indirect result will receive this
+treatment. Furthermore, low-level lowering may decide that a direct result
+must be returned indirectly; if so, this will take priority over the
+<code class="docutils literal notranslate"><span class="pre">swift_indirect_result</span></code> parameters.</p>
+<p>A <code class="docutils literal notranslate"><span class="pre">swift_indirect_result</span></code> parameter must either be the first parameter or
+follow another <code class="docutils literal notranslate"><span class="pre">swift_indirect_result</span></code> parameter.</p>
+<p>A <code class="docutils literal notranslate"><span class="pre">swift_indirect_result</span></code> parameter must have type <code class="docutils literal notranslate"><span class="pre">T*</span></code> or <code class="docutils literal notranslate"><span class="pre">T&</span></code> for
+some object type <code class="docutils literal notranslate"><span class="pre">T</span></code>. If <code class="docutils literal notranslate"><span class="pre">T</span></code> is a complete type at the point of
+definition of a function, it is undefined behavior if the argument
+value does not point to storage of adequate size and alignment for a
+value of type <code class="docutils literal notranslate"><span class="pre">T</span></code>.</p>
+<p>Making indirect results explicit in the signature allows C functions to
+directly construct objects into them without relying on language
+optimizations like C++âs named return value optimization (NRVO).</p>
+</div>
+<div class="section" id="swiftcall-clang-swiftcall-clang-swiftcall">
+<h3><a class="toc-backref" href="#id217">swiftcall (clang::swiftcall, clang::swiftcall)</a><a class="headerlink" href="#swiftcall-clang-swiftcall-clang-swiftcall" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id82">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id82" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">swiftcall</span></code> attribute indicates that a function should be called
+using the Swift calling convention for a function or function pointer.</p>
+<p>The lowering for the Swift calling convention, as described by the Swift
+ABI documentation, occurs in multiple phases. The first, âhigh-levelâ
+phase breaks down the formal parameters and results into innately direct
+and indirect components, adds implicit paraameters for the generic
+signature, and assigns the context and error ABI treatments to parameters
+where applicable. The second phase breaks down the direct parameters
+and results from the first phase and assigns them to registers or the
+stack. The <code class="docutils literal notranslate"><span class="pre">swiftcall</span></code> convention only handles this second phase of
+lowering; the C function type must accurately reflect the results
+of the first phase, as follows:</p>
+<ul class="simple">
+<li>Results classified as indirect by high-level lowering should be
+represented as parameters with the <code class="docutils literal notranslate"><span class="pre">swift_indirect_result</span></code> attribute.</li>
+<li>Results classified as direct by high-level lowering should be represented
+as follows:<ul>
+<li>First, remove any empty direct results.</li>
+<li>If there are no direct results, the C result type should be <code class="docutils literal notranslate"><span class="pre">void</span></code>.</li>
+<li>If there is one direct result, the C result type should be a type with
+the exact layout of that result type.</li>
+<li>If there are a multiple direct results, the C result type should be
+a struct type with the exact layout of a tuple of those results.</li>
+</ul>
+</li>
+<li>Parameters classified as indirect by high-level lowering should be
+represented as parameters of pointer type.</li>
+<li>Parameters classified as direct by high-level lowering should be
+omitted if they are empty types; otherwise, they should be represented
+as a parameter type with a layout exactly matching the layout of the
+Swift parameter type.</li>
+<li>The context parameter, if present, should be represented as a trailing
+parameter with the <code class="docutils literal notranslate"><span class="pre">swift_context</span></code> attribute.</li>
+<li>The error result parameter, if present, should be represented as a
+trailing parameter (always following a context parameter) with the
+<code class="docutils literal notranslate"><span class="pre">swift_error_result</span></code> attribute.</li>
+</ul>
+<p><code class="docutils literal notranslate"><span class="pre">swiftcall</span></code> does not support variadic arguments or unprototyped functions.</p>
+<p>The parameter ABI treatment attributes are aspects of the function type.
+A function type which which applies an ABI treatment attribute to a
+parameter is a different type from an otherwise-identical function type
+that does not. A single parameter may not have multiple ABI treatment
+attributes.</p>
+<p>Support for this feature is target-dependent, although it should be
+supported on every target that Swift supports. Query for this support
+with <code class="docutils literal notranslate"><span class="pre">__has_attribute(swiftcall)</span></code>. This implies support for the
+<code class="docutils literal notranslate"><span class="pre">swift_context</span></code>, <code class="docutils literal notranslate"><span class="pre">swift_error_result</span></code>, and <code class="docutils literal notranslate"><span class="pre">swift_indirect_result</span></code>
+attributes.</p>
+</div>
+<div class="section" id="thread">
+<h3><a class="toc-backref" href="#id218">thread</a><a class="headerlink" href="#thread" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id83">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id83" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">__declspec(thread)</span></code> attribute declares a variable with thread local
+storage. It is available under the <code class="docutils literal notranslate"><span class="pre">-fms-extensions</span></code> flag for MSVC
+compatibility. See the documentation for <a class="reference external" href="http://msdn.microsoft.com/en-us/library/9w1sdazb.aspx">__declspec(thread)</a> on MSDN.</p>
+<p>In Clang, <code class="docutils literal notranslate"><span class="pre">__declspec(thread)</span></code> is generally equivalent in functionality to the
+GNU <code class="docutils literal notranslate"><span class="pre">__thread</span></code> keyword. The variable must not have a destructor and must have
+a constant initializer, if any. The attribute only applies to variables
+declared with static storage duration, such as globals, class static data
+members, and static locals.</p>
+</div>
+<div class="section" id="tls-model-gnu-tls-model">
+<h3><a class="toc-backref" href="#id219">tls_model (gnu::tls_model)</a><a class="headerlink" href="#tls-model-gnu-tls-model" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id84">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id84" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">tls_model</span></code> attribute allows you to specify which thread-local storage
+model to use. It accepts the following strings:</p>
+<ul class="simple">
+<li>global-dynamic</li>
+<li>local-dynamic</li>
+<li>initial-exec</li>
+<li>local-exec</li>
+</ul>
+<p>TLS models are mutually exclusive.</p>
+</div>
+<div class="section" id="trivial-abi-clang-trivial-abi">
+<h3><a class="toc-backref" href="#id220">trivial_abi (clang::trivial_abi)</a><a class="headerlink" href="#trivial-abi-clang-trivial-abi" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id85">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id85" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">trivial_abi</span></code> attribute can be applied to a C++ class, struct, or union.
+It instructs the compiler to pass and return the type using the C ABI for the
+underlying type when the type would otherwise be considered non-trivial for the
+purpose of calls.
+A class annotated with <cite>trivial_abi</cite> can have non-trivial destructors or copy/move constructors without automatically becoming non-trivial for the purposes of calls. For example:</p>
+<blockquote>
+<div><div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="c1">// A is trivial for the purposes of calls because `trivial_abi` makes the</span>
+<span class="c1">// user-provided special functions trivial.</span>
+<span class="k">struct</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">trivial_abi</span><span class="p">))</span> <span class="n">A</span> <span class="p">{</span>
+ <span class="o">~</span><span class="n">A</span><span class="p">();</span>
+ <span class="n">A</span><span class="p">(</span><span class="k">const</span> <span class="n">A</span> <span class="o">&</span><span class="p">);</span>
+ <span class="n">A</span><span class="p">(</span><span class="n">A</span> <span class="o">&&</span><span class="p">);</span>
+ <span class="kt">int</span> <span class="n">x</span><span class="p">;</span>
+<span class="p">};</span>
+
+<span class="c1">// B's destructor and copy/move constructor are considered trivial for the</span>
+<span class="c1">// purpose of calls because A is trivial.</span>
+<span class="k">struct</span> <span class="n">B</span> <span class="p">{</span>
+ <span class="n">A</span> <span class="n">a</span><span class="p">;</span>
+<span class="p">};</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>If a type is trivial for the purposes of calls, has a non-trivial destructor,
+and is passed as an argument by value, the convention is that the callee will
+destroy the object before returning.</p>
+<p>Attribute <code class="docutils literal notranslate"><span class="pre">trivial_abi</span></code> has no effect in the following cases:</p>
+<ul class="simple">
+<li>The class directly declares a virtual base or virtual methods.</li>
+<li>The class has a base class that is non-trivial for the purposes of calls.</li>
+<li>The class has a non-static data member whose type is non-trivial for the purposes of calls, which includes:<ul>
+<li>classes that are non-trivial for the purposes of calls</li>
+<li>__weak-qualified types in Objective-C++</li>
+<li>arrays of any of the above</li>
+</ul>
+</li>
+</ul>
+</div>
+</div>
+<div class="section" id="type-attributes">
+<h2><a class="toc-backref" href="#id221">Type Attributes</a><a class="headerlink" href="#type-attributes" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="single-inhertiance-multiple-inheritance-virtual-inheritance">
+<h3><a class="toc-backref" href="#id222">__single_inhertiance, __multiple_inheritance, __virtual_inheritance</a><a class="headerlink" href="#single-inhertiance-multiple-inheritance-virtual-inheritance" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id86">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id86" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>This collection of keywords is enabled under <code class="docutils literal notranslate"><span class="pre">-fms-extensions</span></code> and controls
+the pointer-to-member representation used on <code class="docutils literal notranslate"><span class="pre">*-*-win32</span></code> targets.</p>
+<p>The <code class="docutils literal notranslate"><span class="pre">*-*-win32</span></code> targets utilize a pointer-to-member representation which
+varies in size and alignment depending on the definition of the underlying
+class.</p>
+<p>However, this is problematic when a forward declaration is only available and
+no definition has been made yet. In such cases, Clang is forced to utilize the
+most general representation that is available to it.</p>
+<p>These keywords make it possible to use a pointer-to-member representation other
+than the most general one regardless of whether or not the definition will ever
+be present in the current translation unit.</p>
+<p>This family of keywords belong between the <code class="docutils literal notranslate"><span class="pre">class-key</span></code> and <code class="docutils literal notranslate"><span class="pre">class-name</span></code>:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="k">struct</span> <span class="n">__single_inheritance</span> <span class="n">S</span><span class="p">;</span>
+<span class="kt">int</span> <span class="n">S</span><span class="o">::*</span><span class="n">i</span><span class="p">;</span>
+<span class="k">struct</span> <span class="n">S</span> <span class="p">{};</span>
+</pre></div>
+</div>
+<p>This keyword can be applied to class templates but only has an effect when used
+on full specializations:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="k">template</span> <span class="o"><</span><span class="k">typename</span> <span class="n">T</span><span class="p">,</span> <span class="k">typename</span> <span class="n">U</span><span class="o">></span> <span class="k">struct</span> <span class="n">__single_inheritance</span> <span class="n">A</span><span class="p">;</span> <span class="c1">// warning: inheritance model ignored on primary template</span>
+<span class="k">template</span> <span class="o"><</span><span class="k">typename</span> <span class="n">T</span><span class="o">></span> <span class="k">struct</span> <span class="n">__multiple_inheritance</span> <span class="n">A</span><span class="o"><</span><span class="n">T</span><span class="p">,</span> <span class="n">T</span><span class="o">></span><span class="p">;</span> <span class="c1">// warning: inheritance model ignored on partial specialization</span>
+<span class="k">template</span> <span class="o"><></span> <span class="k">struct</span> <span class="n">__single_inheritance</span> <span class="n">A</span><span class="o"><</span><span class="kt">int</span><span class="p">,</span> <span class="kt">float</span><span class="o">></span><span class="p">;</span>
+</pre></div>
+</div>
+<p>Note that choosing an inheritance model less general than strictly necessary is
+an error:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="k">struct</span> <span class="n">__multiple_inheritance</span> <span class="n">S</span><span class="p">;</span> <span class="c1">// error: inheritance model does not match definition</span>
+<span class="kt">int</span> <span class="n">S</span><span class="o">::*</span><span class="n">i</span><span class="p">;</span>
+<span class="k">struct</span> <span class="n">S</span> <span class="p">{};</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="align-value">
+<h3><a class="toc-backref" href="#id223">align_value</a><a class="headerlink" href="#align-value" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id87">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id87" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The align_value attribute can be added to the typedef of a pointer type or the
+declaration of a variable of pointer or reference type. It specifies that the
+pointer will point to, or the reference will bind to, only objects with at
+least the provided alignment. This alignment value must be some positive power
+of 2.</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">typedef</span> <span class="kt">double</span> <span class="o">*</span> <span class="n">aligned_double_ptr</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">align_value</span><span class="p">(</span><span class="mi">64</span><span class="p">)));</span>
+<span class="kt">void</span> <span class="nf">foo</span><span class="p">(</span><span class="kt">double</span> <span class="o">&</span> <span class="n">x</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">align_value</span><span class="p">(</span><span class="mi">128</span><span class="p">)),</span>
+ <span class="n">aligned_double_ptr</span> <span class="n">y</span><span class="p">)</span> <span class="p">{</span> <span class="p">...</span> <span class="p">}</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>If the pointer value does not have the specified alignment at runtime, the
+behavior of the program is undefined.</p>
+</div>
+<div class="section" id="empty-bases">
+<h3><a class="toc-backref" href="#id224">empty_bases</a><a class="headerlink" href="#empty-bases" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id88">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id88" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The empty_bases attribute permits the compiler to utilize the
+empty-base-optimization more frequently.
+This attribute only applies to struct, class, and union types.
+It is only supported when using the Microsoft C++ ABI.</p>
+</div>
+<div class="section" id="enum-extensibility-clang-enum-extensibility-clang-enum-extensibility">
+<h3><a class="toc-backref" href="#id225">enum_extensibility (clang::enum_extensibility, clang::enum_extensibility)</a><a class="headerlink" href="#enum-extensibility-clang-enum-extensibility-clang-enum-extensibility" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id89">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id89" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Attribute <code class="docutils literal notranslate"><span class="pre">enum_extensibility</span></code> is used to distinguish between enum definitions
+that are extensible and those that are not. The attribute can take either
+<code class="docutils literal notranslate"><span class="pre">closed</span></code> or <code class="docutils literal notranslate"><span class="pre">open</span></code> as an argument. <code class="docutils literal notranslate"><span class="pre">closed</span></code> indicates a variable of the
+enum type takes a value that corresponds to one of the enumerators listed in the
+enum definition or, when the enum is annotated with <code class="docutils literal notranslate"><span class="pre">flag_enum</span></code>, a value that
+can be constructed using values corresponding to the enumerators. <code class="docutils literal notranslate"><span class="pre">open</span></code>
+indicates a variable of the enum type can take any values allowed by the
+standard and instructs clang to be more lenient when issuing warnings.</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">enum</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">enum_extensibility</span><span class="p">(</span><span class="n">closed</span><span class="p">)))</span> <span class="n">ClosedEnum</span> <span class="p">{</span>
+ <span class="n">A0</span><span class="p">,</span> <span class="n">A1</span>
+<span class="p">};</span>
+
+<span class="k">enum</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">enum_extensibility</span><span class="p">(</span><span class="n">open</span><span class="p">)))</span> <span class="n">OpenEnum</span> <span class="p">{</span>
+ <span class="n">B0</span><span class="p">,</span> <span class="n">B1</span>
+<span class="p">};</span>
+
+<span class="k">enum</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">enum_extensibility</span><span class="p">(</span><span class="n">closed</span><span class="p">),</span><span class="n">flag_enum</span><span class="p">))</span> <span class="n">ClosedFlagEnum</span> <span class="p">{</span>
+ <span class="n">C0</span> <span class="o">=</span> <span class="mi">1</span> <span class="o"><<</span> <span class="mi">0</span><span class="p">,</span> <span class="n">C1</span> <span class="o">=</span> <span class="mi">1</span> <span class="o"><<</span> <span class="mi">1</span>
+<span class="p">};</span>
+
+<span class="k">enum</span> <span class="nf">__attribute__</span><span class="p">((</span><span class="n">enum_extensibility</span><span class="p">(</span><span class="n">open</span><span class="p">),</span><span class="n">flag_enum</span><span class="p">))</span> <span class="n">OpenFlagEnum</span> <span class="p">{</span>
+ <span class="n">D0</span> <span class="o">=</span> <span class="mi">1</span> <span class="o"><<</span> <span class="mi">0</span><span class="p">,</span> <span class="n">D1</span> <span class="o">=</span> <span class="mi">1</span> <span class="o"><<</span> <span class="mi">1</span>
+<span class="p">};</span>
+
+<span class="kt">void</span> <span class="nf">foo1</span><span class="p">()</span> <span class="p">{</span>
+ <span class="k">enum</span> <span class="n">ClosedEnum</span> <span class="n">ce</span><span class="p">;</span>
+ <span class="k">enum</span> <span class="n">OpenEnum</span> <span class="n">oe</span><span class="p">;</span>
+ <span class="k">enum</span> <span class="n">ClosedFlagEnum</span> <span class="n">cfe</span><span class="p">;</span>
+ <span class="k">enum</span> <span class="n">OpenFlagEnum</span> <span class="n">ofe</span><span class="p">;</span>
+
+ <span class="n">ce</span> <span class="o">=</span> <span class="n">A1</span><span class="p">;</span> <span class="c1">// no warnings</span>
+ <span class="n">ce</span> <span class="o">=</span> <span class="mi">100</span><span class="p">;</span> <span class="c1">// warning issued</span>
+ <span class="n">oe</span> <span class="o">=</span> <span class="n">B1</span><span class="p">;</span> <span class="c1">// no warnings</span>
+ <span class="n">oe</span> <span class="o">=</span> <span class="mi">100</span><span class="p">;</span> <span class="c1">// no warnings</span>
+ <span class="n">cfe</span> <span class="o">=</span> <span class="n">C0</span> <span class="o">|</span> <span class="n">C1</span><span class="p">;</span> <span class="c1">// no warnings</span>
+ <span class="n">cfe</span> <span class="o">=</span> <span class="n">C0</span> <span class="o">|</span> <span class="n">C1</span> <span class="o">|</span> <span class="mi">4</span><span class="p">;</span> <span class="c1">// warning issued</span>
+ <span class="n">ofe</span> <span class="o">=</span> <span class="n">D0</span> <span class="o">|</span> <span class="n">D1</span><span class="p">;</span> <span class="c1">// no warnings</span>
+ <span class="n">ofe</span> <span class="o">=</span> <span class="n">D0</span> <span class="o">|</span> <span class="n">D1</span> <span class="o">|</span> <span class="mi">4</span><span class="p">;</span> <span class="c1">// no warnings</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="flag-enum-clang-flag-enum-clang-flag-enum">
+<h3><a class="toc-backref" href="#id226">flag_enum (clang::flag_enum, clang::flag_enum)</a><a class="headerlink" href="#flag-enum-clang-flag-enum-clang-flag-enum" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id90">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id90" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>This attribute can be added to an enumerator to signal to the compiler that it
+is intended to be used as a flag type. This will cause the compiler to assume
+that the range of the type includes all of the values that you can get by
+manipulating bits of the enumerator when issuing warnings.</p>
+</div>
+<div class="section" id="layout-version">
+<h3><a class="toc-backref" href="#id227">layout_version</a><a class="headerlink" href="#layout-version" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id91">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id91" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The layout_version attribute requests that the compiler utilize the class
+layout rules of a particular compiler version.
+This attribute only applies to struct, class, and union types.
+It is only supported when using the Microsoft C++ ABI.</p>
+</div>
+<div class="section" id="lto-visibility-public-clang-lto-visibility-public-clang-lto-visibility-public">
+<h3><a class="toc-backref" href="#id228">lto_visibility_public (clang::lto_visibility_public, clang::lto_visibility_public)</a><a class="headerlink" href="#lto-visibility-public-clang-lto-visibility-public-clang-lto-visibility-public" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id92">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id92" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>See <a class="reference internal" href="LTOVisibility.html"><span class="doc">LTO Visibility</span></a>.</p>
+</div>
+<div class="section" id="novtable">
+<h3><a class="toc-backref" href="#id229">novtable</a><a class="headerlink" href="#novtable" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id93">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id93" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>This attribute can be added to a class declaration or definition to signal to
+the compiler that constructors and destructors will not reference the virtual
+function table. It is only supported when using the Microsoft C++ ABI.</p>
+</div>
+<div class="section" id="objc-subclassing-restricted-clang-objc-subclassing-restricted-clang-objc-subclassing-restricted">
+<h3><a class="toc-backref" href="#id230">objc_subclassing_restricted (clang::objc_subclassing_restricted, clang::objc_subclassing_restricted)</a><a class="headerlink" href="#objc-subclassing-restricted-clang-objc-subclassing-restricted-clang-objc-subclassing-restricted" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id94">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id94" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>This attribute can be added to an Objective-C <code class="docutils literal notranslate"><span class="pre">@interface</span></code> declaration to
+ensure that this class cannot be subclassed.</p>
+</div>
+<div class="section" id="selectany-gnu-selectany">
+<h3><a class="toc-backref" href="#id231">selectany (gnu::selectany)</a><a class="headerlink" href="#selectany-gnu-selectany" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id95">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id95" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>This attribute appertains to a global symbol, causing it to have a weak
+definition (
+<a class="reference external" href="https://llvm.org/docs/LangRef.html#linkage-types">linkonce</a>
+), allowing the linker to select any definition.</p>
+<p>For more information see
+<a class="reference external" href="https://gcc.gnu.org/onlinedocs/gcc-7.2.0/gcc/Microsoft-Windows-Variable-Attributes.html">gcc documentation</a>
+or <a class="reference external" href="https://docs.microsoft.com/pl-pl/cpp/cpp/selectany">msvc documentation</a>.</p>
+</div>
+<div class="section" id="transparent-union-gnu-transparent-union">
+<h3><a class="toc-backref" href="#id232">transparent_union (gnu::transparent_union)</a><a class="headerlink" href="#transparent-union-gnu-transparent-union" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id96">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id96" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>This attribute can be applied to a union to change the behaviour of calls to
+functions that have an argument with a transparent union type. The compiler
+behaviour is changed in the following manner:</p>
+<ul class="simple">
+<li>A value whose type is any member of the transparent union can be passed as an
+argument without the need to cast that value.</li>
+<li>The argument is passed to the function using the calling convention of the
+first member of the transparent union. Consequently, all the members of the
+transparent union should have the same calling convention as its first member.</li>
+</ul>
+<p>Transparent unions are not supported in C++.</p>
+</div>
+</div>
+<div class="section" id="statement-attributes">
+<h2><a class="toc-backref" href="#id233">Statement Attributes</a><a class="headerlink" href="#statement-attributes" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="pragma-clang-loop">
+<h3><a class="toc-backref" href="#id234">#pragma clang loop</a><a class="headerlink" href="#pragma-clang-loop" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id97">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id97" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">clang</span> <span class="pre">loop</span></code> directive allows loop optimization hints to be
+specified for the subsequent loop. The directive allows vectorization,
+interleaving, and unrolling to be enabled or disabled. Vector width as well
+as interleave and unrolling count can be manually specified. See
+<a class="reference external" href="http://clang.llvm.org/docs/LanguageExtensions.html#extensions-for-loop-hint-optimizations">language extensions</a>
+for details.</p>
+</div>
+<div class="section" id="pragma-unroll-pragma-nounroll">
+<h3><a class="toc-backref" href="#id235">#pragma unroll, #pragma nounroll</a><a class="headerlink" href="#pragma-unroll-pragma-nounroll" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id98">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id98" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Loop unrolling optimization hints can be specified with <code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">unroll</span></code> and
+<code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">nounroll</span></code>. The pragma is placed immediately before a for, while,
+do-while, or c++11 range-based for loop.</p>
+<p>Specifying <code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">unroll</span></code> without a parameter directs the loop unroller to
+attempt to fully unroll the loop if the trip count is known at compile time and
+attempt to partially unroll the loop if the trip count is not known at compile
+time:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#pragma unroll</span>
+<span class="k">for</span> <span class="p">(...)</span> <span class="p">{</span>
+ <span class="p">...</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Specifying the optional parameter, <code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">unroll</span> <span class="pre">_value_</span></code>, directs the
+unroller to unroll the loop <code class="docutils literal notranslate"><span class="pre">_value_</span></code> times. The parameter may optionally be
+enclosed in parentheses:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#pragma unroll 16</span>
+<span class="k">for</span> <span class="p">(...)</span> <span class="p">{</span>
+ <span class="p">...</span>
+<span class="p">}</span>
+
+<span class="cp">#pragma unroll(16)</span>
+<span class="k">for</span> <span class="p">(...)</span> <span class="p">{</span>
+ <span class="p">...</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Specifying <code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">nounroll</span></code> indicates that the loop should not be unrolled:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#pragma nounroll</span>
+<span class="k">for</span> <span class="p">(...)</span> <span class="p">{</span>
+ <span class="p">...</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p><code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">unroll</span></code> and <code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">unroll</span> <span class="pre">_value_</span></code> have identical semantics to
+<code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">clang</span> <span class="pre">loop</span> <span class="pre">unroll(full)</span></code> and
+<code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">clang</span> <span class="pre">loop</span> <span class="pre">unroll_count(_value_)</span></code> respectively. <code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">nounroll</span></code>
+is equivalent to <code class="docutils literal notranslate"><span class="pre">#pragma</span> <span class="pre">clang</span> <span class="pre">loop</span> <span class="pre">unroll(disable)</span></code>. See
+<a class="reference external" href="http://clang.llvm.org/docs/LanguageExtensions.html#extensions-for-loop-hint-optimizations">language extensions</a>
+for further details including limitations of the unroll hints.</p>
+</div>
+<div class="section" id="attribute-intel-reqd-sub-group-size">
+<h3><a class="toc-backref" href="#id236">__attribute__((intel_reqd_sub_group_size))</a><a class="headerlink" href="#attribute-intel-reqd-sub-group-size" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id99">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id99" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The optional attribute intel_reqd_sub_group_size can be used to indicate that
+the kernel must be compiled and executed with the specified subgroup size. When
+this attribute is present, get_max_sub_group_size() is guaranteed to return the
+specified integer value. This is important for the correctness of many subgroup
+algorithms, and in some cases may be used by the compiler to generate more optimal
+code. See <cite>cl_intel_required_subgroup_size
+<https://www.khronos.org/registry/OpenCL/extensions/intel/cl_intel_required_subgroup_size.txt></cite>
+for details.</p>
+</div>
+<div class="section" id="attribute-opencl-unroll-hint">
+<h3><a class="toc-backref" href="#id237">__attribute__((opencl_unroll_hint))</a><a class="headerlink" href="#attribute-opencl-unroll-hint" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id100">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id100" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The opencl_unroll_hint attribute qualifier can be used to specify that a loop
+(for, while and do loops) can be unrolled. This attribute qualifier can be
+used to specify full unrolling or partial unrolling by a specified amount.
+This is a compiler hint and the compiler may ignore this directive. See
+<a class="reference external" href="https://www.khronos.org/registry/cl/specs/opencl-2.0.pdf">OpenCL v2.0</a>
+s6.11.5 for details.</p>
+</div>
+<div class="section" id="read-only-write-only-read-write-read-only-write-only-read-write">
+<h3><a class="toc-backref" href="#id238">__read_only, __write_only, __read_write (read_only, write_only, read_write)</a><a class="headerlink" href="#read-only-write-only-read-write-read-only-write-only-read-write" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id101">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id101" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The access qualifiers must be used with image object arguments or pipe arguments
+to declare if they are being read or written by a kernel or function.</p>
+<p>The read_only/__read_only, write_only/__write_only and read_write/__read_write
+names are reserved for use as access qualifiers and shall not be used otherwise.</p>
+<div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="n">kernel</span> <span class="kt">void</span>
+<span class="nf">foo</span> <span class="p">(</span><span class="n">read_only</span> <span class="n">image2d_t</span> <span class="n">imageA</span><span class="p">,</span>
+ <span class="n">write_only</span> <span class="n">image2d_t</span> <span class="n">imageB</span><span class="p">)</span> <span class="p">{</span>
+ <span class="p">...</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>In the above example imageA is a read-only 2D image object, and imageB is a
+write-only 2D image object.</p>
+<p>The read_write (or __read_write) qualifier can not be used with pipe.</p>
+<p>More details can be found in the OpenCL C language Spec v2.0, Section 6.6.</p>
+</div>
+<div class="section" id="fallthrough-clang-fallthrough">
+<h3><a class="toc-backref" href="#id239">fallthrough, clang::fallthrough</a><a class="headerlink" href="#fallthrough-clang-fallthrough" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id102">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id102" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">fallthrough</span></code> (or <code class="docutils literal notranslate"><span class="pre">clang::fallthrough</span></code>) attribute is used
+to annotate intentional fall-through
+between switch labels. It can only be applied to a null statement placed at a
+point of execution between any statement and the next switch label. It is
+common to mark these places with a specific comment, but this attribute is
+meant to replace comments with a more strict annotation, which can be checked
+by the compiler. This attribute doesnât change semantics of the code and can
+be used wherever an intended fall-through occurs. It is designed to mimic
+control-flow statements like <code class="docutils literal notranslate"><span class="pre">break;</span></code>, so it can be placed in most places
+where <code class="docutils literal notranslate"><span class="pre">break;</span></code> can, but only if there are no statements on the execution path
+between it and the next switch label.</p>
+<p>By default, Clang does not warn on unannotated fallthrough from one <code class="docutils literal notranslate"><span class="pre">switch</span></code>
+case to another. Diagnostics on fallthrough without a corresponding annotation
+can be enabled with the <code class="docutils literal notranslate"><span class="pre">-Wimplicit-fallthrough</span></code> argument.</p>
+<p>Here is an example:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="c1">// compile with -Wimplicit-fallthrough</span>
+<span class="k">switch</span> <span class="p">(</span><span class="n">n</span><span class="p">)</span> <span class="p">{</span>
+<span class="k">case</span> <span class="mi">22</span><span class="o">:</span>
+<span class="k">case</span> <span class="mi">33</span><span class="o">:</span> <span class="c1">// no warning: no statements between case labels</span>
+ <span class="n">f</span><span class="p">();</span>
+<span class="k">case</span> <span class="mi">44</span><span class="o">:</span> <span class="c1">// warning: unannotated fall-through</span>
+ <span class="n">g</span><span class="p">();</span>
+ <span class="p">[[</span><span class="n">clang</span><span class="o">::</span><span class="n">fallthrough</span><span class="p">]];</span>
+<span class="k">case</span> <span class="mi">55</span><span class="o">:</span> <span class="c1">// no warning</span>
+ <span class="k">if</span> <span class="p">(</span><span class="n">x</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">h</span><span class="p">();</span>
+ <span class="k">break</span><span class="p">;</span>
+ <span class="p">}</span>
+ <span class="k">else</span> <span class="p">{</span>
+ <span class="n">i</span><span class="p">();</span>
+ <span class="p">[[</span><span class="n">clang</span><span class="o">::</span><span class="n">fallthrough</span><span class="p">]];</span>
+ <span class="p">}</span>
+<span class="k">case</span> <span class="mi">66</span><span class="o">:</span> <span class="c1">// no warning</span>
+ <span class="n">p</span><span class="p">();</span>
+ <span class="p">[[</span><span class="n">clang</span><span class="o">::</span><span class="n">fallthrough</span><span class="p">]];</span> <span class="c1">// warning: fallthrough annotation does not</span>
+ <span class="c1">// directly precede case label</span>
+ <span class="n">q</span><span class="p">();</span>
+<span class="k">case</span> <span class="mi">77</span><span class="o">:</span> <span class="c1">// warning: unannotated fall-through</span>
+ <span class="n">r</span><span class="p">();</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="suppress-gsl-suppress">
+<h3><a class="toc-backref" href="#id240">suppress (gsl::suppress)</a><a class="headerlink" href="#suppress-gsl-suppress" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id103">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id103" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">[[gsl::suppress]]</span></code> attribute suppresses specific
+clang-tidy diagnostics for rules of the <a class="reference external" href="https://github.com/isocpp/CppCoreGuidelines/blob/master/CppCoreGuidelines.md#inforce-enforcement">C++ Core Guidelines</a> in a portable
+way. The attribute can be attached to declarations, statements, and at
+namespace scope.</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="p">[[</span><span class="n">gsl</span><span class="o">::</span><span class="n">suppress</span><span class="p">(</span><span class="s">"Rh-public"</span><span class="p">)]]</span>
+<span class="kt">void</span> <span class="n">f_</span><span class="p">()</span> <span class="p">{</span>
+ <span class="kt">int</span> <span class="o">*</span><span class="n">p</span><span class="p">;</span>
+ <span class="p">[[</span><span class="n">gsl</span><span class="o">::</span><span class="n">suppress</span><span class="p">(</span><span class="s">"type"</span><span class="p">)]]</span> <span class="p">{</span>
+ <span class="n">p</span> <span class="o">=</span> <span class="k">reinterpret_cast</span><span class="o"><</span><span class="kt">int</span><span class="o">*></span><span class="p">(</span><span class="mi">7</span><span class="p">);</span>
+ <span class="p">}</span>
+<span class="p">}</span>
+<span class="k">namespace</span> <span class="n">N</span> <span class="p">{</span>
+ <span class="p">[[</span><span class="n">clang</span><span class="o">::</span><span class="n">suppress</span><span class="p">(</span><span class="s">"type"</span><span class="p">,</span> <span class="s">"bounds"</span><span class="p">)]];</span>
+ <span class="p">...</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div>
+</div>
+<div class="section" id="amd-gpu-attributes">
+<h2><a class="toc-backref" href="#id241">AMD GPU Attributes</a><a class="headerlink" href="#amd-gpu-attributes" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="amdgpu-flat-work-group-size-clang-amdgpu-flat-work-group-size">
+<h3><a class="toc-backref" href="#id242">amdgpu_flat_work_group_size (clang::amdgpu_flat_work_group_size)</a><a class="headerlink" href="#amdgpu-flat-work-group-size-clang-amdgpu-flat-work-group-size" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id104">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id104" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The flat work-group size is the number of work-items in the work-group size
+specified when the kernel is dispatched. It is the product of the sizes of the
+x, y, and z dimension of the work-group.</p>
+<p>Clang supports the
+<code class="docutils literal notranslate"><span class="pre">__attribute__((amdgpu_flat_work_group_size(<min>,</span> <span class="pre"><max>)))</span></code> attribute for the
+AMDGPU target. This attribute may be attached to a kernel function definition
+and is an optimization hint.</p>
+<p><code class="docutils literal notranslate"><span class="pre"><min></span></code> parameter specifies the minimum flat work-group size, and <code class="docutils literal notranslate"><span class="pre"><max></span></code>
+parameter specifies the maximum flat work-group size (must be greater than
+<code class="docutils literal notranslate"><span class="pre"><min></span></code>) to which all dispatches of the kernel will conform. Passing <code class="docutils literal notranslate"><span class="pre">0,</span> <span class="pre">0</span></code>
+as <code class="docutils literal notranslate"><span class="pre"><min>,</span> <span class="pre"><max></span></code> implies the default behavior (<code class="docutils literal notranslate"><span class="pre">128,</span> <span class="pre">256</span></code>).</p>
+<p>If specified, the AMDGPU target backend might be able to produce better machine
+code for barriers and perform scratch promotion by estimating available group
+segment size.</p>
+<dl class="docutils">
+<dt>An error will be given if:</dt>
+<dd><ul class="first last simple">
+<li>Specified values violate subtarget specifications;</li>
+<li>Specified values are not compatible with values provided through other
+attributes.</li>
+</ul>
+</dd>
+</dl>
+</div>
+<div class="section" id="amdgpu-num-sgpr-clang-amdgpu-num-sgpr">
+<h3><a class="toc-backref" href="#id243">amdgpu_num_sgpr (clang::amdgpu_num_sgpr)</a><a class="headerlink" href="#amdgpu-num-sgpr-clang-amdgpu-num-sgpr" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id105">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id105" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the <code class="docutils literal notranslate"><span class="pre">__attribute__((amdgpu_num_sgpr(<num_sgpr>)))</span></code> and
+<code class="docutils literal notranslate"><span class="pre">__attribute__((amdgpu_num_vgpr(<num_vgpr>)))</span></code> attributes for the AMDGPU
+target. These attributes may be attached to a kernel function definition and are
+an optimization hint.</p>
+<p>If these attributes are specified, then the AMDGPU target backend will attempt
+to limit the number of SGPRs and/or VGPRs used to the specified value(s). The
+number of used SGPRs and/or VGPRs may further be rounded up to satisfy the
+allocation requirements or constraints of the subtarget. Passing <code class="docutils literal notranslate"><span class="pre">0</span></code> as
+<code class="docutils literal notranslate"><span class="pre">num_sgpr</span></code> and/or <code class="docutils literal notranslate"><span class="pre">num_vgpr</span></code> implies the default behavior (no limits).</p>
+<p>These attributes can be used to test the AMDGPU target backend. It is
+recommended that the <code class="docutils literal notranslate"><span class="pre">amdgpu_waves_per_eu</span></code> attribute be used to control
+resources such as SGPRs and VGPRs since it is aware of the limits for different
+subtargets.</p>
+<dl class="docutils">
+<dt>An error will be given if:</dt>
+<dd><ul class="first last simple">
+<li>Specified values violate subtarget specifications;</li>
+<li>Specified values are not compatible with values provided through other
+attributes;</li>
+<li>The AMDGPU target backend is unable to create machine code that can meet the
+request.</li>
+</ul>
+</dd>
+</dl>
+</div>
+<div class="section" id="amdgpu-num-vgpr-clang-amdgpu-num-vgpr">
+<h3><a class="toc-backref" href="#id244">amdgpu_num_vgpr (clang::amdgpu_num_vgpr)</a><a class="headerlink" href="#amdgpu-num-vgpr-clang-amdgpu-num-vgpr" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id106">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id106" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Clang supports the <code class="docutils literal notranslate"><span class="pre">__attribute__((amdgpu_num_sgpr(<num_sgpr>)))</span></code> and
+<code class="docutils literal notranslate"><span class="pre">__attribute__((amdgpu_num_vgpr(<num_vgpr>)))</span></code> attributes for the AMDGPU
+target. These attributes may be attached to a kernel function definition and are
+an optimization hint.</p>
+<p>If these attributes are specified, then the AMDGPU target backend will attempt
+to limit the number of SGPRs and/or VGPRs used to the specified value(s). The
+number of used SGPRs and/or VGPRs may further be rounded up to satisfy the
+allocation requirements or constraints of the subtarget. Passing <code class="docutils literal notranslate"><span class="pre">0</span></code> as
+<code class="docutils literal notranslate"><span class="pre">num_sgpr</span></code> and/or <code class="docutils literal notranslate"><span class="pre">num_vgpr</span></code> implies the default behavior (no limits).</p>
+<p>These attributes can be used to test the AMDGPU target backend. It is
+recommended that the <code class="docutils literal notranslate"><span class="pre">amdgpu_waves_per_eu</span></code> attribute be used to control
+resources such as SGPRs and VGPRs since it is aware of the limits for different
+subtargets.</p>
+<dl class="docutils">
+<dt>An error will be given if:</dt>
+<dd><ul class="first last simple">
+<li>Specified values violate subtarget specifications;</li>
+<li>Specified values are not compatible with values provided through other
+attributes;</li>
+<li>The AMDGPU target backend is unable to create machine code that can meet the
+request.</li>
+</ul>
+</dd>
+</dl>
+</div>
+<div class="section" id="amdgpu-waves-per-eu-clang-amdgpu-waves-per-eu">
+<h3><a class="toc-backref" href="#id245">amdgpu_waves_per_eu (clang::amdgpu_waves_per_eu)</a><a class="headerlink" href="#amdgpu-waves-per-eu-clang-amdgpu-waves-per-eu" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id107">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id107" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>A compute unit (CU) is responsible for executing the wavefronts of a work-group.
+It is composed of one or more execution units (EU), which are responsible for
+executing the wavefronts. An EU can have enough resources to maintain the state
+of more than one executing wavefront. This allows an EU to hide latency by
+switching between wavefronts in a similar way to symmetric multithreading on a
+CPU. In order to allow the state for multiple wavefronts to fit on an EU, the
+resources used by a single wavefront have to be limited. For example, the number
+of SGPRs and VGPRs. Limiting such resources can allow greater latency hiding,
+but can result in having to spill some register state to memory.</p>
+<p>Clang supports the <code class="docutils literal notranslate"><span class="pre">__attribute__((amdgpu_waves_per_eu(<min>[,</span> <span class="pre"><max>])))</span></code>
+attribute for the AMDGPU target. This attribute may be attached to a kernel
+function definition and is an optimization hint.</p>
+<p><code class="docutils literal notranslate"><span class="pre"><min></span></code> parameter specifies the requested minimum number of waves per EU, and
+<em>optional</em> <code class="docutils literal notranslate"><span class="pre"><max></span></code> parameter specifies the requested maximum number of waves
+per EU (must be greater than <code class="docutils literal notranslate"><span class="pre"><min></span></code> if specified). If <code class="docutils literal notranslate"><span class="pre"><max></span></code> is omitted,
+then there is no restriction on the maximum number of waves per EU other than
+the one dictated by the hardware for which the kernel is compiled. Passing
+<code class="docutils literal notranslate"><span class="pre">0,</span> <span class="pre">0</span></code> as <code class="docutils literal notranslate"><span class="pre"><min>,</span> <span class="pre"><max></span></code> implies the default behavior (no limits).</p>
+<p>If specified, this attribute allows an advanced developer to tune the number of
+wavefronts that are capable of fitting within the resources of an EU. The AMDGPU
+target backend can use this information to limit resources, such as number of
+SGPRs, number of VGPRs, size of available group and private memory segments, in
+such a way that guarantees that at least <code class="docutils literal notranslate"><span class="pre"><min></span></code> wavefronts and at most
+<code class="docutils literal notranslate"><span class="pre"><max></span></code> wavefronts are able to fit within the resources of an EU. Requesting
+more wavefronts can hide memory latency but limits available registers which
+can result in spilling. Requesting fewer wavefronts can help reduce cache
+thrashing, but can reduce memory latency hiding.</p>
+<p>This attribute controls the machine code generated by the AMDGPU target backend
+to ensure it is capable of meeting the requested values. However, when the
+kernel is executed, there may be other reasons that prevent meeting the request,
+for example, there may be wavefronts from other kernels executing on the EU.</p>
+<dl class="docutils">
+<dt>An error will be given if:</dt>
+<dd><ul class="first last simple">
+<li>Specified values violate subtarget specifications;</li>
+<li>Specified values are not compatible with values provided through other
+attributes;</li>
+<li>The AMDGPU target backend is unable to create machine code that can meet the
+request.</li>
+</ul>
+</dd>
+</dl>
+</div>
+</div>
+<div class="section" id="calling-conventions">
+<h2><a class="toc-backref" href="#id246">Calling Conventions</a><a class="headerlink" href="#calling-conventions" title="Permalink to this headline">¶</a></h2>
+<p>Clang supports several different calling conventions, depending on the target
+platform and architecture. The calling convention used for a function determines
+how parameters are passed, how results are returned to the caller, and other
+low-level details of calling a function.</p>
+<div class="section" id="fastcall-gnu-fastcall-fastcall-fastcall">
+<h3><a class="toc-backref" href="#id247">fastcall (gnu::fastcall, __fastcall, _fastcall)</a><a class="headerlink" href="#fastcall-gnu-fastcall-fastcall-fastcall" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id108">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id108" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On 32-bit x86 targets, this attribute changes the calling convention of a
+function to use ECX and EDX as register parameters and clear parameters off of
+the stack on return. This convention does not support variadic calls or
+unprototyped functions in C, and has no effect on x86_64 targets. This calling
+convention is supported primarily for compatibility with existing code. Users
+seeking register parameters should use the <code class="docutils literal notranslate"><span class="pre">regparm</span></code> attribute, which does
+not require callee-cleanup. See the documentation for <a class="reference external" href="http://msdn.microsoft.com/en-us/library/6xa169sk.aspx">__fastcall</a> on MSDN.</p>
+</div>
+<div class="section" id="ms-abi-gnu-ms-abi">
+<h3><a class="toc-backref" href="#id248">ms_abi (gnu::ms_abi)</a><a class="headerlink" href="#ms-abi-gnu-ms-abi" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id109">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id109" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On non-Windows x86_64 targets, this attribute changes the calling convention of
+a function to match the default convention used on Windows x86_64. This
+attribute has no effect on Windows targets or non-x86_64 targets.</p>
+</div>
+<div class="section" id="pcs-gnu-pcs">
+<h3><a class="toc-backref" href="#id249">pcs (gnu::pcs)</a><a class="headerlink" href="#pcs-gnu-pcs" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id110">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id110" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On ARM targets, this attribute can be used to select calling conventions
+similar to <code class="docutils literal notranslate"><span class="pre">stdcall</span></code> on x86. Valid parameter values are âaapcsâ and
+âaapcs-vfpâ.</p>
+</div>
+<div class="section" id="preserve-all-clang-preserve-all-clang-preserve-all">
+<h3><a class="toc-backref" href="#id250">preserve_all (clang::preserve_all, clang::preserve_all)</a><a class="headerlink" href="#preserve-all-clang-preserve-all-clang-preserve-all" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id111">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id111" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On X86-64 and AArch64 targets, this attribute changes the calling convention of
+a function. The <code class="docutils literal notranslate"><span class="pre">preserve_all</span></code> calling convention attempts to make the code
+in the caller even less intrusive than the <code class="docutils literal notranslate"><span class="pre">preserve_most</span></code> calling convention.
+This calling convention also behaves identical to the <code class="docutils literal notranslate"><span class="pre">C</span></code> 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 doesnât 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 donât need to call out to any other functions.</p>
+<p>This calling convention, like the <code class="docutils literal notranslate"><span class="pre">preserve_most</span></code> calling convention, will be
+used by a future version of the Objective-C runtime and should be considered
+experimental at this time.</p>
+</div>
+<div class="section" id="preserve-most-clang-preserve-most-clang-preserve-most">
+<h3><a class="toc-backref" href="#id251">preserve_most (clang::preserve_most, clang::preserve_most)</a><a class="headerlink" href="#preserve-most-clang-preserve-most-clang-preserve-most" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id112">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id112" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On X86-64 and AArch64 targets, this attribute changes the calling convention of
+a function. The <code class="docutils literal notranslate"><span class="pre">preserve_most</span></code> calling convention attempts to make the code
+in the caller as unintrusive as possible. This convention behaves identically
+to the <code class="docutils literal notranslate"><span class="pre">C</span></code> 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 doesnât
+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 doesnât 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 havenât already been saved by the caller. The
+<cite>preserve_most</cite> calling convention is very similar to the <code class="docutils literal notranslate"><span class="pre">cold</span></code> calling
+convention in terms of caller/callee-saved registers, but they are used for
+different types of function calls. <code class="docutils literal notranslate"><span class="pre">coldcc</span></code> is for function calls that are
+rarely executed, whereas <cite>preserve_most</cite> function calls are intended to be
+on the hot path and definitely executed a lot. Furthermore <code class="docutils literal notranslate"><span class="pre">preserve_most</span></code>
+doesnât prevent the inliner from inlining the function call.</p>
+<p>This calling convention will be used by a future version of the Objective-C
+runtime and should therefore still be considered experimental at this time.
+Although this convention was created to optimize certain runtime calls to
+the Objective-C 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 and AArch64, but the intention is to support more architectures
+in the future.</p>
+</div>
+<div class="section" id="regcall-gnu-regcall-regcall">
+<h3><a class="toc-backref" href="#id252">regcall (gnu::regcall, __regcall)</a><a class="headerlink" href="#regcall-gnu-regcall-regcall" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id113">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id113" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On x86 targets, this attribute changes the calling convention to
+<a class="reference external" href="https://software.intel.com/en-us/node/693069">__regcall</a> convention. This convention aims to pass as many arguments
+as possible in registers. It also tries to utilize registers for the
+return value whenever it is possible.</p>
+</div>
+<div class="section" id="regparm-gnu-regparm">
+<h3><a class="toc-backref" href="#id253">regparm (gnu::regparm)</a><a class="headerlink" href="#regparm-gnu-regparm" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id114">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id114" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On 32-bit x86 targets, the regparm attribute causes the compiler to pass
+the first three integer parameters in EAX, EDX, and ECX instead of on the
+stack. This attribute has no effect on variadic functions, and all parameters
+are passed via the stack as normal.</p>
+</div>
+<div class="section" id="stdcall-gnu-stdcall-stdcall-stdcall">
+<h3><a class="toc-backref" href="#id254">stdcall (gnu::stdcall, __stdcall, _stdcall)</a><a class="headerlink" href="#stdcall-gnu-stdcall-stdcall-stdcall" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id115">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id115" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On 32-bit x86 targets, this attribute changes the calling convention of a
+function to clear parameters off of the stack on return. This convention does
+not support variadic calls or unprototyped functions in C, and has no effect on
+x86_64 targets. This calling convention is used widely by the Windows API and
+COM applications. See the documentation for <a class="reference external" href="http://msdn.microsoft.com/en-us/library/zxk0tw93.aspx">__stdcall</a> on MSDN.</p>
+</div>
+<div class="section" id="thiscall-gnu-thiscall-thiscall-thiscall">
+<h3><a class="toc-backref" href="#id255">thiscall (gnu::thiscall, __thiscall, _thiscall)</a><a class="headerlink" href="#thiscall-gnu-thiscall-thiscall-thiscall" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id116">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id116" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On 32-bit x86 targets, this attribute changes the calling convention of a
+function to use ECX for the first parameter (typically the implicit <code class="docutils literal notranslate"><span class="pre">this</span></code>
+parameter of C++ methods) and clear parameters off of the stack on return. This
+convention does not support variadic calls or unprototyped functions in C, and
+has no effect on x86_64 targets. See the documentation for <a class="reference external" href="http://msdn.microsoft.com/en-us/library/ek8tkfbw.aspx">__thiscall</a> on
+MSDN.</p>
+</div>
+<div class="section" id="vectorcall-clang-vectorcall-clang-vectorcall-vectorcall-vectorcall">
+<h3><a class="toc-backref" href="#id256">vectorcall (clang::vectorcall, clang::vectorcall, __vectorcall, _vectorcall)</a><a class="headerlink" href="#vectorcall-clang-vectorcall-clang-vectorcall-vectorcall-vectorcall" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id117">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id117" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>On 32-bit x86 <em>and</em> x86_64 targets, this attribute changes the calling
+convention of a function to pass vector parameters in SSE registers.</p>
+<p>On 32-bit x86 targets, this calling convention is similar to <code class="docutils literal notranslate"><span class="pre">__fastcall</span></code>.
+The first two integer parameters are passed in ECX and EDX. Subsequent integer
+parameters are passed in memory, and callee clears the stack. On x86_64
+targets, the callee does <em>not</em> clear the stack, and integer parameters are
+passed in RCX, RDX, R8, and R9 as is done for the default Windows x64 calling
+convention.</p>
+<p>On both 32-bit x86 and x86_64 targets, vector and floating point arguments are
+passed in XMM0-XMM5. Homogeneous vector aggregates of up to four elements are
+passed in sequential SSE registers if enough are available. If AVX is enabled,
+256 bit vectors are passed in YMM0-YMM5. Any vector or aggregate type that
+cannot be passed in registers for any reason is passed by reference, which
+allows the caller to align the parameter memory.</p>
+<p>See the documentation for <a class="reference external" href="http://msdn.microsoft.com/en-us/library/dn375768.aspx">__vectorcall</a> on MSDN for more details.</p>
+</div>
+</div>
+<div class="section" id="consumed-annotation-checking">
+<h2><a class="toc-backref" href="#id257">Consumed Annotation Checking</a><a class="headerlink" href="#consumed-annotation-checking" title="Permalink to this headline">¶</a></h2>
+<p>Clang supports additional attributes for checking basic resource management
+properties, specifically for unique objects that have a single owning reference.
+The following attributes are currently supported, although <strong>the implementation
+for these annotations is currently in development and are subject to change.</strong></p>
+<div class="section" id="callable-when-clang-callable-when">
+<h3><a class="toc-backref" href="#id258">callable_when (clang::callable_when)</a><a class="headerlink" href="#callable-when-clang-callable-when" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id118">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id118" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Use <code class="docutils literal notranslate"><span class="pre">__attribute__((callable_when(...)))</span></code> to indicate what states a method
+may be called in. Valid states are unconsumed, consumed, or unknown. Each
+argument to this attribute must be a quoted string. E.g.:</p>
+<p><code class="docutils literal notranslate"><span class="pre">__attribute__((callable_when("unconsumed",</span> <span class="pre">"unknown")))</span></code></p>
+</div>
+<div class="section" id="consumable-clang-consumable">
+<h3><a class="toc-backref" href="#id259">consumable (clang::consumable)</a><a class="headerlink" href="#consumable-clang-consumable" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id119">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id119" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Each <code class="docutils literal notranslate"><span class="pre">class</span></code> that uses any of the typestate annotations must first be marked
+using the <code class="docutils literal notranslate"><span class="pre">consumable</span></code> attribute. Failure to do so will result in a warning.</p>
+<p>This attribute accepts a single parameter that must be one of the following:
+<code class="docutils literal notranslate"><span class="pre">unknown</span></code>, <code class="docutils literal notranslate"><span class="pre">consumed</span></code>, or <code class="docutils literal notranslate"><span class="pre">unconsumed</span></code>.</p>
+</div>
+<div class="section" id="param-typestate-clang-param-typestate">
+<h3><a class="toc-backref" href="#id260">param_typestate (clang::param_typestate)</a><a class="headerlink" href="#param-typestate-clang-param-typestate" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id120">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id120" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>This attribute specifies expectations about function parameters. Calls to an
+function with annotated parameters will issue a warning if the corresponding
+argument isnât in the expected state. The attribute is also used to set the
+initial state of the parameter when analyzing the functionâs body.</p>
+</div>
+<div class="section" id="return-typestate-clang-return-typestate">
+<h3><a class="toc-backref" href="#id261">return_typestate (clang::return_typestate)</a><a class="headerlink" href="#return-typestate-clang-return-typestate" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id121">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id121" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">return_typestate</span></code> attribute can be applied to functions or parameters.
+When applied to a function the attribute specifies the state of the returned
+value. The functionâs body is checked to ensure that it always returns a value
+in the specified state. On the caller side, values returned by the annotated
+function are initialized to the given state.</p>
+<p>When applied to a function parameter it modifies the state of an argument after
+a call to the function returns. The functionâs body is checked to ensure that
+the parameter is in the expected state before returning.</p>
+</div>
+<div class="section" id="set-typestate-clang-set-typestate">
+<h3><a class="toc-backref" href="#id262">set_typestate (clang::set_typestate)</a><a class="headerlink" href="#set-typestate-clang-set-typestate" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id122">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id122" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Annotate methods that transition an object into a new state with
+<code class="docutils literal notranslate"><span class="pre">__attribute__((set_typestate(new_state)))</span></code>. The new state must be
+unconsumed, consumed, or unknown.</p>
+</div>
+<div class="section" id="test-typestate-clang-test-typestate">
+<h3><a class="toc-backref" href="#id263">test_typestate (clang::test_typestate)</a><a class="headerlink" href="#test-typestate-clang-test-typestate" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id123">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id123" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>Use <code class="docutils literal notranslate"><span class="pre">__attribute__((test_typestate(tested_state)))</span></code> to indicate that a method
+returns true if the object is in the specified state..</p>
+</div>
+</div>
+<div class="section" id="type-safety-checking">
+<h2><a class="toc-backref" href="#id264">Type Safety Checking</a><a class="headerlink" href="#type-safety-checking" title="Permalink to this headline">¶</a></h2>
+<p>Clang supports additional attributes to enable checking type safety properties
+that canât be enforced by the C type system. To see warnings produced by these
+checks, ensure that -Wtype-safety is enabled. Use cases include:</p>
+<ul class="simple">
+<li>MPI library implementations, where these attributes enable checking that
+the buffer type matches the passed <code class="docutils literal notranslate"><span class="pre">MPI_Datatype</span></code>;</li>
+<li>for HDF5 library there is a similar use case to MPI;</li>
+<li>checking types of variadic functionsâ arguments for functions like
+<code class="docutils literal notranslate"><span class="pre">fcntl()</span></code> and <code class="docutils literal notranslate"><span class="pre">ioctl()</span></code>.</li>
+</ul>
+<p>You can detect support for these attributes with <code class="docutils literal notranslate"><span class="pre">__has_attribute()</span></code>. For
+example:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cp">#if defined(__has_attribute)</span>
+<span class="cp"># if __has_attribute(argument_with_type_tag) && \</span>
+<span class="cp"> __has_attribute(pointer_with_type_tag) && \</span>
+<span class="cp"> __has_attribute(type_tag_for_datatype)</span>
+<span class="cp"># define ATTR_MPI_PWT(buffer_idx, type_idx) __attribute__((pointer_with_type_tag(mpi,buffer_idx,type_idx)))</span>
+<span class="cm">/* ... other macros ... */</span>
+<span class="cp"># endif</span>
+<span class="cp">#endif</span>
+
+<span class="cp">#if !defined(ATTR_MPI_PWT)</span>
+<span class="cp"># define ATTR_MPI_PWT(buffer_idx, type_idx)</span>
+<span class="cp">#endif</span>
+
+<span class="kt">int</span> <span class="nf">MPI_Send</span><span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">buf</span><span class="p">,</span> <span class="kt">int</span> <span class="n">count</span><span class="p">,</span> <span class="n">MPI_Datatype</span> <span class="n">datatype</span> <span class="cm">/*, other args omitted */</span><span class="p">)</span>
+ <span class="n">ATTR_MPI_PWT</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span><span class="mi">3</span><span class="p">);</span>
+</pre></div>
+</div>
+<div class="section" id="argument-with-type-tag">
+<h3><a class="toc-backref" href="#id265">argument_with_type_tag</a><a class="headerlink" href="#argument-with-type-tag" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id124">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id124" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Use <code class="docutils literal notranslate"><span class="pre">__attribute__((argument_with_type_tag(arg_kind,</span> <span class="pre">arg_idx,</span>
+<span class="pre">type_tag_idx)))</span></code> on a function declaration to specify that the function
+accepts a type tag that determines the type of some other argument.</p>
+<p>This attribute is primarily useful for checking arguments of variadic functions
+(<code class="docutils literal notranslate"><span class="pre">pointer_with_type_tag</span></code> can be used in most non-variadic cases).</p>
+<dl class="docutils">
+<dt>In the attribute prototype above:</dt>
+<dd><ul class="first last simple">
+<li><code class="docutils literal notranslate"><span class="pre">arg_kind</span></code> is an identifier that should be used when annotating all
+applicable type tags.</li>
+<li><code class="docutils literal notranslate"><span class="pre">arg_idx</span></code> provides the position of a function argument. The expected type of
+this function argument will be determined by the function argument specified
+by <code class="docutils literal notranslate"><span class="pre">type_tag_idx</span></code>. In the code example below, â3â means that the type of the
+functionâs third argument will be determined by <code class="docutils literal notranslate"><span class="pre">type_tag_idx</span></code>.</li>
+<li><code class="docutils literal notranslate"><span class="pre">type_tag_idx</span></code> provides the position of a function argument. This function
+argument will be a type tag. The type tag will determine the expected type of
+the argument specified by <code class="docutils literal notranslate"><span class="pre">arg_idx</span></code>. In the code example below, â2â means
+that the type tag associated with the functionâs second argument should agree
+with the type of the argument specified by <code class="docutils literal notranslate"><span class="pre">arg_idx</span></code>.</li>
+</ul>
+</dd>
+</dl>
+<p>For example:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">fcntl</span><span class="p">(</span><span class="kt">int</span> <span class="n">fd</span><span class="p">,</span> <span class="kt">int</span> <span class="n">cmd</span><span class="p">,</span> <span class="p">...)</span>
+ <span class="n">__attribute__</span><span class="p">((</span> <span class="n">argument_with_type_tag</span><span class="p">(</span><span class="n">fcntl</span><span class="p">,</span><span class="mi">3</span><span class="p">,</span><span class="mi">2</span><span class="p">)</span> <span class="p">));</span>
+<span class="c1">// The function's second argument will be a type tag; this type tag will</span>
+<span class="c1">// determine the expected type of the function's third argument.</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="pointer-with-type-tag">
+<h3><a class="toc-backref" href="#id266">pointer_with_type_tag</a><a class="headerlink" href="#pointer-with-type-tag" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id125">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id125" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>Use <code class="docutils literal notranslate"><span class="pre">__attribute__((pointer_with_type_tag(ptr_kind,</span> <span class="pre">ptr_idx,</span> <span class="pre">type_tag_idx)))</span></code>
+on a function declaration to specify that the function accepts a type tag that
+determines the pointee type of some other pointer argument.</p>
+<dl class="docutils">
+<dt>In the attribute prototype above:</dt>
+<dd><ul class="first last simple">
+<li><code class="docutils literal notranslate"><span class="pre">ptr_kind</span></code> is an identifier that should be used when annotating all
+applicable type tags.</li>
+<li><code class="docutils literal notranslate"><span class="pre">ptr_idx</span></code> provides the position of a function argument; this function
+argument will have a pointer type. The expected pointee type of this pointer
+type will be determined by the function argument specified by
+<code class="docutils literal notranslate"><span class="pre">type_tag_idx</span></code>. In the code example below, â1â means that the pointee type
+of the functionâs first argument will be determined by <code class="docutils literal notranslate"><span class="pre">type_tag_idx</span></code>.</li>
+<li><code class="docutils literal notranslate"><span class="pre">type_tag_idx</span></code> provides the position of a function argument; this function
+argument will be a type tag. The type tag will determine the expected pointee
+type of the pointer argument specified by <code class="docutils literal notranslate"><span class="pre">ptr_idx</span></code>. In the code example
+below, â3â means that the type tag associated with the functionâs third
+argument should agree with the pointee type of the pointer argument specified
+by <code class="docutils literal notranslate"><span class="pre">ptr_idx</span></code>.</li>
+</ul>
+</dd>
+</dl>
+<p>For example:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="k">typedef</span> <span class="kt">int</span> <span class="n">MPI_Datatype</span><span class="p">;</span>
+<span class="kt">int</span> <span class="nf">MPI_Send</span><span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">buf</span><span class="p">,</span> <span class="kt">int</span> <span class="n">count</span><span class="p">,</span> <span class="n">MPI_Datatype</span> <span class="n">datatype</span> <span class="cm">/*, other args omitted */</span><span class="p">)</span>
+ <span class="n">__attribute__</span><span class="p">((</span> <span class="n">pointer_with_type_tag</span><span class="p">(</span><span class="n">mpi</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">3</span><span class="p">)</span> <span class="p">));</span>
+<span class="c1">// The function's 3rd argument will be a type tag; this type tag will</span>
+<span class="c1">// determine the expected pointee type of the function's 1st argument.</span>
+</pre></div>
+</div>
+</div>
+<div class="section" id="type-tag-for-datatype-clang-type-tag-for-datatype-clang-type-tag-for-datatype">
+<h3><a class="toc-backref" href="#id267">type_tag_for_datatype (clang::type_tag_for_datatype, clang::type_tag_for_datatype)</a><a class="headerlink" href="#type-tag-for-datatype-clang-type-tag-for-datatype-clang-type-tag-for-datatype" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id126">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id126" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>When declaring a variable, use
+<code class="docutils literal notranslate"><span class="pre">__attribute__((type_tag_for_datatype(kind,</span> <span class="pre">type)))</span></code> to create a type tag that
+is tied to the <code class="docutils literal notranslate"><span class="pre">type</span></code> argument given to the attribute.</p>
+<dl class="docutils">
+<dt>In the attribute prototype above:</dt>
+<dd><ul class="first last simple">
+<li><code class="docutils literal notranslate"><span class="pre">kind</span></code> is an identifier that should be used when annotating all applicable
+type tags.</li>
+<li><code class="docutils literal notranslate"><span class="pre">type</span></code> indicates the name of the type.</li>
+</ul>
+</dd>
+</dl>
+<p>Clang supports annotating type tags of two forms.</p>
+<blockquote>
+<div><ul>
+<li><p class="first"><strong>Type tag that is a reference to a declared identifier.</strong>
+Use <code class="docutils literal notranslate"><span class="pre">__attribute__((type_tag_for_datatype(kind,</span> <span class="pre">type)))</span></code> when declaring that
+identifier:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="k">typedef</span> <span class="kt">int</span> <span class="n">MPI_Datatype</span><span class="p">;</span>
+<span class="k">extern</span> <span class="k">struct</span> <span class="n">mpi_datatype</span> <span class="n">mpi_datatype_int</span>
+ <span class="nf">__attribute__</span><span class="p">((</span> <span class="n">type_tag_for_datatype</span><span class="p">(</span><span class="n">mpi</span><span class="p">,</span><span class="kt">int</span><span class="p">)</span> <span class="p">));</span>
+<span class="cp">#define MPI_INT ((MPI_Datatype) &mpi_datatype_int)</span>
+<span class="c1">// &mpi_datatype_int is a type tag. It is tied to type "int".</span>
+</pre></div>
+</div>
+</li>
+<li><p class="first"><strong>Type tag that is an integral literal.</strong>
+Declare a <code class="docutils literal notranslate"><span class="pre">static</span> <span class="pre">const</span></code> variable with an initializer value and attach
+<code class="docutils literal notranslate"><span class="pre">__attribute__((type_tag_for_datatype(kind,</span> <span class="pre">type)))</span></code> on that declaration:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="k">typedef</span> <span class="kt">int</span> <span class="n">MPI_Datatype</span><span class="p">;</span>
+<span class="k">static</span> <span class="k">const</span> <span class="n">MPI_Datatype</span> <span class="n">mpi_datatype_int</span>
+ <span class="nf">__attribute__</span><span class="p">((</span> <span class="n">type_tag_for_datatype</span><span class="p">(</span><span class="n">mpi</span><span class="p">,</span><span class="kt">int</span><span class="p">)</span> <span class="p">))</span> <span class="o">=</span> <span class="mi">42</span><span class="p">;</span>
+<span class="cp">#define MPI_INT ((MPI_Datatype) 42)</span>
+<span class="c1">// The number 42 is a type tag. It is tied to type "int".</span>
+</pre></div>
+</div>
+</li>
+</ul>
+</div></blockquote>
+<p>The <code class="docutils literal notranslate"><span class="pre">type_tag_for_datatype</span></code> attribute also accepts an optional third argument
+that determines how the type of the function argument specified by either
+<code class="docutils literal notranslate"><span class="pre">arg_idx</span></code> or <code class="docutils literal notranslate"><span class="pre">ptr_idx</span></code> is compared against the type associated with the type
+tag. (Recall that for the <code class="docutils literal notranslate"><span class="pre">argument_with_type_tag</span></code> attribute, the type of the
+function argument specified by <code class="docutils literal notranslate"><span class="pre">arg_idx</span></code> is compared against the type
+associated with the type tag. Also recall that for the <code class="docutils literal notranslate"><span class="pre">pointer_with_type_tag</span></code>
+attribute, the pointee type of the function argument specified by <code class="docutils literal notranslate"><span class="pre">ptr_idx</span></code> is
+compared against the type associated with the type tag.) There are two supported
+values for this optional third argument:</p>
+<blockquote>
+<div><ul>
+<li><p class="first"><code class="docutils literal notranslate"><span class="pre">layout_compatible</span></code> will cause types to be compared according to
+layout-compatibility rules (In C++11 [class.mem] p 17, 18, see the
+layout-compatibility rules for two standard-layout struct types and for two
+standard-layout union types). This is useful when creating a type tag
+associated with a struct or union type. For example:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cm">/* In mpi.h */</span>
+<span class="k">typedef</span> <span class="kt">int</span> <span class="n">MPI_Datatype</span><span class="p">;</span>
+<span class="k">struct</span> <span class="n">internal_mpi_double_int</span> <span class="p">{</span> <span class="kt">double</span> <span class="n">d</span><span class="p">;</span> <span class="kt">int</span> <span class="n">i</span><span class="p">;</span> <span class="p">};</span>
+<span class="k">extern</span> <span class="k">struct</span> <span class="n">mpi_datatype</span> <span class="n">mpi_datatype_double_int</span>
+ <span class="nf">__attribute__</span><span class="p">((</span> <span class="n">type_tag_for_datatype</span><span class="p">(</span><span class="n">mpi</span><span class="p">,</span>
+ <span class="k">struct</span> <span class="n">internal_mpi_double_int</span><span class="p">,</span> <span class="n">layout_compatible</span><span class="p">)</span> <span class="p">));</span>
+
+<span class="cp">#define MPI_DOUBLE_INT ((MPI_Datatype) &mpi_datatype_double_int)</span>
+
+<span class="kt">int</span> <span class="nf">MPI_Send</span><span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">buf</span><span class="p">,</span> <span class="kt">int</span> <span class="n">count</span><span class="p">,</span> <span class="n">MPI_Datatype</span> <span class="n">datatype</span><span class="p">,</span> <span class="p">...)</span>
+ <span class="n">__attribute__</span><span class="p">((</span> <span class="n">pointer_with_type_tag</span><span class="p">(</span><span class="n">mpi</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">3</span><span class="p">)</span> <span class="p">));</span>
+
+<span class="cm">/* In user code */</span>
+<span class="k">struct</span> <span class="n">my_pair</span> <span class="p">{</span> <span class="kt">double</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="p">};</span>
+<span class="k">struct</span> <span class="n">my_pair</span> <span class="o">*</span><span class="n">buffer</span><span class="p">;</span>
+<span class="n">MPI_Send</span><span class="p">(</span><span class="n">buffer</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">MPI_DOUBLE_INT</span> <span class="cm">/*, ... */</span><span class="p">);</span> <span class="c1">// no warning because the</span>
+ <span class="c1">// layout of my_pair is</span>
+ <span class="c1">// compatible with that of</span>
+ <span class="c1">// internal_mpi_double_int</span>
+
+<span class="k">struct</span> <span class="n">my_int_pair</span> <span class="p">{</span> <span class="kt">int</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="p">}</span>
+<span class="k">struct</span> <span class="n">my_int_pair</span> <span class="o">*</span><span class="n">buffer2</span><span class="p">;</span>
+<span class="n">MPI_Send</span><span class="p">(</span><span class="n">buffer2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">MPI_DOUBLE_INT</span> <span class="cm">/*, ... */</span><span class="p">);</span> <span class="c1">// warning because the</span>
+ <span class="c1">// layout of my_int_pair</span>
+ <span class="c1">// does not match that of</span>
+ <span class="c1">// internal_mpi_double_int</span>
+</pre></div>
+</div>
+</li>
+<li><p class="first"><code class="docutils literal notranslate"><span class="pre">must_be_null</span></code> specifies that the function argument specified by either
+<code class="docutils literal notranslate"><span class="pre">arg_idx</span></code> (for the <code class="docutils literal notranslate"><span class="pre">argument_with_type_tag</span></code> attribute) or <code class="docutils literal notranslate"><span class="pre">ptr_idx</span></code> (for
+the <code class="docutils literal notranslate"><span class="pre">pointer_with_type_tag</span></code> attribute) should be a null pointer constant.
+The second argument to the <code class="docutils literal notranslate"><span class="pre">type_tag_for_datatype</span></code> attribute is ignored. For
+example:</p>
+<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="cm">/* In mpi.h */</span>
+<span class="k">typedef</span> <span class="kt">int</span> <span class="n">MPI_Datatype</span><span class="p">;</span>
+<span class="k">extern</span> <span class="k">struct</span> <span class="n">mpi_datatype</span> <span class="n">mpi_datatype_null</span>
+ <span class="nf">__attribute__</span><span class="p">((</span> <span class="n">type_tag_for_datatype</span><span class="p">(</span><span class="n">mpi</span><span class="p">,</span> <span class="kt">void</span><span class="p">,</span> <span class="n">must_be_null</span><span class="p">)</span> <span class="p">));</span>
+
+<span class="cp">#define MPI_DATATYPE_NULL ((MPI_Datatype) &mpi_datatype_null)</span>
+<span class="kt">int</span> <span class="nf">MPI_Send</span><span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">buf</span><span class="p">,</span> <span class="kt">int</span> <span class="n">count</span><span class="p">,</span> <span class="n">MPI_Datatype</span> <span class="n">datatype</span><span class="p">,</span> <span class="p">...)</span>
+ <span class="n">__attribute__</span><span class="p">((</span> <span class="n">pointer_with_type_tag</span><span class="p">(</span><span class="n">mpi</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">3</span><span class="p">)</span> <span class="p">));</span>
+
+<span class="cm">/* In user code */</span>
+<span class="k">struct</span> <span class="n">my_pair</span> <span class="p">{</span> <span class="kt">double</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="p">};</span>
+<span class="k">struct</span> <span class="n">my_pair</span> <span class="o">*</span><span class="n">buffer</span><span class="p">;</span>
+<span class="n">MPI_Send</span><span class="p">(</span><span class="n">buffer</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="n">MPI_DATATYPE_NULL</span> <span class="cm">/*, ... */</span><span class="p">);</span> <span class="c1">// warning: MPI_DATATYPE_NULL</span>
+ <span class="c1">// was specified but buffer</span>
+ <span class="c1">// is not a null pointer</span>
+</pre></div>
+</div>
+</li>
+</ul>
+</div></blockquote>
+</div>
+</div>
+<div class="section" id="opencl-address-spaces">
+<h2><a class="toc-backref" href="#id268">OpenCL Address Spaces</a><a class="headerlink" href="#opencl-address-spaces" title="Permalink to this headline">¶</a></h2>
+<p>The address space qualifier may be used to specify the region of memory that is
+used to allocate the object. OpenCL supports the following address spaces:
+__generic(generic), __global(global), __local(local), __private(private),
+__constant(constant).</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="n">__constant</span> <span class="kt">int</span> <span class="n">c</span> <span class="o">=</span> <span class="p">...;</span>
+
+<span class="n">__generic</span> <span class="kt">int</span><span class="o">*</span> <span class="nf">foo</span><span class="p">(</span><span class="n">global</span> <span class="kt">int</span><span class="o">*</span> <span class="n">g</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">__local</span> <span class="kt">int</span><span class="o">*</span> <span class="n">l</span><span class="p">;</span>
+ <span class="n">private</span> <span class="kt">int</span> <span class="n">p</span><span class="p">;</span>
+ <span class="p">...</span>
+ <span class="k">return</span> <span class="n">l</span><span class="p">;</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>More details can be found in the OpenCL C language Spec v2.0, Section 6.5.</p>
+<div class="section" id="constant-constant">
+<h3><a class="toc-backref" href="#id269">constant (__constant)</a><a class="headerlink" href="#constant-constant" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id127">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id127" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The constant address space attribute signals that an object is located in
+a constant (non-modifiable) memory region. It is available to all work items.
+Any type can be annotated with the constant address space attribute. Objects
+with the constant address space qualifier can be declared in any scope and must
+have an initializer.</p>
+</div>
+<div class="section" id="generic-generic">
+<h3><a class="toc-backref" href="#id270">generic (__generic)</a><a class="headerlink" href="#generic-generic" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id128">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id128" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The generic address space attribute is only available with OpenCL v2.0 and later.
+It can be used with pointer types. Variables in global and local scope and
+function parameters in non-kernel functions can have the generic address space
+type attribute. It is intended to be a placeholder for any other address space
+except for â__constantâ in OpenCL code which can be used with multiple address
+spaces.</p>
+</div>
+<div class="section" id="global-global">
+<h3><a class="toc-backref" href="#id271">global (__global)</a><a class="headerlink" href="#global-global" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id129">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id129" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The global address space attribute specifies that an object is allocated in
+global memory, which is accessible by all work items. The content stored in this
+memory area persists between kernel executions. Pointer types to the global
+address space are allowed as function parameters or local variables. Starting
+with OpenCL v2.0, the global address space can be used with global (program
+scope) variables and static local variable as well.</p>
+</div>
+<div class="section" id="local-local">
+<h3><a class="toc-backref" href="#id272">local (__local)</a><a class="headerlink" href="#local-local" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id130">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id130" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The local address space specifies that an object is allocated in the local (work
+group) memory area, which is accessible to all work items in the same work
+group. The content stored in this memory region is not accessible after
+the kernel execution ends. In a kernel function scope, any variable can be in
+the local address space. In other scopes, only pointer types to the local address
+space are allowed. Local address space variables cannot have an initializer.</p>
+</div>
+<div class="section" id="private-private">
+<h3><a class="toc-backref" href="#id273">private (__private)</a><a class="headerlink" href="#private-private" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id131">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id131" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The private address space specifies that an object is allocated in the private
+(work item) memory. Other work items cannot access the same memory area and its
+content is destroyed after work item execution ends. Local variables can be
+declared in the private address space. Function arguments are always in the
+private address space. Kernel function arguments of a pointer or an array type
+cannot point to the private address space.</p>
+</div>
+</div>
+<div class="section" id="nullability-attributes">
+<h2><a class="toc-backref" href="#id274">Nullability Attributes</a><a class="headerlink" href="#nullability-attributes" title="Permalink to this headline">¶</a></h2>
+<p>Whether a particular pointer may be ânullâ is an important concern when working with pointers in the C family of languages. The various nullability attributes indicate whether a particular pointer can be null or not, which makes APIs more expressive and can help static analysis tools identify bugs involving null pointers. Clang supports several kinds of nullability attributes: the <code class="docutils literal notranslate"><span class="pre">nonnull</span></code> and <code class="docutils literal notranslate"><span class="pre">returns_nonnull</span></code> attributes indicate which function or method parameters and result types can never be null, while nullability type qualifiers indicate which pointer types can be null (<code class="docutils literal notranslate"><span class="pre">_Nullable</span></code>) or cannot be null (<code class="docutils literal notranslate"><span class="pre">_Nonnull</span></code>).</p>
+<p>The nullability (type) qualifiers express whether a value of a given pointer type can be null (the <code class="docutils literal notranslate"><span class="pre">_Nullable</span></code> qualifier), doesnât have a defined meaning for null (the <code class="docutils literal notranslate"><span class="pre">_Nonnull</span></code> qualifier), or for which the purpose of null is unclear (the <code class="docutils literal notranslate"><span class="pre">_Null_unspecified</span></code> qualifier). Because nullability qualifiers are expressed within the type system, they are more general than the <code class="docutils literal notranslate"><span class="pre">nonnull</span></code> and <code class="docutils literal notranslate"><span class="pre">returns_nonnull</span></code> attributes, allowing one to express (for example) a nullable pointer to an array of nonnull pointers. Nullability qualifiers are written to the right of the pointer to which they apply. For example:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="c1">// No meaningful result when 'ptr' is null (here, it happens to be undefined behavior).</span>
+<span class="kt">int</span> <span class="nf">fetch</span><span class="p">(</span><span class="kt">int</span> <span class="o">*</span> <span class="n">_Nonnull</span> <span class="n">ptr</span><span class="p">)</span> <span class="p">{</span> <span class="k">return</span> <span class="o">*</span><span class="n">ptr</span><span class="p">;</span> <span class="p">}</span>
+
+<span class="c1">// 'ptr' may be null.</span>
+<span class="kt">int</span> <span class="nf">fetch_or_zero</span><span class="p">(</span><span class="kt">int</span> <span class="o">*</span> <span class="n">_Nullable</span> <span class="n">ptr</span><span class="p">)</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="n">ptr</span> <span class="o">?</span> <span class="o">*</span><span class="nl">ptr</span> <span class="p">:</span> <span class="mi">0</span><span class="p">;</span>
+<span class="p">}</span>
+
+<span class="c1">// A nullable pointer to non-null pointers to const characters.</span>
+<span class="k">const</span> <span class="kt">char</span> <span class="o">*</span><span class="nf">join_strings</span><span class="p">(</span><span class="k">const</span> <span class="kt">char</span> <span class="o">*</span> <span class="n">_Nonnull</span> <span class="o">*</span> <span class="n">_Nullable</span> <span class="n">strings</span><span class="p">,</span> <span class="kt">unsigned</span> <span class="n">n</span><span class="p">);</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>In Objective-C, there is an alternate spelling for the nullability qualifiers that can be used in Objective-C methods and properties using context-sensitive, non-underscored keywords. For example:</p>
+<blockquote>
+<div><div class="highlight-objective-c notranslate"><div class="highlight"><pre><span></span><span class="k">@interface</span> <span class="nc">NSView</span> : <span class="nc">NSResponder</span>
+ <span class="o">-</span> <span class="p">(</span><span class="n">nullable</span> <span class="n">NSView</span> <span class="o">*</span><span class="p">)</span><span class="nl">ancestorSharedWithView</span><span class="p">:(</span><span class="n">nonnull</span> <span class="n">NSView</span> <span class="o">*</span><span class="p">)</span><span class="n">aView</span><span class="p">;</span>
+ <span class="k">@property</span> <span class="p">(</span><span class="k">assign</span><span class="p">,</span> <span class="n">nullable</span><span class="p">)</span> <span class="n">NSView</span> <span class="o">*</span><span class="n">superview</span><span class="p">;</span>
+ <span class="k">@property</span> <span class="p">(</span><span class="k">readonly</span><span class="p">,</span> <span class="n">nonnull</span><span class="p">)</span> <span class="bp">NSArray</span> <span class="o">*</span><span class="n">subviews</span><span class="p">;</span>
+<span class="k">@end</span>
+</pre></div>
+</div>
+</div></blockquote>
+<div class="section" id="nonnull">
+<h3><a class="toc-backref" href="#id275">_Nonnull</a><a class="headerlink" href="#nonnull" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id132">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id132" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">_Nonnull</span></code> nullability qualifier indicates that null is not a meaningful value for a value of the <code class="docutils literal notranslate"><span class="pre">_Nonnull</span></code> pointer type. For example, given a declaration such as:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">fetch</span><span class="p">(</span><span class="kt">int</span> <span class="o">*</span> <span class="n">_Nonnull</span> <span class="n">ptr</span><span class="p">);</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>a caller of <code class="docutils literal notranslate"><span class="pre">fetch</span></code> should not provide a null value, and the compiler will produce a warning if it sees a literal null value passed to <code class="docutils literal notranslate"><span class="pre">fetch</span></code>. Note that, unlike the declaration attribute <code class="docutils literal notranslate"><span class="pre">nonnull</span></code>, the presence of <code class="docutils literal notranslate"><span class="pre">_Nonnull</span></code> does not imply that passing null is undefined behavior: <code class="docutils literal notranslate"><span class="pre">fetch</span></code> is free to consider null undefined behavior or (perhaps for backward-compatibility reasons) defensively handle null.</p>
+</div>
+<div class="section" id="null-unspecified">
+<h3><a class="toc-backref" href="#id276">_Null_unspecified</a><a class="headerlink" href="#null-unspecified" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id133">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id133" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">_Null_unspecified</span></code> nullability qualifier indicates that neither the <code class="docutils literal notranslate"><span class="pre">_Nonnull</span></code> nor <code class="docutils literal notranslate"><span class="pre">_Nullable</span></code> qualifiers make sense for a particular pointer type. It is used primarily to indicate that the role of null with specific pointers in a nullability-annotated header is unclear, e.g., due to overly-complex implementations or historical factors with a long-lived API.</p>
+</div>
+<div class="section" id="nullable">
+<h3><a class="toc-backref" href="#id277">_Nullable</a><a class="headerlink" href="#nullable" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id134">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id134" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">_Nullable</span></code> nullability qualifier indicates that a value of the <code class="docutils literal notranslate"><span class="pre">_Nullable</span></code> pointer type can be null. For example, given:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="kt">int</span> <span class="nf">fetch_or_zero</span><span class="p">(</span><span class="kt">int</span> <span class="o">*</span> <span class="n">_Nullable</span> <span class="n">ptr</span><span class="p">);</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>a caller of <code class="docutils literal notranslate"><span class="pre">fetch_or_zero</span></code> can provide null.</p>
+</div>
+<div class="section" id="nonnull-gnu-nonnull">
+<h3><a class="toc-backref" href="#id278">nonnull (gnu::nonnull)</a><a class="headerlink" href="#nonnull-gnu-nonnull" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id135">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id135" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">nonnull</span></code> attribute indicates that some function parameters must not be null, and can be used in several different ways. Itâs original usage (<a class="reference external" href="https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#Common-Function-Attributes">from GCC</a>) is as a function (or Objective-C method) attribute that specifies which parameters of the function are nonnull in a comma-separated list. For example:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">extern</span> <span class="kt">void</span> <span class="o">*</span> <span class="nf">my_memcpy</span> <span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">dest</span><span class="p">,</span> <span class="k">const</span> <span class="kt">void</span> <span class="o">*</span><span class="n">src</span><span class="p">,</span> <span class="kt">size_t</span> <span class="n">len</span><span class="p">)</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">nonnull</span> <span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">)));</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>Here, the <code class="docutils literal notranslate"><span class="pre">nonnull</span></code> attribute indicates that parameters 1 and 2
+cannot have a null value. Omitting the parenthesized list of parameter indices means that all parameters of pointer type cannot be null:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">extern</span> <span class="kt">void</span> <span class="o">*</span> <span class="nf">my_memcpy</span> <span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">dest</span><span class="p">,</span> <span class="k">const</span> <span class="kt">void</span> <span class="o">*</span><span class="n">src</span><span class="p">,</span> <span class="kt">size_t</span> <span class="n">len</span><span class="p">)</span>
+ <span class="n">__attribute__</span><span class="p">((</span><span class="n">nonnull</span><span class="p">));</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>Clang also allows the <code class="docutils literal notranslate"><span class="pre">nonnull</span></code> attribute to be placed directly on a function (or Objective-C method) parameter, eliminating the need to specify the parameter index ahead of type. For example:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">extern</span> <span class="kt">void</span> <span class="o">*</span> <span class="nf">my_memcpy</span> <span class="p">(</span><span class="kt">void</span> <span class="o">*</span><span class="n">dest</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">nonnull</span><span class="p">)),</span>
+ <span class="k">const</span> <span class="kt">void</span> <span class="o">*</span><span class="n">src</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">nonnull</span><span class="p">)),</span> <span class="kt">size_t</span> <span class="n">len</span><span class="p">);</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>Note that the <code class="docutils literal notranslate"><span class="pre">nonnull</span></code> attribute indicates that passing null to a non-null parameter is undefined behavior, which the optimizer may take advantage of to, e.g., remove null checks. The <code class="docutils literal notranslate"><span class="pre">_Nonnull</span></code> type qualifier indicates that a pointer cannot be null in a more general manner (because it is part of the type system) and does not imply undefined behavior, making it more widely applicable.</p>
+</div>
+<div class="section" id="returns-nonnull-gnu-returns-nonnull">
+<h3><a class="toc-backref" href="#id279">returns_nonnull (gnu::returns_nonnull)</a><a class="headerlink" href="#returns-nonnull-gnu-returns-nonnull" title="Permalink to this headline">¶</a></h3>
+<table border="1" class="docutils" id="id136">
+<caption><span class="caption-text">Supported Syntaxes</span><a class="headerlink" href="#id136" title="Permalink to this table">¶</a></caption>
+<colgroup>
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+<col width="14%" />
+</colgroup>
+<thead valign="bottom">
+<tr class="row-odd"><th class="head">GNU</th>
+<th class="head">C++11</th>
+<th class="head">C2x</th>
+<th class="head">__declspec</th>
+<th class="head">Keyword</th>
+<th class="head">Pragma</th>
+<th class="head">Pragma clang attribute</th>
+</tr>
+</thead>
+<tbody valign="top">
+<tr class="row-even"><td>X</td>
+<td>X</td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td> </td>
+<td>X</td>
+</tr>
+</tbody>
+</table>
+<p>The <code class="docutils literal notranslate"><span class="pre">returns_nonnull</span></code> attribute indicates that a particular function (or Objective-C method) always returns a non-null pointer. For example, a particular system <code class="docutils literal notranslate"><span class="pre">malloc</span></code> might be defined to terminate a process when memory is not available rather than returning a null pointer:</p>
+<blockquote>
+<div><div class="highlight-c notranslate"><div class="highlight"><pre><span></span><span class="k">extern</span> <span class="kt">void</span> <span class="o">*</span> <span class="nf">malloc</span> <span class="p">(</span><span class="kt">size_t</span> <span class="n">size</span><span class="p">)</span> <span class="n">__attribute__</span><span class="p">((</span><span class="n">returns_nonnull</span><span class="p">));</span>
+</pre></div>
+</div>
+</div></blockquote>
+<p>The <code class="docutils literal notranslate"><span class="pre">returns_nonnull</span></code> attribute implies that returning a null pointer is undefined behavior, which the optimizer may take advantage of. The <code class="docutils literal notranslate"><span class="pre">_Nonnull</span></code> type qualifier indicates that a pointer cannot be null in a more general manner (because it is part of the type system) and does not imply undefined behavior, making it more widely applicable</p>
+</div>
+</div>
+</div>
+
+
+ </div>
+ <div class="bottomnav" role="navigation" aria-label="bottom navigation">
+
+ <p>
+ « <a href="ClangCommandLineReference.html">Clang command line argument reference</a>
+ ::
+ <a class="uplink" href="index.html">Contents</a>
+ ::
+ <a href="DiagnosticsReference.html">Diagnostic flags in Clang</a> »
+ </p>
+
+ </div>
+
+ <div class="footer" role="contentinfo">
+ © Copyright 2007-2018, The Clang Team.
+ Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.7.6.
+ </div>
+ </body>
+</html>
\ No newline at end of file
Added: www-releases/trunk/7.0.1/tools/clang/docs/AutomaticReferenceCounting.html
URL: http://llvm.org/viewvc/llvm-project/www-releases/trunk/7.0.1/tools/clang/docs/AutomaticReferenceCounting.html?rev=349965&view=auto
==============================================================================
--- www-releases/trunk/7.0.1/tools/clang/docs/AutomaticReferenceCounting.html (added)
+++ www-releases/trunk/7.0.1/tools/clang/docs/AutomaticReferenceCounting.html Fri Dec 21 13:53:02 2018
@@ -0,0 +1,2110 @@
+
+<!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="X-UA-Compatible" content="IE=Edge" />
+ <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
+ <title>Objective-C Automatic Reference Counting (ARC) — Clang 7 documentation</title>
+ <link rel="stylesheet" href="_static/haiku.css" type="text/css" />
+ <link rel="stylesheet" href="_static/pygments.css" type="text/css" />
+ <script type="text/javascript" id="documentation_options" data-url_root="./" src="_static/documentation_options.js"></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>
+ <script type="text/javascript" src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.1/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
+ <link rel="index" title="Index" href="genindex.html" />
+ <link rel="search" title="Search" href="search.html" />
+ <link rel="next" title="Clang command line argument reference" href="ClangCommandLineReference.html" />
+ <link rel="prev" title="Block Implementation Specification" href="Block-ABI-Apple.html" />
+ </head><body>
+ <div class="header" role="banner"><h1 class="heading"><a href="index.html">
+ <span>Clang 7 documentation</span></a></h1>
+ <h2 class="heading"><span>Objective-C Automatic Reference Counting (ARC)</span></h2>
+ </div>
+ <div class="topnav" role="navigation" aria-label="top navigation">
+
+ <p>
+ « <a href="Block-ABI-Apple.html">Block Implementation Specification</a>
+ ::
+ <a class="uplink" href="index.html">Contents</a>
+ ::
+ <a href="ClangCommandLineReference.html">Clang command line argument reference</a> »
+ </p>
+
+ </div>
+ <div class="content">
+
+
+ <style>
+ .arc-term { font-style: italic; font-weight: bold; }
+ .revision { font-style: italic; }
+ .when-revised { font-weight: bold; font-style: normal; }
+
+ /*
+ * Automatic numbering is described in this article:
+ * http://dev.opera.com/articles/view/automatic-numbering-with-css-counters/
+ */
+ /*
+ * Automatic numbering for the TOC.
+ * This is wrong from the semantics point of view, since it is an ordered
+ * list, but uses "ul" tag.
+ */
+ div#contents.contents.local ul {
+ counter-reset: toc-section;
+ list-style-type: none;
+ }
+ div#contents.contents.local ul li {
+ counter-increment: toc-section;
+ background: none; // Remove bullets
+ }
+ div#contents.contents.local ul li a.reference:before {
+ content: counters(toc-section, ".") " ";
+ }
+
+ /* Automatic numbering for the body. */
+ body {
+ counter-reset: section subsection subsubsection;
+ }
+ .section h2 {
+ counter-reset: subsection subsubsection;
+ counter-increment: section;
+ }
+ .section h2 a.toc-backref:before {
+ content: counter(section) " ";
+ }
+ .section h3 {
+ counter-reset: subsubsection;
+ counter-increment: subsection;
+ }
+ .section h3 a.toc-backref:before {
+ content: counter(section) "." counter(subsection) " ";
+ }
+ .section h4 {
+ counter-increment: subsubsection;
+ }
+ .section h4 a.toc-backref:before {
+ content: counter(section) "." counter(subsection) "." counter(subsubsection) " ";
+ }
+</style><div class="section" id="objective-c-automatic-reference-counting-arc">
+<h1>Objective-C Automatic Reference Counting (ARC)<a class="headerlink" href="#objective-c-automatic-reference-counting-arc" title="Permalink to this headline">¶</a></h1>
+<div class="contents local topic" id="contents">
+<ul class="simple">
+<li><a class="reference internal" href="#about-this-document" id="id4">About this document</a><ul>
+<li><a class="reference internal" href="#purpose" id="id5">Purpose</a></li>
+<li><a class="reference internal" href="#background" id="id6">Background</a></li>
+<li><a class="reference internal" href="#evolution" id="id7">Evolution</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#general" id="id8">General</a></li>
+<li><a class="reference internal" href="#retainable-object-pointers" id="id9">Retainable object pointers</a><ul>
+<li><a class="reference internal" href="#retain-count-semantics" id="id10">Retain count semantics</a></li>
+<li><a class="reference internal" href="#retainable-object-pointers-as-operands-and-arguments" id="id11">Retainable object pointers as operands and arguments</a><ul>
+<li><a class="reference internal" href="#consumed-parameters" id="id12">Consumed parameters</a></li>
+<li><a class="reference internal" href="#retained-return-values" id="id13">Retained return values</a></li>
+<li><a class="reference internal" href="#unretained-return-values" id="id14">Unretained return values</a></li>
+<li><a class="reference internal" href="#bridged-casts" id="id15">Bridged casts</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#restrictions" id="id16">Restrictions</a><ul>
+<li><a class="reference internal" href="#conversion-of-retainable-object-pointers" id="id17">Conversion of retainable object pointers</a></li>
+<li><a class="reference internal" href="#conversion-to-retainable-object-pointer-type-of-expressions-with-known-semantics" id="id18">Conversion to retainable object pointer type of expressions with known semantics</a></li>
+<li><a class="reference internal" href="#conversion-from-retainable-object-pointer-type-in-certain-contexts" id="id19">Conversion from retainable object pointer type in certain contexts</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#ownership-qualification" id="id20">Ownership qualification</a><ul>
+<li><a class="reference internal" href="#spelling" id="id21">Spelling</a><ul>
+<li><a class="reference internal" href="#property-declarations" id="id22">Property declarations</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#semantics" id="id23">Semantics</a></li>
+<li><a class="reference internal" href="#arc-ownership-restrictions" id="id24">Restrictions</a><ul>
+<li><a class="reference internal" href="#weak-unavailable-types" id="id25">Weak-unavailable types</a></li>
+<li><a class="reference internal" href="#storage-duration-of-autoreleasing-objects" id="id26">Storage duration of <code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> objects</a></li>
+<li><a class="reference internal" href="#conversion-of-pointers-to-ownership-qualified-types" id="id27">Conversion of pointers to ownership-qualified types</a></li>
+<li><a class="reference internal" href="#passing-to-an-out-parameter-by-writeback" id="id28">Passing to an out parameter by writeback</a></li>
+<li><a class="reference internal" href="#ownership-qualified-fields-of-structs-and-unions" id="id29">Ownership-qualified fields of structs and unions</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#ownership-inference" id="id30">Ownership inference</a><ul>
+<li><a class="reference internal" href="#objects" id="id31">Objects</a></li>
+<li><a class="reference internal" href="#indirect-parameters" id="id32">Indirect parameters</a></li>
+<li><a class="reference internal" href="#template-arguments" id="id33">Template arguments</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#method-families" id="id34">Method families</a><ul>
+<li><a class="reference internal" href="#explicit-method-family-control" id="id35">Explicit method family control</a></li>
+<li><a class="reference internal" href="#semantics-of-method-families" id="id36">Semantics of method families</a><ul>
+<li><a class="reference internal" href="#semantics-of-init" id="id37">Semantics of <code class="docutils literal notranslate"><span class="pre">init</span></code></a></li>
+<li><a class="reference internal" href="#related-result-types" id="id38">Related result types</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#optimization" id="id39">Optimization</a><ul>
+<li><a class="reference internal" href="#object-liveness" id="id40">Object liveness</a></li>
+<li><a class="reference internal" href="#no-object-lifetime-extension" id="id41">No object lifetime extension</a></li>
+<li><a class="reference internal" href="#precise-lifetime-semantics" id="id42">Precise lifetime semantics</a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#miscellaneous" id="id43">Miscellaneous</a><ul>
+<li><a class="reference internal" href="#special-methods" id="id44">Special methods</a><ul>
+<li><a class="reference internal" href="#memory-management-methods" id="id45">Memory management methods</a></li>
+<li><a class="reference internal" href="#dealloc" id="id46"><code class="docutils literal notranslate"><span class="pre">dealloc</span></code></a></li>
+</ul>
+</li>
+<li><a class="reference internal" href="#autoreleasepool" id="id47"><code class="docutils literal notranslate"><span class="pre">@autoreleasepool</span></code></a></li>
+<li><a class="reference internal" href="#self" id="id48"><code class="docutils literal notranslate"><span class="pre">self</span></code></a></li>
+<li><a class="reference internal" href="#fast-enumeration-iteration-variables" id="id49">Fast enumeration iteration variables</a></li>
+<li><a class="reference internal" href="#blocks" id="id50">Blocks</a></li>
+<li><a class="reference internal" href="#exceptions" id="id51">Exceptions</a></li>
+<li><a class="reference internal" href="#interior-pointers" id="id52">Interior pointers</a></li>
+<li><a class="reference internal" href="#c-retainable-pointer-types" id="id53">C retainable pointer types</a><ul>
+<li><a class="reference internal" href="#auditing-of-c-retainable-pointer-interfaces" id="id54">Auditing of C retainable pointer interfaces</a></li>
+</ul>
+</li>
+</ul>
+</li>
+<li><a class="reference internal" href="#runtime-support" id="id55">Runtime support</a><ul>
+<li><a class="reference internal" href="#arc-runtime-objc-autorelease" id="id56"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_autorelease(id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#void-objc-autoreleasepoolpop-void-pool" id="id57"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_autoreleasePoolPop(void</span> <span class="pre">*pool);</span></code></a></li>
+<li><a class="reference internal" href="#void-objc-autoreleasepoolpush-void" id="id58"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">*objc_autoreleasePoolPush(void);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-autoreleasereturnvalue" id="id59"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_autoreleaseReturnValue(id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#void-objc-copyweak-id-dest-id-src" id="id60"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_copyWeak(id</span> <span class="pre">*dest,</span> <span class="pre">id</span> <span class="pre">*src);</span></code></a></li>
+<li><a class="reference internal" href="#void-objc-destroyweak-id-object" id="id61"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_destroyWeak(id</span> <span class="pre">*object);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-initweak" id="id62"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_initWeak(id</span> <span class="pre">*object,</span> <span class="pre">id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-loadweak" id="id63"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_loadWeak(id</span> <span class="pre">*object);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-loadweakretained" id="id64"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_loadWeakRetained(id</span> <span class="pre">*object);</span></code></a></li>
+<li><a class="reference internal" href="#void-objc-moveweak-id-dest-id-src" id="id65"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_moveWeak(id</span> <span class="pre">*dest,</span> <span class="pre">id</span> <span class="pre">*src);</span></code></a></li>
+<li><a class="reference internal" href="#void-objc-release-id-value" id="id66"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_release(id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-retain" id="id67"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retain(id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-retainautorelease" id="id68"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retainAutorelease(id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-retainautoreleasereturnvalue" id="id69"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retainAutoreleaseReturnValue(id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-retainautoreleasedreturnvalue" id="id70"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retainAutoreleasedReturnValue(id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-retainblock" id="id71"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retainBlock(id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-storestrong" id="id72"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_storeStrong(id</span> <span class="pre">*object,</span> <span class="pre">id</span> <span class="pre">value);</span></code></a></li>
+<li><a class="reference internal" href="#arc-runtime-objc-storeweak" id="id73"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_storeWeak(id</span> <span class="pre">*object,</span> <span class="pre">id</span> <span class="pre">value);</span></code></a></li>
+</ul>
+</li>
+</ul>
+</div>
+<div class="section" id="about-this-document">
+<span id="arc-meta"></span><h2><a class="toc-backref" href="#id4">About this document</a><a class="headerlink" href="#about-this-document" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="purpose">
+<span id="arc-meta-purpose"></span><h3><a class="toc-backref" href="#id5">Purpose</a><a class="headerlink" href="#purpose" title="Permalink to this headline">¶</a></h3>
+<p>The first and primary purpose of this document is to serve as a complete
+technical specification of Automatic Reference Counting. Given a core
+Objective-C compiler and runtime, it should be possible to write a compiler and
+runtime which implements these new semantics.</p>
+<p>The secondary purpose is to act as a rationale for why ARC was designed in this
+way. This should remain tightly focused on the technical design and should not
+stray into marketing speculation.</p>
+</div>
+<div class="section" id="background">
+<span id="arc-meta-background"></span><h3><a class="toc-backref" href="#id6">Background</a><a class="headerlink" href="#background" title="Permalink to this headline">¶</a></h3>
+<p>This document assumes a basic familiarity with C.</p>
+<p><span class="arc-term">Blocks</span> are a C language extension for creating anonymous functions.
+Users interact with and transfer block objects using <span class="arc-term">block
+pointers</span>, which are represented like a normal pointer. A block may capture
+values from local variables; when this occurs, memory must be dynamically
+allocated. The initial allocation is done on the stack, but the runtime
+provides a <code class="docutils literal notranslate"><span class="pre">Block_copy</span></code> function which, given a block pointer, either copies
+the underlying block object to the heap, setting its reference count to 1 and
+returning the new block pointer, or (if the block object is already on the
+heap) increases its reference count by 1. The paired function is
+<code class="docutils literal notranslate"><span class="pre">Block_release</span></code>, which decreases the reference count by 1 and destroys the
+object if the count reaches zero and is on the heap.</p>
+<p>Objective-C is a set of language extensions, significant enough to be
+considered a different language. It is a strict superset of C. The extensions
+can also be imposed on C++, producing a language called Objective-C++. The
+primary feature is a single-inheritance object system; we briefly describe the
+modern dialect.</p>
+<p>Objective-C defines a new type kind, collectively called the <span class="arc-term">object
+pointer types</span>. This kind has two notable builtin members, <code class="docutils literal notranslate"><span class="pre">id</span></code> and
+<code class="docutils literal notranslate"><span class="pre">Class</span></code>; <code class="docutils literal notranslate"><span class="pre">id</span></code> is the final supertype of all object pointers. The validity
+of conversions between object pointer types is not checked at runtime. Users
+may define <span class="arc-term">classes</span>; each class is a type, and the pointer to that
+type is an object pointer type. A class may have a superclass; its pointer
+type is a subtype of its superclassâs pointer type. A class has a set of
+<span class="arc-term">ivars</span>, fields which appear on all instances of that class. For
+every class <em>T</em> thereâs an associated metaclass; it has no fields, its
+superclass is the metaclass of <em>T</em>âs superclass, and its metaclass is a global
+class. Every class has a global object whose class is the classâs metaclass;
+metaclasses have no associated type, so pointers to this object have type
+<code class="docutils literal notranslate"><span class="pre">Class</span></code>.</p>
+<p>A class declaration (<code class="docutils literal notranslate"><span class="pre">@interface</span></code>) declares a set of <span class="arc-term">methods</span>. A
+method has a return type, a list of argument types, and a <span class="arc-term">selector</span>:
+a name like <code class="docutils literal notranslate"><span class="pre">foo:bar:baz:</span></code>, where the number of colons corresponds to the
+number of formal arguments. A method may be an instance method, in which case
+it can be invoked on objects of the class, or a class method, in which case it
+can be invoked on objects of the metaclass. A method may be invoked by
+providing an object (called the <span class="arc-term">receiver</span>) and a list of formal
+arguments interspersed with the selector, like so:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="p">[</span><span class="n">receiver</span> <span class="nl">foo</span><span class="p">:</span> <span class="n">fooArg</span> <span class="nl">bar</span><span class="p">:</span> <span class="n">barArg</span> <span class="nl">baz</span><span class="p">:</span> <span class="n">bazArg</span><span class="p">]</span>
+</pre></div>
+</div>
+<p>This looks in the dynamic class of the receiver for a method with this name,
+then in that classâs superclass, etc., until it finds something it can execute.
+The receiver âexpressionâ may also be the name of a class, in which case the
+actual receiver is the class object for that class, or (within method
+definitions) it may be <code class="docutils literal notranslate"><span class="pre">super</span></code>, in which case the lookup algorithm starts
+with the static superclass instead of the dynamic class. The actual methods
+dynamically found in a class are not those declared in the <code class="docutils literal notranslate"><span class="pre">@interface</span></code>, but
+those defined in a separate <code class="docutils literal notranslate"><span class="pre">@implementation</span></code> declaration; however, when
+compiling a call, typechecking is done based on the methods declared in the
+<code class="docutils literal notranslate"><span class="pre">@interface</span></code>.</p>
+<p>Method declarations may also be grouped into <span class="arc-term">protocols</span>, which are not
+inherently associated with any class, but which classes may claim to follow.
+Object pointer types may be qualified with additional protocols that the object
+is known to support.</p>
+<p><span class="arc-term">Class extensions</span> are collections of ivars and methods, designed to
+allow a classâs <code class="docutils literal notranslate"><span class="pre">@interface</span></code> to be split across multiple files; however,
+there is still a primary implementation file which must see the
+<code class="docutils literal notranslate"><span class="pre">@interface</span></code>s of all class extensions. <span class="arc-term">Categories</span> allow
+methods (but not ivars) to be declared <em>post hoc</em> on an arbitrary class; the
+methods in the categoryâs <code class="docutils literal notranslate"><span class="pre">@implementation</span></code> will be dynamically added to that
+classâs method tables which the category is loaded at runtime, replacing those
+methods in case of a collision.</p>
+<p>In the standard environment, objects are allocated on the heap, and their
+lifetime is manually managed using a reference count. This is done using two
+instance methods which all classes are expected to implement: <code class="docutils literal notranslate"><span class="pre">retain</span></code>
+increases the objectâs reference count by 1, whereas <code class="docutils literal notranslate"><span class="pre">release</span></code> decreases it
+by 1 and calls the instance method <code class="docutils literal notranslate"><span class="pre">dealloc</span></code> if the count reaches 0. To
+simplify certain operations, there is also an <span class="arc-term">autorelease pool</span>, a
+thread-local list of objects to call <code class="docutils literal notranslate"><span class="pre">release</span></code> on later; an object can be
+added to this pool by calling <code class="docutils literal notranslate"><span class="pre">autorelease</span></code> on it.</p>
+<p>Block pointers may be converted to type <code class="docutils literal notranslate"><span class="pre">id</span></code>; block objects are laid out in a
+way that makes them compatible with Objective-C objects. There is a builtin
+class that all block objects are considered to be objects of; this class
+implements <code class="docutils literal notranslate"><span class="pre">retain</span></code> by adjusting the reference count, not by calling
+<code class="docutils literal notranslate"><span class="pre">Block_copy</span></code>.</p>
+</div>
+<div class="section" id="evolution">
+<span id="arc-meta-evolution"></span><h3><a class="toc-backref" href="#id7">Evolution</a><a class="headerlink" href="#evolution" title="Permalink to this headline">¶</a></h3>
+<p>ARC is under continual evolution, and this document must be updated as the
+language progresses.</p>
+<p>If a change increases the expressiveness of the language, for example by
+lifting a restriction or by adding new syntax, the change will be annotated
+with a revision marker, like so:</p>
+<blockquote>
+<div>ARC applies to Objective-C pointer types, block pointer types, and
+<span class="when-revised">[beginning Apple 8.0, LLVM 3.8]</span> <span class="revision">BPTRs declared
+within</span> <code class="docutils literal notranslate"><span class="pre">extern</span> <span class="pre">"BCPL"</span></code> blocks.</div></blockquote>
+<p>For now, it is sensible to version this document by the releases of its sole
+implementation (and its host project), clang. âLLVM X.Yâ refers to an
+open-source release of clang from the LLVM project. âApple X.Yâ refers to an
+Apple-provided release of the Apple LLVM Compiler. Other organizations that
+prepare their own, separately-versioned clang releases and wish to maintain
+similar information in this document should send requests to cfe-dev.</p>
+<p>If a change decreases the expressiveness of the language, for example by
+imposing a new restriction, this should be taken as an oversight in the
+original specification and something to be avoided in all versions. Such
+changes are generally to be avoided.</p>
+</div>
+</div>
+<div class="section" id="general">
+<span id="arc-general"></span><h2><a class="toc-backref" href="#id8">General</a><a class="headerlink" href="#general" title="Permalink to this headline">¶</a></h2>
+<p>Automatic Reference Counting implements automatic memory management for
+Objective-C objects and blocks, freeing the programmer from the need to
+explicitly insert retains and releases. It does not provide a cycle collector;
+users must explicitly manage the lifetime of their objects, breaking cycles
+manually or with weak or unsafe references.</p>
+<p>ARC may be explicitly enabled with the compiler flag <code class="docutils literal notranslate"><span class="pre">-fobjc-arc</span></code>. It may
+also be explicitly disabled with the compiler flag <code class="docutils literal notranslate"><span class="pre">-fno-objc-arc</span></code>. The last
+of these two flags appearing on the compile line âwinsâ.</p>
+<p>If ARC is enabled, <code class="docutils literal notranslate"><span class="pre">__has_feature(objc_arc)</span></code> will expand to 1 in the
+preprocessor. For more information about <code class="docutils literal notranslate"><span class="pre">__has_feature</span></code>, see the
+<a class="reference internal" href="LanguageExtensions.html#langext-has-feature-has-extension"><span class="std std-ref">language extensions</span></a> document.</p>
+</div>
+<div class="section" id="retainable-object-pointers">
+<span id="arc-objects"></span><h2><a class="toc-backref" href="#id9">Retainable object pointers</a><a class="headerlink" href="#retainable-object-pointers" title="Permalink to this headline">¶</a></h2>
+<p>This section describes retainable object pointers, their basic operations, and
+the restrictions imposed on their use under ARC. Note in particular that it
+covers the rules for pointer <em>values</em> (patterns of bits indicating the location
+of a pointed-to object), not pointer <em>objects</em> (locations in memory which store
+pointer values). The rules for objects are covered in the next section.</p>
+<p>A <span class="arc-term">retainable object pointer</span> (or âretainable pointerâ) is a value of
+a <span class="arc-term">retainable object pointer type</span> (âretainable typeâ). There are
+three kinds of retainable object pointer types:</p>
+<ul class="simple">
+<li>block pointers (formed by applying the caret (<code class="docutils literal notranslate"><span class="pre">^</span></code>) declarator sigil to a
+function type)</li>
+<li>Objective-C object pointers (<code class="docutils literal notranslate"><span class="pre">id</span></code>, <code class="docutils literal notranslate"><span class="pre">Class</span></code>, <code class="docutils literal notranslate"><span class="pre">NSFoo*</span></code>, etc.)</li>
+<li>typedefs marked with <code class="docutils literal notranslate"><span class="pre">__attribute__((NSObject))</span></code></li>
+</ul>
+<p>Other pointer types, such as <code class="docutils literal notranslate"><span class="pre">int*</span></code> and <code class="docutils literal notranslate"><span class="pre">CFStringRef</span></code>, are not subject to
+ARCâs semantics and restrictions.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>We are not at liberty to require all code to be recompiled with ARC;
+therefore, ARC must interoperate with Objective-C code which manages retains
+and releases manually. In general, there are three requirements in order for
+a compiler-supported reference-count system to provide reliable
+interoperation:</p>
+<ul class="last simple">
+<li>The type system must reliably identify which objects are to be managed. An
+<code class="docutils literal notranslate"><span class="pre">int*</span></code> might be a pointer to a <code class="docutils literal notranslate"><span class="pre">malloc</span></code>âed array, or it might be an
+interior pointer to such an array, or it might point to some field or local
+variable. In contrast, values of the retainable object pointer types are
+never interior.</li>
+<li>The type system must reliably indicate how to manage objects of a type.
+This usually means that the type must imply a procedure for incrementing
+and decrementing retain counts. Supporting single-ownership objects
+requires a lot more explicit mediation in the language.</li>
+<li>There must be reliable conventions for whether and when âownershipâ is
+passed between caller and callee, for both arguments and return values.
+Objective-C methods follow such a convention very reliably, at least for
+system libraries on Mac OS X, and functions always pass objects at +0. The
+C-based APIs for Core Foundation objects, on the other hand, have much more
+varied transfer semantics.</li>
+</ul>
+</div>
+<p>The use of <code class="docutils literal notranslate"><span class="pre">__attribute__((NSObject))</span></code> typedefs is not recommended. If itâs
+absolutely necessary to use this attribute, be very explicit about using the
+typedef, and do not assume that it will be preserved by language features like
+<code class="docutils literal notranslate"><span class="pre">__typeof</span></code> and C++ template argument substitution.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Any compiler operation which incidentally strips type âsugarâ from a type
+will yield a type without the attribute, which may result in unexpected
+behavior.</p>
+</div>
+<div class="section" id="retain-count-semantics">
+<span id="arc-objects-retains"></span><h3><a class="toc-backref" href="#id10">Retain count semantics</a><a class="headerlink" href="#retain-count-semantics" title="Permalink to this headline">¶</a></h3>
+<p>A retainable object pointer is either a <span class="arc-term">null pointer</span> or a pointer
+to a valid object. Furthermore, if it has block pointer type and is not
+<code class="docutils literal notranslate"><span class="pre">null</span></code> then it must actually be a pointer to a block object, and if it has
+<code class="docutils literal notranslate"><span class="pre">Class</span></code> type (possibly protocol-qualified) then it must actually be a pointer
+to a class object. Otherwise ARC does not enforce the Objective-C type system
+as long as the implementing methods follow the signature of the static type.
+It is undefined behavior if ARC is exposed to an invalid pointer.</p>
+<p>For ARCâs purposes, a valid object is one with âwell-behavedâ retaining
+operations. Specifically, the object must be laid out such that the
+Objective-C message send machinery can successfully send it the following
+messages:</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">retain</span></code>, taking no arguments and returning a pointer to the object.</li>
+<li><code class="docutils literal notranslate"><span class="pre">release</span></code>, taking no arguments and returning <code class="docutils literal notranslate"><span class="pre">void</span></code>.</li>
+<li><code class="docutils literal notranslate"><span class="pre">autorelease</span></code>, taking no arguments and returning a pointer to the object.</li>
+</ul>
+<p>The behavior of these methods is constrained in the following ways. The term
+<span class="arc-term">high-level semantics</span> is an intentionally vague term; the intent is
+that programmers must implement these methods in a way such that the compiler,
+modifying code in ways it deems safe according to these constraints, will not
+violate their requirements. For example, if the user puts logging statements
+in <code class="docutils literal notranslate"><span class="pre">retain</span></code>, they should not be surprised if those statements are executed
+more or less often depending on optimization settings. These constraints are
+not exhaustive of the optimization opportunities: values held in local
+variables are subject to additional restrictions, described later in this
+document.</p>
+<p>It is undefined behavior if a computation history featuring a send of
+<code class="docutils literal notranslate"><span class="pre">retain</span></code> followed by a send of <code class="docutils literal notranslate"><span class="pre">release</span></code> to the same object, with no
+intervening <code class="docutils literal notranslate"><span class="pre">release</span></code> on that object, is not equivalent under the high-level
+semantics to a computation history in which these sends are removed. Note that
+this implies that these methods may not raise exceptions.</p>
+<p>It is undefined behavior if a computation history features any use whatsoever
+of an object following the completion of a send of <code class="docutils literal notranslate"><span class="pre">release</span></code> that is not
+preceded by a send of <code class="docutils literal notranslate"><span class="pre">retain</span></code> to the same object.</p>
+<p>The behavior of <code class="docutils literal notranslate"><span class="pre">autorelease</span></code> must be equivalent to sending <code class="docutils literal notranslate"><span class="pre">release</span></code> when
+one of the autorelease pools currently in scope is popped. It may not throw an
+exception.</p>
+<p>When the semantics call for performing one of these operations on a retainable
+object pointer, if that pointer is <code class="docutils literal notranslate"><span class="pre">null</span></code> then the effect is a no-op.</p>
+<p>All of the semantics described in this document are subject to additional
+<a class="reference internal" href="#arc-optimization"><span class="std std-ref">optimization rules</span></a> which permit the removal or
+optimization of operations based on local knowledge of data flow. The
+semantics describe the high-level behaviors that the compiler implements, not
+an exact sequence of operations that a program will be compiled into.</p>
+</div>
+<div class="section" id="retainable-object-pointers-as-operands-and-arguments">
+<span id="arc-objects-operands"></span><h3><a class="toc-backref" href="#id11">Retainable object pointers as operands and arguments</a><a class="headerlink" href="#retainable-object-pointers-as-operands-and-arguments" title="Permalink to this headline">¶</a></h3>
+<p>In general, ARC does not perform retain or release operations when simply using
+a retainable object pointer as an operand within an expression. This includes:</p>
+<ul class="simple">
+<li>loading a retainable pointer from an object with non-weak <a class="reference internal" href="#arc-ownership"><span class="std std-ref">ownership</span></a>,</li>
+<li>passing a retainable pointer as an argument to a function or method, and</li>
+<li>receiving a retainable pointer as the result of a function or method call.</li>
+</ul>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">While this might seem uncontroversial, it is actually unsafe when multiple
+expressions are evaluated in âparallelâ, as with binary operators and calls,
+because (for example) one expression might load from an object while another
+writes to it. However, C and C++ already call this undefined behavior
+because the evaluations are unsequenced, and ARC simply exploits that here to
+avoid needing to retain arguments across a large number of calls.</p>
+</div>
+<p>The remainder of this section describes exceptions to these rules, how those
+exceptions are detected, and what those exceptions imply semantically.</p>
+<div class="section" id="consumed-parameters">
+<span id="arc-objects-operands-consumed"></span><h4><a class="toc-backref" href="#id12">Consumed parameters</a><a class="headerlink" href="#consumed-parameters" title="Permalink to this headline">¶</a></h4>
+<p>A function or method parameter of retainable object pointer type may be marked
+as <span class="arc-term">consumed</span>, signifying that the callee expects to take ownership
+of a +1 retain count. This is done by adding the <code class="docutils literal notranslate"><span class="pre">ns_consumed</span></code> attribute to
+the parameter declaration, like so:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">foo</span><span class="p">(</span><span class="n">__attribute</span><span class="p">((</span><span class="n">ns_consumed</span><span class="p">))</span> <span class="kt">id</span> <span class="n">x</span><span class="p">);</span>
+<span class="p">-</span> <span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="nf">foo:</span> <span class="p">(</span><span class="kt">id</span><span class="p">)</span> <span class="nv">__attribute</span><span class="p">((</span><span class="n">ns_consumed</span><span class="p">))</span> <span class="nv">x</span><span class="p">;</span>
+</pre></div>
+</div>
+<p>This attribute is part of the type of the function or method, not the type of
+the parameter. It controls only how the argument is passed and received.</p>
+<p>When passing such an argument, ARC retains the argument prior to making the
+call.</p>
+<p>When receiving such an argument, ARC releases the argument at the end of the
+function, subject to the usual optimizations for local values.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">This formalizes direct transfers of ownership from a caller to a callee. The
+most common scenario here is passing the <code class="docutils literal notranslate"><span class="pre">self</span></code> parameter to <code class="docutils literal notranslate"><span class="pre">init</span></code>, but
+it is useful to generalize. Typically, local optimization will remove any
+extra retains and releases: on the caller side the retain will be merged with
+a +1 source, and on the callee side the release will be rolled into the
+initialization of the parameter.</p>
+</div>
+<p>The implicit <code class="docutils literal notranslate"><span class="pre">self</span></code> parameter of a method may be marked as consumed by adding
+<code class="docutils literal notranslate"><span class="pre">__attribute__((ns_consumes_self))</span></code> to the method declaration. Methods in
+the <code class="docutils literal notranslate"><span class="pre">init</span></code> <a class="reference internal" href="#arc-method-families"><span class="std std-ref">family</span></a> are treated as if they were
+implicitly marked with this attribute.</p>
+<p>It is undefined behavior if an Objective-C message send to a method with
+<code class="docutils literal notranslate"><span class="pre">ns_consumed</span></code> parameters (other than self) is made with a null receiver. It
+is undefined behavior if the method to which an Objective-C message send
+statically resolves to has a different set of <code class="docutils literal notranslate"><span class="pre">ns_consumed</span></code> parameters than
+the method it dynamically resolves to. It is undefined behavior if a block or
+function call is made through a static type with a different set of
+<code class="docutils literal notranslate"><span class="pre">ns_consumed</span></code> parameters than the implementation of the called block or
+function.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Consumed parameters with null receiver are a guaranteed leak. Mismatches
+with consumed parameters will cause over-retains or over-releases, depending
+on the direction. The rule about function calls is really just an
+application of the existing C/C++ rule about calling functions through an
+incompatible function type, but itâs useful to state it explicitly.</p>
+</div>
+</div>
+<div class="section" id="retained-return-values">
+<span id="arc-object-operands-retained-return-values"></span><h4><a class="toc-backref" href="#id13">Retained return values</a><a class="headerlink" href="#retained-return-values" title="Permalink to this headline">¶</a></h4>
+<p>A function or method which returns a retainable object pointer type may be
+marked as returning a retained value, signifying that the caller expects to take
+ownership of a +1 retain count. This is done by adding the
+<code class="docutils literal notranslate"><span class="pre">ns_returns_retained</span></code> attribute to the function or method declaration, like
+so:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">id</span> <span class="nf">foo</span><span class="p">(</span><span class="kt">void</span><span class="p">)</span> <span class="n">__attribute</span><span class="p">((</span><span class="n">ns_returns_retained</span><span class="p">));</span>
+<span class="p">-</span> <span class="p">(</span><span class="kt">id</span><span class="p">)</span> <span class="nf">foo</span> <span class="n">__attribute</span><span class="p">((</span><span class="n">ns_returns_retained</span><span class="p">));</span>
+</pre></div>
+</div>
+<p>This attribute is part of the type of the function or method.</p>
+<p>When returning from such a function or method, ARC retains the value at the
+point of evaluation of the return statement, before leaving all local scopes.</p>
+<p>When receiving a return result from such a function or method, ARC releases the
+value at the end of the full-expression it is contained within, subject to the
+usual optimizations for local values.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">This formalizes direct transfers of ownership from a callee to a caller. The
+most common scenario this models is the retained return from <code class="docutils literal notranslate"><span class="pre">init</span></code>,
+<code class="docutils literal notranslate"><span class="pre">alloc</span></code>, <code class="docutils literal notranslate"><span class="pre">new</span></code>, and <code class="docutils literal notranslate"><span class="pre">copy</span></code> methods, but there are other cases in the
+frameworks. After optimization there are typically no extra retains and
+releases required.</p>
+</div>
+<p>Methods in the <code class="docutils literal notranslate"><span class="pre">alloc</span></code>, <code class="docutils literal notranslate"><span class="pre">copy</span></code>, <code class="docutils literal notranslate"><span class="pre">init</span></code>, <code class="docutils literal notranslate"><span class="pre">mutableCopy</span></code>, and <code class="docutils literal notranslate"><span class="pre">new</span></code>
+<a class="reference internal" href="#arc-method-families"><span class="std std-ref">families</span></a> are implicitly marked
+<code class="docutils literal notranslate"><span class="pre">__attribute__((ns_returns_retained))</span></code>. This may be suppressed by explicitly
+marking the method <code class="docutils literal notranslate"><span class="pre">__attribute__((ns_returns_not_retained))</span></code>.</p>
+<p>It is undefined behavior if the method to which an Objective-C message send
+statically resolves has different retain semantics on its result from the
+method it dynamically resolves to. It is undefined behavior if a block or
+function call is made through a static type with different retain semantics on
+its result from the implementation of the called block or function.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Mismatches with returned results will cause over-retains or over-releases,
+depending on the direction. Again, the rule about function calls is really
+just an application of the existing C/C++ rule about calling functions
+through an incompatible function type.</p>
+</div>
+</div>
+<div class="section" id="unretained-return-values">
+<span id="arc-objects-operands-unretained-returns"></span><h4><a class="toc-backref" href="#id14">Unretained return values</a><a class="headerlink" href="#unretained-return-values" title="Permalink to this headline">¶</a></h4>
+<p>A method or function which returns a retainable object type but does not return
+a retained value must ensure that the object is still valid across the return
+boundary.</p>
+<p>When returning from such a function or method, ARC retains the value at the
+point of evaluation of the return statement, then leaves all local scopes, and
+then balances out the retain while ensuring that the value lives across the
+call boundary. In the worst case, this may involve an <code class="docutils literal notranslate"><span class="pre">autorelease</span></code>, but
+callers must not assume that the value is actually in the autorelease pool.</p>
+<p>ARC performs no extra mandatory work on the caller side, although it may elect
+to do something to shorten the lifetime of the returned value.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">It is common in non-ARC code to not return an autoreleased value; therefore
+the convention does not force either path. It is convenient to not be
+required to do unnecessary retains and autoreleases; this permits
+optimizations such as eliding retain/autoreleases when it can be shown that
+the original pointer will still be valid at the point of return.</p>
+</div>
+<p>A method or function may be marked with
+<code class="docutils literal notranslate"><span class="pre">__attribute__((ns_returns_autoreleased))</span></code> to indicate that it returns a
+pointer which is guaranteed to be valid at least as long as the innermost
+autorelease pool. There are no additional semantics enforced in the definition
+of such a method; it merely enables optimizations in callers.</p>
+</div>
+<div class="section" id="bridged-casts">
+<span id="arc-objects-operands-casts"></span><h4><a class="toc-backref" href="#id15">Bridged casts</a><a class="headerlink" href="#bridged-casts" title="Permalink to this headline">¶</a></h4>
+<p>A <span class="arc-term">bridged cast</span> is a C-style cast annotated with one of three
+keywords:</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">(__bridge</span> <span class="pre">T)</span> <span class="pre">op</span></code> casts the operand to the destination type <code class="docutils literal notranslate"><span class="pre">T</span></code>. If
+<code class="docutils literal notranslate"><span class="pre">T</span></code> is a retainable object pointer type, then <code class="docutils literal notranslate"><span class="pre">op</span></code> must have a
+non-retainable pointer type. If <code class="docutils literal notranslate"><span class="pre">T</span></code> is a non-retainable pointer type,
+then <code class="docutils literal notranslate"><span class="pre">op</span></code> must have a retainable object pointer type. Otherwise the cast
+is ill-formed. There is no transfer of ownership, and ARC inserts no retain
+operations.</li>
+<li><code class="docutils literal notranslate"><span class="pre">(__bridge_retained</span> <span class="pre">T)</span> <span class="pre">op</span></code> casts the operand, which must have retainable
+object pointer type, to the destination type, which must be a non-retainable
+pointer type. ARC retains the value, subject to the usual optimizations on
+local values, and the recipient is responsible for balancing that +1.</li>
+<li><code class="docutils literal notranslate"><span class="pre">(__bridge_transfer</span> <span class="pre">T)</span> <span class="pre">op</span></code> casts the operand, which must have
+non-retainable pointer type, to the destination type, which must be a
+retainable object pointer type. ARC will release the value at the end of
+the enclosing full-expression, subject to the usual optimizations on local
+values.</li>
+</ul>
+<p>These casts are required in order to transfer objects in and out of ARC
+control; see the rationale in the section on <a class="reference internal" href="#arc-objects-restrictions-conversion"><span class="std std-ref">conversion of retainable
+object pointers</span></a>.</p>
+<p>Using a <code class="docutils literal notranslate"><span class="pre">__bridge_retained</span></code> or <code class="docutils literal notranslate"><span class="pre">__bridge_transfer</span></code> cast purely to convince
+ARC to emit an unbalanced retain or release, respectively, is poor form.</p>
+</div>
+</div>
+<div class="section" id="restrictions">
+<span id="arc-objects-restrictions"></span><h3><a class="toc-backref" href="#id16">Restrictions</a><a class="headerlink" href="#restrictions" title="Permalink to this headline">¶</a></h3>
+<div class="section" id="conversion-of-retainable-object-pointers">
+<span id="arc-objects-restrictions-conversion"></span><h4><a class="toc-backref" href="#id17">Conversion of retainable object pointers</a><a class="headerlink" href="#conversion-of-retainable-object-pointers" title="Permalink to this headline">¶</a></h4>
+<p>In general, a program which attempts to implicitly or explicitly convert a
+value of retainable object pointer type to any non-retainable type, or
+vice-versa, is ill-formed. For example, an Objective-C object pointer shall
+not be converted to <code class="docutils literal notranslate"><span class="pre">void*</span></code>. As an exception, cast to <code class="docutils literal notranslate"><span class="pre">intptr_t</span></code> is
+allowed because such casts are not transferring ownership. The <a class="reference internal" href="#arc-objects-operands-casts"><span class="std std-ref">bridged
+casts</span></a> may be used to perform these conversions
+where necessary.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">We cannot ensure the correct management of the lifetime of objects if they
+may be freely passed around as unmanaged types. The bridged casts are
+provided so that the programmer may explicitly describe whether the cast
+transfers control into or out of ARC.</p>
+</div>
+<p>However, the following exceptions apply.</p>
+</div>
+<div class="section" id="conversion-to-retainable-object-pointer-type-of-expressions-with-known-semantics">
+<span id="arc-objects-restrictions-conversion-with-known-semantics"></span><h4><a class="toc-backref" href="#id18">Conversion to retainable object pointer type of expressions with known semantics</a><a class="headerlink" href="#conversion-to-retainable-object-pointer-type-of-expressions-with-known-semantics" title="Permalink to this headline">¶</a></h4>
+<p><span class="when-revised">[beginning Apple 4.0, LLVM 3.1]</span>
+<span class="revision">These exceptions have been greatly expanded; they previously applied
+only to a much-reduced subset which is difficult to categorize but which
+included null pointers, message sends (under the given rules), and the various
+global constants.</span></p>
+<p>An unbridged conversion to a retainable object pointer type from a type other
+than a retainable object pointer type is ill-formed, as discussed above, unless
+the operand of the cast has a syntactic form which is known retained, known
+unretained, or known retain-agnostic.</p>
+<p>An expression is <span class="arc-term">known retain-agnostic</span> if it is:</p>
+<ul class="simple">
+<li>an Objective-C string literal,</li>
+<li>a load from a <code class="docutils literal notranslate"><span class="pre">const</span></code> system global variable of <a class="reference internal" href="#arc-misc-c-retainable"><span class="std std-ref">C retainable pointer
+type</span></a>, or</li>
+<li>a null pointer constant.</li>
+</ul>
+<p>An expression is <span class="arc-term">known unretained</span> if it is an rvalue of <a class="reference internal" href="#arc-misc-c-retainable"><span class="std std-ref">C
+retainable pointer type</span></a> and it is:</p>
+<ul class="simple">
+<li>a direct call to a function, and either that function has the
+<code class="docutils literal notranslate"><span class="pre">cf_returns_not_retained</span></code> attribute or it is an <a class="reference internal" href="#arc-misc-c-retainable-audit"><span class="std std-ref">audited</span></a> function that does not have the
+<code class="docutils literal notranslate"><span class="pre">cf_returns_retained</span></code> attribute and does not follow the create/copy naming
+convention,</li>
+<li>a message send, and the declared method either has the
+<code class="docutils literal notranslate"><span class="pre">cf_returns_not_retained</span></code> attribute or it has neither the
+<code class="docutils literal notranslate"><span class="pre">cf_returns_retained</span></code> attribute nor a <a class="reference internal" href="#arc-method-families"><span class="std std-ref">selector family</span></a> that implies a retained result, or</li>
+<li><span class="when-revised">[beginning LLVM 3.6]</span> <span class="revision">a load from a</span> <code class="docutils literal notranslate"><span class="pre">const</span></code>
+<span class="revision">non-system global variable.</span></li>
+</ul>
+<p>An expression is <span class="arc-term">known retained</span> if it is an rvalue of <a class="reference internal" href="#arc-misc-c-retainable"><span class="std std-ref">C
+retainable pointer type</span></a> and it is:</p>
+<ul class="simple">
+<li>a message send, and the declared method either has the
+<code class="docutils literal notranslate"><span class="pre">cf_returns_retained</span></code> attribute, or it does not have the
+<code class="docutils literal notranslate"><span class="pre">cf_returns_not_retained</span></code> attribute but it does have a <a class="reference internal" href="#arc-method-families"><span class="std std-ref">selector
+family</span></a> that implies a retained result.</li>
+</ul>
+<p>Furthermore:</p>
+<ul class="simple">
+<li>a comma expression is classified according to its right-hand side,</li>
+<li>a statement expression is classified according to its result expression, if
+it has one,</li>
+<li>an lvalue-to-rvalue conversion applied to an Objective-C property lvalue is
+classified according to the underlying message send, and</li>
+<li>a conditional operator is classified according to its second and third
+operands, if they agree in classification, or else the other if one is known
+retain-agnostic.</li>
+</ul>
+<p>If the cast operand is known retained, the conversion is treated as a
+<code class="docutils literal notranslate"><span class="pre">__bridge_transfer</span></code> cast. If the cast operand is known unretained or known
+retain-agnostic, the conversion is treated as a <code class="docutils literal notranslate"><span class="pre">__bridge</span></code> cast.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>Bridging casts are annoying. Absent the ability to completely automate the
+management of CF objects, however, we are left with relatively poor attempts
+to reduce the need for a glut of explicit bridges. Hence these rules.</p>
+<p>Weâve so far consciously refrained from implicitly turning retained CF
+results from function calls into <code class="docutils literal notranslate"><span class="pre">__bridge_transfer</span></code> casts. The worry is
+that some code patterns â for example, creating a CF value, assigning it
+to an ObjC-typed local, and then calling <code class="docutils literal notranslate"><span class="pre">CFRelease</span></code> when done â are a
+bit too likely to be accidentally accepted, leading to mysterious behavior.</p>
+<p class="last">For loads from <code class="docutils literal notranslate"><span class="pre">const</span></code> global variables of <a class="reference internal" href="#arc-misc-c-retainable"><span class="std std-ref">C retainable pointer type</span></a>, it is reasonable to assume that global system
+constants were initialitzed with true constants (e.g. string literals), but
+user constants might have been initialized with something dynamically
+allocated, using a global initializer.</p>
+</div>
+</div>
+<div class="section" id="conversion-from-retainable-object-pointer-type-in-certain-contexts">
+<span id="arc-objects-restrictions-conversion-exception-contextual"></span><h4><a class="toc-backref" href="#id19">Conversion from retainable object pointer type in certain contexts</a><a class="headerlink" href="#conversion-from-retainable-object-pointer-type-in-certain-contexts" title="Permalink to this headline">¶</a></h4>
+<p><span class="when-revised">[beginning Apple 4.0, LLVM 3.1]</span></p>
+<p>If an expression of retainable object pointer type is explicitly cast to a
+<a class="reference internal" href="#arc-misc-c-retainable"><span class="std std-ref">C retainable pointer type</span></a>, the program is
+ill-formed as discussed above unless the result is immediately used:</p>
+<ul class="simple">
+<li>to initialize a parameter in an Objective-C message send where the parameter
+is not marked with the <code class="docutils literal notranslate"><span class="pre">cf_consumed</span></code> attribute, or</li>
+<li>to initialize a parameter in a direct call to an
+<a class="reference internal" href="#arc-misc-c-retainable-audit"><span class="std std-ref">audited</span></a> function where the parameter is
+not marked with the <code class="docutils literal notranslate"><span class="pre">cf_consumed</span></code> attribute.</li>
+</ul>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Consumed parameters are left out because ARC would naturally balance them
+with a retain, which was judged too treacherous. This is in part because
+several of the most common consuming functions are in the <code class="docutils literal notranslate"><span class="pre">Release</span></code> family,
+and it would be quite unfortunate for explicit releases to be silently
+balanced out in this way.</p>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="ownership-qualification">
+<span id="arc-ownership"></span><h2><a class="toc-backref" href="#id20">Ownership qualification</a><a class="headerlink" href="#ownership-qualification" title="Permalink to this headline">¶</a></h2>
+<p>This section describes the behavior of <em>objects</em> of retainable object pointer
+type; that is, locations in memory which store retainable object pointers.</p>
+<p>A type is a <span class="arc-term">retainable object owner type</span> if it is a retainable
+object pointer type or an array type whose element type is a retainable object
+owner type.</p>
+<p>An <span class="arc-term">ownership qualifier</span> is a type qualifier which applies only to
+retainable object owner types. An array type is ownership-qualified according
+to its element type, and adding an ownership qualifier to an array type so
+qualifies its element type.</p>
+<p>A program is ill-formed if it attempts to apply an ownership qualifier to a
+type which is already ownership-qualified, even if it is the same qualifier.
+There is a single exception to this rule: an ownership qualifier may be applied
+to a substituted template type parameter, which overrides the ownership
+qualifier provided by the template argument.</p>
+<p>When forming a function type, the result type is adjusted so that any
+top-level ownership qualifier is deleted.</p>
+<p>Except as described under the <a class="reference internal" href="#arc-ownership-inference"><span class="std std-ref">inference rules</span></a>,
+a program is ill-formed if it attempts to form a pointer or reference type to a
+retainable object owner type which lacks an ownership qualifier.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">These rules, together with the inference rules, ensure that all objects and
+lvalues of retainable object pointer type have an ownership qualifier. The
+ability to override an ownership qualifier during template substitution is
+required to counteract the <a class="reference internal" href="#arc-ownership-inference-template-arguments"><span class="std std-ref">inference of __strong for template type
+arguments</span></a>. Ownership qualifiers
+on return types are dropped because they serve no purpose there except to
+cause spurious problems with overloading and templates.</p>
+</div>
+<p>There are four ownership qualifiers:</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code></li>
+<li><code class="docutils literal notranslate"><span class="pre">__strong</span></code></li>
+<li><code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code></li>
+<li><code class="docutils literal notranslate"><span class="pre">__weak</span></code></li>
+</ul>
+<p>A type is <span class="arc-term">nontrivially ownership-qualified</span> if it is qualified with
+<code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code>, <code class="docutils literal notranslate"><span class="pre">__strong</span></code>, or <code class="docutils literal notranslate"><span class="pre">__weak</span></code>.</p>
+<div class="section" id="spelling">
+<span id="arc-ownership-spelling"></span><h3><a class="toc-backref" href="#id21">Spelling</a><a class="headerlink" href="#spelling" title="Permalink to this headline">¶</a></h3>
+<p>The names of the ownership qualifiers are reserved for the implementation. A
+program may not assume that they are or are not implemented with macros, or
+what those macros expand to.</p>
+<p>An ownership qualifier may be written anywhere that any other type qualifier
+may be written.</p>
+<p>If an ownership qualifier appears in the <em>declaration-specifiers</em>, the
+following rules apply:</p>
+<ul class="simple">
+<li>if the type specifier is a retainable object owner type, the qualifier
+initially applies to that type;</li>
+<li>otherwise, if the outermost non-array declarator is a pointer
+or block pointer declarator, the qualifier initially applies to
+that type;</li>
+<li>otherwise the program is ill-formed.</li>
+<li>If the qualifier is so applied at a position in the declaration
+where the next-innermost declarator is a function declarator, and
+there is an block declarator within that function declarator, then
+the qualifier applies instead to that block declarator and this rule
+is considered afresh beginning from the new position.</li>
+</ul>
+<p>If an ownership qualifier appears on the declarator name, or on the declared
+object, it is applied to the innermost pointer or block-pointer type.</p>
+<p>If an ownership qualifier appears anywhere else in a declarator, it applies to
+the type there.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Ownership qualifiers are like <code class="docutils literal notranslate"><span class="pre">const</span></code> and <code class="docutils literal notranslate"><span class="pre">volatile</span></code> in the sense
+that they may sensibly apply at multiple distinct positions within a
+declarator. However, unlike those qualifiers, there are many
+situations where they are not meaningful, and so we make an effort
+to âmoveâ the qualifier to a place where it will be meaningful. The
+general goal is to allow the programmer to write, say, <code class="docutils literal notranslate"><span class="pre">__strong</span></code>
+before the entire declaration and have it apply in the leftmost
+sensible place.</p>
+</div>
+<div class="section" id="property-declarations">
+<span id="arc-ownership-spelling-property"></span><h4><a class="toc-backref" href="#id22">Property declarations</a><a class="headerlink" href="#property-declarations" title="Permalink to this headline">¶</a></h4>
+<p>A property of retainable object pointer type may have ownership. If the
+propertyâs type is ownership-qualified, then the property has that ownership.
+If the property has one of the following modifiers, then the property has the
+corresponding ownership. A property is ill-formed if it has conflicting
+sources of ownership, or if it has redundant ownership modifiers, or if it has
+<code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> ownership.</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">assign</span></code> implies <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code> ownership.</li>
+<li><code class="docutils literal notranslate"><span class="pre">copy</span></code> implies <code class="docutils literal notranslate"><span class="pre">__strong</span></code> ownership, as well as the usual behavior of
+copy semantics on the setter.</li>
+<li><code class="docutils literal notranslate"><span class="pre">retain</span></code> implies <code class="docutils literal notranslate"><span class="pre">__strong</span></code> ownership.</li>
+<li><code class="docutils literal notranslate"><span class="pre">strong</span></code> implies <code class="docutils literal notranslate"><span class="pre">__strong</span></code> ownership.</li>
+<li><code class="docutils literal notranslate"><span class="pre">unsafe_unretained</span></code> implies <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code> ownership.</li>
+<li><code class="docutils literal notranslate"><span class="pre">weak</span></code> implies <code class="docutils literal notranslate"><span class="pre">__weak</span></code> ownership.</li>
+</ul>
+<p>With the exception of <code class="docutils literal notranslate"><span class="pre">weak</span></code>, these modifiers are available in non-ARC
+modes.</p>
+<p>A propertyâs specified ownership is preserved in its metadata, but otherwise
+the meaning is purely conventional unless the property is synthesized. If a
+property is synthesized, then the <span class="arc-term">associated instance variable</span> is
+the instance variable which is named, possibly implicitly, by the
+<code class="docutils literal notranslate"><span class="pre">@synthesize</span></code> declaration. If the associated instance variable already
+exists, then its ownership qualification must equal the ownership of the
+property; otherwise, the instance variable is created with that ownership
+qualification.</p>
+<p>A property of retainable object pointer type which is synthesized without a
+source of ownership has the ownership of its associated instance variable, if it
+already exists; otherwise, <span class="when-revised">[beginning Apple 3.1, LLVM 3.1]</span>
+<span class="revision">its ownership is implicitly</span> <code class="docutils literal notranslate"><span class="pre">strong</span></code>. Prior to this revision, it
+was ill-formed to synthesize such a property.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Using <code class="docutils literal notranslate"><span class="pre">strong</span></code> by default is safe and consistent with the generic ARC rule
+about <a class="reference internal" href="#arc-ownership-inference-variables"><span class="std std-ref">inferring ownership</span></a>. It is,
+unfortunately, inconsistent with the non-ARC rule which states that such
+properties are implicitly <code class="docutils literal notranslate"><span class="pre">assign</span></code>. However, that rule is clearly
+untenable in ARC, since it leads to default-unsafe code. The main merit to
+banning the properties is to avoid confusion with non-ARC practice, which did
+not ultimately strike us as sufficient to justify requiring extra syntax and
+(more importantly) forcing novices to understand ownership rules just to
+declare a property when the default is so reasonable. Changing the rule away
+from non-ARC practice was acceptable because we had conservatively banned the
+synthesis in order to give ourselves exactly this leeway.</p>
+</div>
+<p>Applying <code class="docutils literal notranslate"><span class="pre">__attribute__((NSObject))</span></code> to a property not of retainable object
+pointer type has the same behavior it does outside of ARC: it requires the
+property type to be some sort of pointer and permits the use of modifiers other
+than <code class="docutils literal notranslate"><span class="pre">assign</span></code>. These modifiers only affect the synthesized getter and
+setter; direct accesses to the ivar (even if synthesized) still have primitive
+semantics, and the value in the ivar will not be automatically released during
+deallocation.</p>
+</div>
+</div>
+<div class="section" id="semantics">
+<span id="arc-ownership-semantics"></span><h3><a class="toc-backref" href="#id23">Semantics</a><a class="headerlink" href="#semantics" title="Permalink to this headline">¶</a></h3>
+<p>There are five <span class="arc-term">managed operations</span> which may be performed on an
+object of retainable object pointer type. Each qualifier specifies different
+semantics for each of these operations. It is still undefined behavior to
+access an object outside of its lifetime.</p>
+<p>A load or store with âprimitive semanticsâ has the same semantics as the
+respective operation would have on an <code class="docutils literal notranslate"><span class="pre">void*</span></code> lvalue with the same alignment
+and non-ownership qualification.</p>
+<p><span class="arc-term">Reading</span> occurs when performing a lvalue-to-rvalue conversion on an
+object lvalue.</p>
+<ul class="simple">
+<li>For <code class="docutils literal notranslate"><span class="pre">__weak</span></code> objects, the current pointee is retained and then released at
+the end of the current full-expression. This must execute atomically with
+respect to assignments and to the final release of the pointee.</li>
+<li>For all other objects, the lvalue is loaded with primitive semantics.</li>
+</ul>
+<p><span class="arc-term">Assignment</span> occurs when evaluating an assignment operator. The
+semantics vary based on the qualification:</p>
+<ul class="simple">
+<li>For <code class="docutils literal notranslate"><span class="pre">__strong</span></code> objects, the new pointee is first retained; second, the
+lvalue is loaded with primitive semantics; third, the new pointee is stored
+into the lvalue with primitive semantics; and finally, the old pointee is
+released. This is not performed atomically; external synchronization must be
+used to make this safe in the face of concurrent loads and stores.</li>
+<li>For <code class="docutils literal notranslate"><span class="pre">__weak</span></code> objects, the lvalue is updated to point to the new pointee,
+unless the new pointee is an object currently undergoing deallocation, in
+which case the lvalue is updated to a null pointer. This must execute
+atomically with respect to other assignments to the object, to reads from the
+object, and to the final release of the new pointee.</li>
+<li>For <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code> objects, the new pointee is stored into the
+lvalue using primitive semantics.</li>
+<li>For <code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> objects, the new pointee is retained, autoreleased,
+and stored into the lvalue using primitive semantics.</li>
+</ul>
+<p><span class="arc-term">Initialization</span> occurs when an objectâs lifetime begins, which
+depends on its storage duration. Initialization proceeds in two stages:</p>
+<ol class="arabic simple">
+<li>First, a null pointer is stored into the lvalue using primitive semantics.
+This step is skipped if the object is <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code>.</li>
+<li>Second, if the object has an initializer, that expression is evaluated and
+then assigned into the object using the usual assignment semantics.</li>
+</ol>
+<p><span class="arc-term">Destruction</span> occurs when an objectâs lifetime ends. In all cases it
+is semantically equivalent to assigning a null pointer to the object, with the
+proviso that of course the object cannot be legally read after the objectâs
+lifetime ends.</p>
+<p><span class="arc-term">Moving</span> occurs in specific situations where an lvalue is âmoved
+fromâ, meaning that its current pointee will be used but the object may be left
+in a different (but still valid) state. This arises with <code class="docutils literal notranslate"><span class="pre">__block</span></code> variables
+and rvalue references in C++. For <code class="docutils literal notranslate"><span class="pre">__strong</span></code> lvalues, moving is equivalent
+to loading the lvalue with primitive semantics, writing a null pointer to it
+with primitive semantics, and then releasing the result of the load at the end
+of the current full-expression. For all other lvalues, moving is equivalent to
+reading the object.</p>
+</div>
+<div class="section" id="arc-ownership-restrictions">
+<span id="id1"></span><h3><a class="toc-backref" href="#id24">Restrictions</a><a class="headerlink" href="#arc-ownership-restrictions" title="Permalink to this headline">¶</a></h3>
+<div class="section" id="weak-unavailable-types">
+<span id="arc-ownership-restrictions-weak"></span><h4><a class="toc-backref" href="#id25">Weak-unavailable types</a><a class="headerlink" href="#weak-unavailable-types" title="Permalink to this headline">¶</a></h4>
+<p>It is explicitly permitted for Objective-C classes to not support <code class="docutils literal notranslate"><span class="pre">__weak</span></code>
+references. It is undefined behavior to perform an operation with weak
+assignment semantics with a pointer to an Objective-C object whose class does
+not support <code class="docutils literal notranslate"><span class="pre">__weak</span></code> references.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Historically, it has been possible for a class to provide its own
+reference-count implementation by overriding <code class="docutils literal notranslate"><span class="pre">retain</span></code>, <code class="docutils literal notranslate"><span class="pre">release</span></code>, etc.
+However, weak references to an object require coordination with its classâs
+reference-count implementation because, among other things, weak loads and
+stores must be atomic with respect to the final release. Therefore, existing
+custom reference-count implementations will generally not support weak
+references without additional effort. This is unavoidable without breaking
+binary compatibility.</p>
+</div>
+<p>A class may indicate that it does not support weak references by providing the
+<code class="docutils literal notranslate"><span class="pre">objc_arc_weak_reference_unavailable</span></code> attribute on the classâs interface declaration. A
+retainable object pointer type is <strong>weak-unavailable</strong> if
+is a pointer to an (optionally protocol-qualified) Objective-C class <code class="docutils literal notranslate"><span class="pre">T</span></code> where
+<code class="docutils literal notranslate"><span class="pre">T</span></code> or one of its superclasses has the <code class="docutils literal notranslate"><span class="pre">objc_arc_weak_reference_unavailable</span></code>
+attribute. A program is ill-formed if it applies the <code class="docutils literal notranslate"><span class="pre">__weak</span></code> ownership
+qualifier to a weak-unavailable type or if the value operand of a weak
+assignment operation has a weak-unavailable type.</p>
+</div>
+<div class="section" id="storage-duration-of-autoreleasing-objects">
+<span id="arc-ownership-restrictions-autoreleasing"></span><h4><a class="toc-backref" href="#id26">Storage duration of <code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> objects</a><a class="headerlink" href="#storage-duration-of-autoreleasing-objects" title="Permalink to this headline">¶</a></h4>
+<p>A program is ill-formed if it declares an <code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> object of
+non-automatic storage duration. A program is ill-formed if it captures an
+<code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> object in a block or, unless by reference, in a C++11
+lambda.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Autorelease pools are tied to the current thread and scope by their nature.
+While it is possible to have temporary objects whose instance variables are
+filled with autoreleased objects, there is no way that ARC can provide any
+sort of safety guarantee there.</p>
+</div>
+<p>It is undefined behavior if a non-null pointer is assigned to an
+<code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> object while an autorelease pool is in scope and then that
+object is read after the autorelease poolâs scope is left.</p>
+</div>
+<div class="section" id="conversion-of-pointers-to-ownership-qualified-types">
+<span id="arc-ownership-restrictions-conversion-indirect"></span><h4><a class="toc-backref" href="#id27">Conversion of pointers to ownership-qualified types</a><a class="headerlink" href="#conversion-of-pointers-to-ownership-qualified-types" title="Permalink to this headline">¶</a></h4>
+<p>A program is ill-formed if an expression of type <code class="docutils literal notranslate"><span class="pre">T*</span></code> is converted,
+explicitly or implicitly, to the type <code class="docutils literal notranslate"><span class="pre">U*</span></code>, where <code class="docutils literal notranslate"><span class="pre">T</span></code> and <code class="docutils literal notranslate"><span class="pre">U</span></code> have
+different ownership qualification, unless:</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">T</span></code> is qualified with <code class="docutils literal notranslate"><span class="pre">__strong</span></code>, <code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code>, or
+<code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code>, and <code class="docutils literal notranslate"><span class="pre">U</span></code> is qualified with both <code class="docutils literal notranslate"><span class="pre">const</span></code> and
+<code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code>; or</li>
+<li>either <code class="docutils literal notranslate"><span class="pre">T</span></code> or <code class="docutils literal notranslate"><span class="pre">U</span></code> is <code class="docutils literal notranslate"><span class="pre">cv</span> <span class="pre">void</span></code>, where <code class="docutils literal notranslate"><span class="pre">cv</span></code> is an optional sequence
+of non-ownership qualifiers; or</li>
+<li>the conversion is requested with a <code class="docutils literal notranslate"><span class="pre">reinterpret_cast</span></code> in Objective-C++; or</li>
+<li>the conversion is a well-formed <a class="reference internal" href="#arc-ownership-restrictions-pass-by-writeback"><span class="std std-ref">pass-by-writeback</span></a>.</li>
+</ul>
+<p>The analogous rule applies to <code class="docutils literal notranslate"><span class="pre">T&</span></code> and <code class="docutils literal notranslate"><span class="pre">U&</span></code> in Objective-C++.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">These rules provide a reasonable level of type-safety for indirect pointers,
+as long as the underlying memory is not deallocated. The conversion to
+<code class="docutils literal notranslate"><span class="pre">const</span> <span class="pre">__unsafe_unretained</span></code> is permitted because the semantics of reads are
+equivalent across all these ownership semantics, and thatâs a very useful and
+common pattern. The interconversion with <code class="docutils literal notranslate"><span class="pre">void*</span></code> is useful for allocating
+memory or otherwise escaping the type system, but use it carefully.
+<code class="docutils literal notranslate"><span class="pre">reinterpret_cast</span></code> is considered to be an obvious enough sign of taking
+responsibility for any problems.</p>
+</div>
+<p>It is undefined behavior to access an ownership-qualified object through an
+lvalue of a differently-qualified type, except that any non-<code class="docutils literal notranslate"><span class="pre">__weak</span></code> object
+may be read through an <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code> lvalue.</p>
+<p>It is undefined behavior if the storage of a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> or <code class="docutils literal notranslate"><span class="pre">__weak</span></code>
+object is not properly initialized before the first managed operation
+is performed on the object, or if the storage of such an object is freed
+or reused before the object has been properly deinitialized. Storage for
+a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> or <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object may be properly initialized by filling
+it with the representation of a null pointer, e.g. by acquiring the memory
+with <code class="docutils literal notranslate"><span class="pre">calloc</span></code> or using <code class="docutils literal notranslate"><span class="pre">bzero</span></code> to zero it out. A <code class="docutils literal notranslate"><span class="pre">__strong</span></code> or
+<code class="docutils literal notranslate"><span class="pre">__weak</span></code> object may be properly deinitialized by assigning a null pointer
+into it. A <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object may also be properly initialized
+by copying into it (e.g. with <code class="docutils literal notranslate"><span class="pre">memcpy</span></code>) the representation of a
+different <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object whose storage has been properly initialized;
+doing this properly deinitializes the source object and causes its storage
+to no longer be properly initialized. A <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object may not be
+representation-copied in this way.</p>
+<p>These requirements are followed automatically for objects whose
+initialization and deinitialization are under the control of ARC:</p>
+<ul class="simple">
+<li>objects of static, automatic, and temporary storage duration</li>
+<li>instance variables of Objective-C objects</li>
+<li>elements of arrays where the array objectâs initialization and
+deinitialization are under the control of ARC</li>
+<li>fields of Objective-C struct types where the struct objectâs
+initialization and deinitialization are under the control of ARC</li>
+<li>non-static data members of Objective-C++ non-union class types</li>
+<li>Objective-C++ objects and arrays of dynamic storage duration created
+with the <code class="docutils literal notranslate"><span class="pre">new</span></code> or <code class="docutils literal notranslate"><span class="pre">new[]</span></code> operators and destroyed with the
+corresponding <code class="docutils literal notranslate"><span class="pre">delete</span></code> or <code class="docutils literal notranslate"><span class="pre">delete[]</span></code> operator</li>
+</ul>
+<p>They are not followed automatically for these objects:</p>
+<ul class="simple">
+<li>objects of dynamic storage duration created in other memory, such as
+that returned by <code class="docutils literal notranslate"><span class="pre">malloc</span></code></li>
+<li>union members</li>
+</ul>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">ARC must perform special operations when initializing an object and
+when destroying it. In many common situations, ARC knows when an
+object is created and when it is destroyed and can ensure that these
+operations are performed correctly. Otherwise, however, ARC requires
+programmer cooperation to establish its initialization invariants
+because it is infeasible for ARC to dynamically infer whether they
+are intact. For example, there is no syntactic difference in C between
+an assignment that is intended by the programmer to initialize a variable
+and one that is intended to replace the existing value stored there,
+but ARC must perform one operation or the other. ARC chooses to always
+assume that objects are initialized (except when it is in charge of
+initializing them) because the only workable alternative would be to
+ban all code patterns that could potentially be used to access
+uninitialized memory, and that would be too limiting. In practice,
+this is rarely a problem because programmers do not generally need to
+work with objects for which the requirements are not handled
+automatically.</p>
+</div>
+<p>Note that dynamically-allocated Objective-C++ arrays of
+nontrivially-ownership-qualified type are not ABI-compatible with non-ARC
+code because the non-ARC code will consider the element type to be POD.
+Such arrays that are <code class="docutils literal notranslate"><span class="pre">new[]</span></code>âd in ARC translation units cannot be
+<code class="docutils literal notranslate"><span class="pre">delete[]</span></code>âd in non-ARC translation units and vice-versa.</p>
+</div>
+<div class="section" id="passing-to-an-out-parameter-by-writeback">
+<span id="arc-ownership-restrictions-pass-by-writeback"></span><h4><a class="toc-backref" href="#id28">Passing to an out parameter by writeback</a><a class="headerlink" href="#passing-to-an-out-parameter-by-writeback" title="Permalink to this headline">¶</a></h4>
+<p>If the argument passed to a parameter of type <code class="docutils literal notranslate"><span class="pre">T</span> <span class="pre">__autoreleasing</span> <span class="pre">*</span></code> has type
+<code class="docutils literal notranslate"><span class="pre">U</span> <span class="pre">oq</span> <span class="pre">*</span></code>, where <code class="docutils literal notranslate"><span class="pre">oq</span></code> is an ownership qualifier, then the argument is a
+candidate for <span class="arc-term">pass-by-writeback`</span> if:</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">oq</span></code> is <code class="docutils literal notranslate"><span class="pre">__strong</span></code> or <code class="docutils literal notranslate"><span class="pre">__weak</span></code>, and</li>
+<li>it would be legal to initialize a <code class="docutils literal notranslate"><span class="pre">T</span> <span class="pre">__strong</span> <span class="pre">*</span></code> with a <code class="docutils literal notranslate"><span class="pre">U</span> <span class="pre">__strong</span> <span class="pre">*</span></code>.</li>
+</ul>
+<p>For purposes of overload resolution, an implicit conversion sequence requiring
+a pass-by-writeback is always worse than an implicit conversion sequence not
+requiring a pass-by-writeback.</p>
+<p>The pass-by-writeback is ill-formed if the argument expression does not have a
+legal form:</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">&var</span></code>, where <code class="docutils literal notranslate"><span class="pre">var</span></code> is a scalar variable of automatic storage duration
+with retainable object pointer type</li>
+<li>a conditional expression where the second and third operands are both legal
+forms</li>
+<li>a cast whose operand is a legal form</li>
+<li>a null pointer constant</li>
+</ul>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">The restriction in the form of the argument serves two purposes. First, it
+makes it impossible to pass the address of an array to the argument, which
+serves to protect against an otherwise serious risk of mis-inferring an
+âarrayâ argument as an out-parameter. Second, it makes it much less likely
+that the user will see confusing aliasing problems due to the implementation,
+below, where their store to the writeback temporary is not immediately seen
+in the original argument variable.</p>
+</div>
+<p>A pass-by-writeback is evaluated as follows:</p>
+<ol class="arabic simple">
+<li>The argument is evaluated to yield a pointer <code class="docutils literal notranslate"><span class="pre">p</span></code> of type <code class="docutils literal notranslate"><span class="pre">U</span> <span class="pre">oq</span> <span class="pre">*</span></code>.</li>
+<li>If <code class="docutils literal notranslate"><span class="pre">p</span></code> is a null pointer, then a null pointer is passed as the argument,
+and no further work is required for the pass-by-writeback.</li>
+<li>Otherwise, a temporary of type <code class="docutils literal notranslate"><span class="pre">T</span> <span class="pre">__autoreleasing</span></code> is created and
+initialized to a null pointer.</li>
+<li>If the parameter is not an Objective-C method parameter marked <code class="docutils literal notranslate"><span class="pre">out</span></code>,
+then <code class="docutils literal notranslate"><span class="pre">*p</span></code> is read, and the result is written into the temporary with
+primitive semantics.</li>
+<li>The address of the temporary is passed as the argument to the actual call.</li>
+<li>After the call completes, the temporary is loaded with primitive
+semantics, and that value is assigned into <code class="docutils literal notranslate"><span class="pre">*p</span></code>.</li>
+</ol>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">This is all admittedly convoluted. In an ideal world, we would see that a
+local variable is being passed to an out-parameter and retroactively modify
+its type to be <code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> rather than <code class="docutils literal notranslate"><span class="pre">__strong</span></code>. This would be
+remarkably difficult and not always well-founded under the C type system.
+However, it was judged unacceptably invasive to require programmers to write
+<code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> on all the variables they intend to use for
+out-parameters. This was the least bad solution.</p>
+</div>
+</div>
+<div class="section" id="ownership-qualified-fields-of-structs-and-unions">
+<span id="arc-ownership-restrictions-records"></span><h4><a class="toc-backref" href="#id29">Ownership-qualified fields of structs and unions</a><a class="headerlink" href="#ownership-qualified-fields-of-structs-and-unions" title="Permalink to this headline">¶</a></h4>
+<p>A program is ill-formed if it declares a member of a C struct or union to have
+a nontrivially ownership-qualified type.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">The resulting type would be non-POD in the C++ sense, but C does not give us
+very good language tools for managing the lifetime of aggregates, so it is
+more convenient to simply forbid them. It is still possible to manage this
+with a <code class="docutils literal notranslate"><span class="pre">void*</span></code> or an <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code> object.</p>
+</div>
+<p>This restriction does not apply in Objective-C++. However, nontrivally
+ownership-qualified types are considered non-POD: in C++11 terms, they are not
+trivially default constructible, copy constructible, move constructible, copy
+assignable, move assignable, or destructible. It is a violation of C++âs One
+Definition Rule to use a class outside of ARC that, under ARC, would have a
+nontrivially ownership-qualified member.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Unlike in C, we can express all the necessary ARC semantics for
+ownership-qualified subobjects as suboperations of the (default) special
+member functions for the class. These functions then become non-trivial.
+This has the non-obvious result that the class will have a non-trivial copy
+constructor and non-trivial destructor; if this would not normally be true
+outside of ARC, objects of the type will be passed and returned in an
+ABI-incompatible manner.</p>
+</div>
+</div>
+</div>
+<div class="section" id="ownership-inference">
+<span id="arc-ownership-inference"></span><h3><a class="toc-backref" href="#id30">Ownership inference</a><a class="headerlink" href="#ownership-inference" title="Permalink to this headline">¶</a></h3>
+<div class="section" id="objects">
+<span id="arc-ownership-inference-variables"></span><h4><a class="toc-backref" href="#id31">Objects</a><a class="headerlink" href="#objects" title="Permalink to this headline">¶</a></h4>
+<p>If an object is declared with retainable object owner type, but without an
+explicit ownership qualifier, its type is implicitly adjusted to have
+<code class="docutils literal notranslate"><span class="pre">__strong</span></code> qualification.</p>
+<p>As a special case, if the objectâs base type is <code class="docutils literal notranslate"><span class="pre">Class</span></code> (possibly
+protocol-qualified), the type is adjusted to have <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code>
+qualification instead.</p>
+</div>
+<div class="section" id="indirect-parameters">
+<span id="arc-ownership-inference-indirect-parameters"></span><h4><a class="toc-backref" href="#id32">Indirect parameters</a><a class="headerlink" href="#indirect-parameters" title="Permalink to this headline">¶</a></h4>
+<p>If a function or method parameter has type <code class="docutils literal notranslate"><span class="pre">T*</span></code>, where <code class="docutils literal notranslate"><span class="pre">T</span></code> is an
+ownership-unqualified retainable object pointer type, then:</p>
+<ul class="simple">
+<li>if <code class="docutils literal notranslate"><span class="pre">T</span></code> is <code class="docutils literal notranslate"><span class="pre">const</span></code>-qualified or <code class="docutils literal notranslate"><span class="pre">Class</span></code>, then it is implicitly
+qualified with <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code>;</li>
+<li>otherwise, it is implicitly qualified with <code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code>.</li>
+</ul>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last"><code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> exists mostly for this case, the Cocoa convention for
+out-parameters. Since a pointer to <code class="docutils literal notranslate"><span class="pre">const</span></code> is obviously not an
+out-parameter, we instead use a type more useful for passing arrays. If the
+user instead intends to pass in a <em>mutable</em> array, inferring
+<code class="docutils literal notranslate"><span class="pre">__autoreleasing</span></code> is the wrong thing to do; this directs some of the
+caution in the following rules about writeback.</p>
+</div>
+<p>Such a type written anywhere else would be ill-formed by the general rule
+requiring ownership qualifiers.</p>
+<p>This rule does not apply in Objective-C++ if a parameterâs type is dependent in
+a template pattern and is only <em>instantiated</em> to a type which would be a
+pointer to an unqualified retainable object pointer type. Such code is still
+ill-formed.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">The convention is very unlikely to be intentional in template code.</p>
+</div>
+</div>
+<div class="section" id="template-arguments">
+<span id="arc-ownership-inference-template-arguments"></span><h4><a class="toc-backref" href="#id33">Template arguments</a><a class="headerlink" href="#template-arguments" title="Permalink to this headline">¶</a></h4>
+<p>If a template argument for a template type parameter is an retainable object
+owner type that does not have an explicit ownership qualifier, it is adjusted
+to have <code class="docutils literal notranslate"><span class="pre">__strong</span></code> qualification. This adjustment occurs regardless of
+whether the template argument was deduced or explicitly specified.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last"><code class="docutils literal notranslate"><span class="pre">__strong</span></code> is a useful default for containers (e.g., <code class="docutils literal notranslate"><span class="pre">std::vector<id></span></code>),
+which would otherwise require explicit qualification. Moreover, unqualified
+retainable object pointer types are unlikely to be useful within templates,
+since they generally need to have a qualifier applied to the before being
+used.</p>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="method-families">
+<span id="arc-method-families"></span><h2><a class="toc-backref" href="#id34">Method families</a><a class="headerlink" href="#method-families" title="Permalink to this headline">¶</a></h2>
+<p>An Objective-C method may fall into a <span class="arc-term">method family</span>, which is a
+conventional set of behaviors ascribed to it by the Cocoa conventions.</p>
+<p>A method is in a certain method family if:</p>
+<ul class="simple">
+<li>it has a <code class="docutils literal notranslate"><span class="pre">objc_method_family</span></code> attribute placing it in that family; or if
+not that,</li>
+<li>it does not have an <code class="docutils literal notranslate"><span class="pre">objc_method_family</span></code> attribute placing it in a
+different or no family, and</li>
+<li>its selector falls into the corresponding selector family, and</li>
+<li>its signature obeys the added restrictions of the method family.</li>
+</ul>
+<p>A selector is in a certain selector family if, ignoring any leading
+underscores, the first component of the selector either consists entirely of
+the name of the method family or it begins with that name followed by a
+character other than a lowercase letter. For example, <code class="docutils literal notranslate"><span class="pre">_perform:with:</span></code> and
+<code class="docutils literal notranslate"><span class="pre">performWith:</span></code> would fall into the <code class="docutils literal notranslate"><span class="pre">perform</span></code> family (if we recognized one),
+but <code class="docutils literal notranslate"><span class="pre">performing:with</span></code> would not.</p>
+<p>The families and their added restrictions are:</p>
+<ul>
+<li><p class="first"><code class="docutils literal notranslate"><span class="pre">alloc</span></code> methods must return a retainable object pointer type.</p>
+</li>
+<li><p class="first"><code class="docutils literal notranslate"><span class="pre">copy</span></code> methods must return a retainable object pointer type.</p>
+</li>
+<li><p class="first"><code class="docutils literal notranslate"><span class="pre">mutableCopy</span></code> methods must return a retainable object pointer type.</p>
+</li>
+<li><p class="first"><code class="docutils literal notranslate"><span class="pre">new</span></code> methods must return a retainable object pointer type.</p>
+</li>
+<li><p class="first"><code class="docutils literal notranslate"><span class="pre">init</span></code> methods must be instance methods and must return an Objective-C
+pointer type. Additionally, a program is ill-formed if it declares or
+contains a call to an <code class="docutils literal notranslate"><span class="pre">init</span></code> method whose return type is neither <code class="docutils literal notranslate"><span class="pre">id</span></code> nor
+a pointer to a super-class or sub-class of the declaring class (if the method
+was declared on a class) or the static receiver type of the call (if it was
+declared on a protocol).</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>There are a fair number of existing methods with <code class="docutils literal notranslate"><span class="pre">init</span></code>-like selectors
+which nonetheless donât follow the <code class="docutils literal notranslate"><span class="pre">init</span></code> conventions. Typically these
+are either accidental naming collisions or helper methods called during
+initialization. Because of the peculiar retain/release behavior of
+<code class="docutils literal notranslate"><span class="pre">init</span></code> methods, itâs very important not to treat these methods as
+<code class="docutils literal notranslate"><span class="pre">init</span></code> methods if they arenât meant to be. It was felt that implicitly
+defining these methods out of the family based on the exact relationship
+between the return type and the declaring class would be much too subtle
+and fragile. Therefore we identify a small number of legitimate-seeming
+return types and call everything else an error. This serves the secondary
+purpose of encouraging programmers not to accidentally give methods names
+in the <code class="docutils literal notranslate"><span class="pre">init</span></code> family.</p>
+<p class="last">Note that a method with an <code class="docutils literal notranslate"><span class="pre">init</span></code>-family selector which returns a
+non-Objective-C type (e.g. <code class="docutils literal notranslate"><span class="pre">void</span></code>) is perfectly well-formed; it simply
+isnât in the <code class="docutils literal notranslate"><span class="pre">init</span></code> family.</p>
+</div>
+</li>
+</ul>
+<p>A program is ill-formed if a methodâs declarations, implementations, and
+overrides do not all have the same method family.</p>
+<div class="section" id="explicit-method-family-control">
+<span id="arc-family-attribute"></span><h3><a class="toc-backref" href="#id35">Explicit method family control</a><a class="headerlink" href="#explicit-method-family-control" title="Permalink to this headline">¶</a></h3>
+<p>A method may be annotated with the <code class="docutils literal notranslate"><span class="pre">objc_method_family</span></code> attribute to
+precisely control which method family it belongs to. If a method in an
+<code class="docutils literal notranslate"><span class="pre">@implementation</span></code> does not have this attribute, but there is a method
+declared in the corresponding <code class="docutils literal notranslate"><span class="pre">@interface</span></code> that does, then the attribute is
+copied to the declaration in the <code class="docutils literal notranslate"><span class="pre">@implementation</span></code>. The attribute is
+available outside of ARC, and may be tested for with the preprocessor query
+<code class="docutils literal notranslate"><span class="pre">__has_attribute(objc_method_family)</span></code>.</p>
+<p>The attribute is spelled
+<code class="docutils literal notranslate"><span class="pre">__attribute__((objc_method_family(</span></code> <em>family</em> <code class="docutils literal notranslate"><span class="pre">)))</span></code>. If <em>family</em> is
+<code class="docutils literal notranslate"><span class="pre">none</span></code>, the method has no family, even if it would otherwise be considered to
+have one based on its selector and type. Otherwise, <em>family</em> must be one of
+<code class="docutils literal notranslate"><span class="pre">alloc</span></code>, <code class="docutils literal notranslate"><span class="pre">copy</span></code>, <code class="docutils literal notranslate"><span class="pre">init</span></code>, <code class="docutils literal notranslate"><span class="pre">mutableCopy</span></code>, or <code class="docutils literal notranslate"><span class="pre">new</span></code>, in which case the
+method is considered to belong to the corresponding family regardless of its
+selector. It is an error if a method that is explicitly added to a family in
+this way does not meet the requirements of the family other than the selector
+naming convention.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">The rules codified in this document describe the standard conventions of
+Objective-C. However, as these conventions have not heretofore been enforced
+by an unforgiving mechanical system, they are only imperfectly kept,
+especially as they havenât always even been precisely defined. While it is
+possible to define low-level ownership semantics with attributes like
+<code class="docutils literal notranslate"><span class="pre">ns_returns_retained</span></code>, this attribute allows the user to communicate
+semantic intent, which is of use both to ARC (which, e.g., treats calls to
+<code class="docutils literal notranslate"><span class="pre">init</span></code> specially) and the static analyzer.</p>
+</div>
+</div>
+<div class="section" id="semantics-of-method-families">
+<span id="arc-family-semantics"></span><h3><a class="toc-backref" href="#id36">Semantics of method families</a><a class="headerlink" href="#semantics-of-method-families" title="Permalink to this headline">¶</a></h3>
+<p>A methodâs membership in a method family may imply non-standard semantics for
+its parameters and return type.</p>
+<p>Methods in the <code class="docutils literal notranslate"><span class="pre">alloc</span></code>, <code class="docutils literal notranslate"><span class="pre">copy</span></code>, <code class="docutils literal notranslate"><span class="pre">mutableCopy</span></code>, and <code class="docutils literal notranslate"><span class="pre">new</span></code> families â
+that is, methods in all the currently-defined families except <code class="docutils literal notranslate"><span class="pre">init</span></code> â
+implicitly <a class="reference internal" href="#arc-object-operands-retained-return-values"><span class="std std-ref">return a retained object</span></a> as if they were annotated with
+the <code class="docutils literal notranslate"><span class="pre">ns_returns_retained</span></code> attribute. This can be overridden by annotating
+the method with either of the <code class="docutils literal notranslate"><span class="pre">ns_returns_autoreleased</span></code> or
+<code class="docutils literal notranslate"><span class="pre">ns_returns_not_retained</span></code> attributes.</p>
+<p>Properties also follow same naming rules as methods. This means that those in
+the <code class="docutils literal notranslate"><span class="pre">alloc</span></code>, <code class="docutils literal notranslate"><span class="pre">copy</span></code>, <code class="docutils literal notranslate"><span class="pre">mutableCopy</span></code>, and <code class="docutils literal notranslate"><span class="pre">new</span></code> families provide access
+to <a class="reference internal" href="#arc-object-operands-retained-return-values"><span class="std std-ref">retained objects</span></a>. This
+can be overridden by annotating the property with <code class="docutils literal notranslate"><span class="pre">ns_returns_not_retained</span></code>
+attribute.</p>
+<div class="section" id="semantics-of-init">
+<span id="arc-family-semantics-init"></span><h4><a class="toc-backref" href="#id37">Semantics of <code class="docutils literal notranslate"><span class="pre">init</span></code></a><a class="headerlink" href="#semantics-of-init" title="Permalink to this headline">¶</a></h4>
+<p>Methods in the <code class="docutils literal notranslate"><span class="pre">init</span></code> family implicitly <a class="reference internal" href="#arc-objects-operands-consumed"><span class="std std-ref">consume</span></a> their <code class="docutils literal notranslate"><span class="pre">self</span></code> parameter and <a class="reference internal" href="#arc-object-operands-retained-return-values"><span class="std std-ref">return a
+retained object</span></a>. Neither of
+these properties can be altered through attributes.</p>
+<p>A call to an <code class="docutils literal notranslate"><span class="pre">init</span></code> method with a receiver that is either <code class="docutils literal notranslate"><span class="pre">self</span></code> (possibly
+parenthesized or casted) or <code class="docutils literal notranslate"><span class="pre">super</span></code> is called a <span class="arc-term">delegate init
+call</span>. It is an error for a delegate init call to be made except from an
+<code class="docutils literal notranslate"><span class="pre">init</span></code> method, and excluding blocks within such methods.</p>
+<p>As an exception to the <a class="reference internal" href="#arc-misc-self"><span class="std std-ref">usual rule</span></a>, the variable <code class="docutils literal notranslate"><span class="pre">self</span></code>
+is mutable in an <code class="docutils literal notranslate"><span class="pre">init</span></code> method and has the usual semantics for a <code class="docutils literal notranslate"><span class="pre">__strong</span></code>
+variable. However, it is undefined behavior and the program is ill-formed, no
+diagnostic required, if an <code class="docutils literal notranslate"><span class="pre">init</span></code> method attempts to use the previous value
+of <code class="docutils literal notranslate"><span class="pre">self</span></code> after the completion of a delegate init call. It is conventional,
+but not required, for an <code class="docutils literal notranslate"><span class="pre">init</span></code> method to return <code class="docutils literal notranslate"><span class="pre">self</span></code>.</p>
+<p>It is undefined behavior for a program to cause two or more calls to <code class="docutils literal notranslate"><span class="pre">init</span></code>
+methods on the same object, except that each <code class="docutils literal notranslate"><span class="pre">init</span></code> method invocation may
+perform at most one delegate init call.</p>
+</div>
+<div class="section" id="related-result-types">
+<span id="arc-family-semantics-result-type"></span><h4><a class="toc-backref" href="#id38">Related result types</a><a class="headerlink" href="#related-result-types" title="Permalink to this headline">¶</a></h4>
+<p>Certain methods are candidates to have <span class="arc-term">related result types</span>:</p>
+<ul class="simple">
+<li>class methods in the <code class="docutils literal notranslate"><span class="pre">alloc</span></code> and <code class="docutils literal notranslate"><span class="pre">new</span></code> method families</li>
+<li>instance methods in the <code class="docutils literal notranslate"><span class="pre">init</span></code> family</li>
+<li>the instance method <code class="docutils literal notranslate"><span class="pre">self</span></code></li>
+<li>outside of ARC, the instance methods <code class="docutils literal notranslate"><span class="pre">retain</span></code> and <code class="docutils literal notranslate"><span class="pre">autorelease</span></code></li>
+</ul>
+<p>If the formal result type of such a method is <code class="docutils literal notranslate"><span class="pre">id</span></code> or protocol-qualified
+<code class="docutils literal notranslate"><span class="pre">id</span></code>, or a type equal to the declaring class or a superclass, then it is said
+to have a related result type. In this case, when invoked in an explicit
+message send, it is assumed to return a type related to the type of the
+receiver:</p>
+<ul class="simple">
+<li>if it is a class method, and the receiver is a class name <code class="docutils literal notranslate"><span class="pre">T</span></code>, the message
+send expression has type <code class="docutils literal notranslate"><span class="pre">T*</span></code>; otherwise</li>
+<li>if it is an instance method, and the receiver has type <code class="docutils literal notranslate"><span class="pre">T</span></code>, the message
+send expression has type <code class="docutils literal notranslate"><span class="pre">T</span></code>; otherwise</li>
+<li>the message send expression has the normal result type of the method.</li>
+</ul>
+<p>This is a new rule of the Objective-C language and applies outside of ARC.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">ARCâs automatic code emission is more prone than most code to signature
+errors, i.e. errors where a call was emitted against one method signature,
+but the implementing method has an incompatible signature. Having more
+precise type information helps drastically lower this risk, as well as
+catching a number of latent bugs.</p>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="optimization">
+<span id="arc-optimization"></span><h2><a class="toc-backref" href="#id39">Optimization</a><a class="headerlink" href="#optimization" title="Permalink to this headline">¶</a></h2>
+<p>Within this section, the word <span class="arc-term">function</span> will be used to
+refer to any structured unit of code, be it a C function, an
+Objective-C method, or a block.</p>
+<p>This specification describes ARC as performing specific <code class="docutils literal notranslate"><span class="pre">retain</span></code> and
+<code class="docutils literal notranslate"><span class="pre">release</span></code> operations on retainable object pointers at specific
+points during the execution of a program. These operations make up a
+non-contiguous subsequence of the computation history of the program.
+The portion of this sequence for a particular retainable object
+pointer for which a specific function execution is directly
+responsible is the <span class="arc-term">formal local retain history</span> of the
+object pointer. The corresponding actual sequence executed is the
+<cite>dynamic local retain history</cite>.</p>
+<p>However, under certain circumstances, ARC is permitted to re-order and
+eliminate operations in a manner which may alter the overall
+computation history beyond what is permitted by the general âas ifâ
+rule of C/C++ and the <a class="reference internal" href="#arc-objects-retains"><span class="std std-ref">restrictions</span></a> on
+the implementation of <code class="docutils literal notranslate"><span class="pre">retain</span></code> and <code class="docutils literal notranslate"><span class="pre">release</span></code>.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>Specifically, ARC is sometimes permitted to optimize <code class="docutils literal notranslate"><span class="pre">release</span></code>
+operations in ways which might cause an object to be deallocated
+before it would otherwise be. Without this, it would be almost
+impossible to eliminate any <code class="docutils literal notranslate"><span class="pre">retain</span></code>/<code class="docutils literal notranslate"><span class="pre">release</span></code> pairs. For
+example, consider the following code:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">id</span> <span class="n">x</span> <span class="o">=</span> <span class="n">_ivar</span><span class="p">;</span>
+<span class="p">[</span><span class="n">x</span> <span class="n">foo</span><span class="p">];</span>
+</pre></div>
+</div>
+<p class="last">If we were not permitted in any event to shorten the lifetime of the
+object in <code class="docutils literal notranslate"><span class="pre">x</span></code>, then we would not be able to eliminate this retain
+and release unless we could prove that the message send could not
+modify <code class="docutils literal notranslate"><span class="pre">_ivar</span></code> (or deallocate <code class="docutils literal notranslate"><span class="pre">self</span></code>). Since message sends are
+opaque to the optimizer, this is not possible, and so ARCâs hands
+would be almost completely tied.</p>
+</div>
+<p>ARC makes no guarantees about the execution of a computation history
+which contains undefined behavior. In particular, ARC makes no
+guarantees in the presence of race conditions.</p>
+<p>ARC may assume that any retainable object pointers it receives or
+generates are instantaneously valid from that point until a point
+which, by the concurrency model of the host language, happens-after
+the generation of the pointer and happens-before a release of that
+object (possibly via an aliasing pointer or indirectly due to
+destruction of a different object).</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">There is very little point in trying to guarantee correctness in the
+presence of race conditions. ARC does not have a stack-scanning
+garbage collector, and guaranteeing the atomicity of every load and
+store operation would be prohibitive and preclude a vast amount of
+optimization.</p>
+</div>
+<p>ARC may assume that non-ARC code engages in sensible balancing
+behavior and does not rely on exact or minimum retain count values
+except as guaranteed by <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object invariants or +1 transfer
+conventions. For example, if an object is provably double-retained
+and double-released, ARC may eliminate the inner retain and release;
+it does not need to guard against code which performs an unbalanced
+release followed by a âbalancingâ retain.</p>
+<div class="section" id="object-liveness">
+<span id="arc-optimization-liveness"></span><h3><a class="toc-backref" href="#id40">Object liveness</a><a class="headerlink" href="#object-liveness" title="Permalink to this headline">¶</a></h3>
+<p>ARC may not allow a retainable object <code class="docutils literal notranslate"><span class="pre">X</span></code> to be deallocated at a
+time <code class="docutils literal notranslate"><span class="pre">T</span></code> in a computation history if:</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">X</span></code> is the value stored in a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object <code class="docutils literal notranslate"><span class="pre">S</span></code> with
+<a class="reference internal" href="#arc-optimization-precise"><span class="std std-ref">precise lifetime semantics</span></a>, or</li>
+<li><code class="docutils literal notranslate"><span class="pre">X</span></code> is the value stored in a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object <code class="docutils literal notranslate"><span class="pre">S</span></code> with
+imprecise lifetime semantics and, at some point after <code class="docutils literal notranslate"><span class="pre">T</span></code> but
+before the next store to <code class="docutils literal notranslate"><span class="pre">S</span></code>, the computation history features a
+load from <code class="docutils literal notranslate"><span class="pre">S</span></code> and in some way depends on the value loaded, or</li>
+<li><code class="docutils literal notranslate"><span class="pre">X</span></code> is a value described as being released at the end of the
+current full-expression and, at some point after <code class="docutils literal notranslate"><span class="pre">T</span></code> but before
+the end of the full-expression, the computation history depends
+on that value.</li>
+</ul>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>The intent of the second rule is to say that objects held in normal
+<code class="docutils literal notranslate"><span class="pre">__strong</span></code> local variables may be released as soon as the value in
+the variable is no longer being used: either the variable stops
+being used completely or a new value is stored in the variable.</p>
+<p class="last">The intent of the third rule is to say that return values may be
+released after theyâve been used.</p>
+</div>
+<p>A computation history depends on a pointer value <code class="docutils literal notranslate"><span class="pre">P</span></code> if it:</p>
+<ul class="simple">
+<li>performs a pointer comparison with <code class="docutils literal notranslate"><span class="pre">P</span></code>,</li>
+<li>loads from <code class="docutils literal notranslate"><span class="pre">P</span></code>,</li>
+<li>stores to <code class="docutils literal notranslate"><span class="pre">P</span></code>,</li>
+<li>depends on a pointer value <code class="docutils literal notranslate"><span class="pre">Q</span></code> derived via pointer arithmetic
+from <code class="docutils literal notranslate"><span class="pre">P</span></code> (including an instance-variable or field access), or</li>
+<li>depends on a pointer value <code class="docutils literal notranslate"><span class="pre">Q</span></code> loaded from <code class="docutils literal notranslate"><span class="pre">P</span></code>.</li>
+</ul>
+<p>Dependency applies only to values derived directly or indirectly from
+a particular expression result and does not occur merely because a
+separate pointer value dynamically aliases <code class="docutils literal notranslate"><span class="pre">P</span></code>. Furthermore, this
+dependency is not carried by values that are stored to objects.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>The restrictions on dependency are intended to make this analysis
+feasible by an optimizer with only incomplete information about a
+program. Essentially, dependence is carried to âobviousâ uses of a
+pointer. Merely passing a pointer argument to a function does not
+itself cause dependence, but since generally the optimizer will not
+be able to prove that the function doesnât depend on that parameter,
+it will be forced to conservatively assume it does.</p>
+<p>Dependency propagates to values loaded from a pointer because those
+values might be invalidated by deallocating the object. For
+example, given the code <code class="docutils literal notranslate"><span class="pre">__strong</span> <span class="pre">id</span> <span class="pre">x</span> <span class="pre">=</span> <span class="pre">p->ivar;</span></code>, ARC must not
+move the release of <code class="docutils literal notranslate"><span class="pre">p</span></code> to between the load of <code class="docutils literal notranslate"><span class="pre">p->ivar</span></code> and the
+retain of that value for storing into <code class="docutils literal notranslate"><span class="pre">x</span></code>.</p>
+<p>Dependency does not propagate through stores of dependent pointer
+values because doing so would allow dependency to outlive the
+full-expression which produced the original value. For example, the
+address of an instance variable could be written to some global
+location and then freely accessed during the lifetime of the local,
+or a function could return an inner pointer of an object and store
+it to a local. These cases would be potentially impossible to
+reason about and so would basically prevent any optimizations based
+on imprecise lifetime. There are also uncommon enough to make it
+reasonable to require the precise-lifetime annotation if someone
+really wants to rely on them.</p>
+<p class="last">Dependency does propagate through return values of pointer type.
+The compelling source of need for this rule is a property accessor
+which returns an un-autoreleased result; the calling function must
+have the chance to operate on the value, e.g. to retain it, before
+ARC releases the original pointer. Note again, however, that
+dependence does not survive a store, so ARC does not guarantee the
+continued validity of the return value past the end of the
+full-expression.</p>
+</div>
+</div>
+<div class="section" id="no-object-lifetime-extension">
+<span id="arc-optimization-object-lifetime"></span><h3><a class="toc-backref" href="#id41">No object lifetime extension</a><a class="headerlink" href="#no-object-lifetime-extension" title="Permalink to this headline">¶</a></h3>
+<p>If, in the formal computation history of the program, an object <code class="docutils literal notranslate"><span class="pre">X</span></code>
+has been deallocated by the time of an observable side-effect, then
+ARC must cause <code class="docutils literal notranslate"><span class="pre">X</span></code> to be deallocated by no later than the occurrence
+of that side-effect, except as influenced by the re-ordering of the
+destruction of objects.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>This rule is intended to prohibit ARC from observably extending the
+lifetime of a retainable object, other than as specified in this
+document. Together with the rule limiting the transformation of
+releases, this rule requires ARC to eliminate retains and release
+only in pairs.</p>
+<p class="last">ARCâs power to reorder the destruction of objects is critical to its
+ability to do any optimization, for essentially the same reason that
+it must retain the power to decrease the lifetime of an object.
+Unfortunately, while itâs generally poor style for the destruction
+of objects to have arbitrary side-effects, itâs certainly possible.
+Hence the caveat.</p>
+</div>
+</div>
+<div class="section" id="precise-lifetime-semantics">
+<span id="arc-optimization-precise"></span><h3><a class="toc-backref" href="#id42">Precise lifetime semantics</a><a class="headerlink" href="#precise-lifetime-semantics" title="Permalink to this headline">¶</a></h3>
+<p>In general, ARC maintains an invariant that a retainable object pointer held in
+a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object will be retained for the full formal lifetime of the
+object. Objects subject to this invariant have <span class="arc-term">precise lifetime
+semantics</span>.</p>
+<p>By default, local variables of automatic storage duration do not have precise
+lifetime semantics. Such objects are simply strong references which hold
+values of retainable object pointer type, and these values are still fully
+subject to the optimizations on values under local control.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Applying these precise-lifetime semantics strictly would be prohibitive.
+Many useful optimizations that might theoretically decrease the lifetime of
+an object would be rendered impossible. Essentially, it promises too much.</p>
+</div>
+<p>A local variable of retainable object owner type and automatic storage duration
+may be annotated with the <code class="docutils literal notranslate"><span class="pre">objc_precise_lifetime</span></code> attribute to indicate that
+it should be considered to be an object with precise lifetime semantics.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Nonetheless, it is sometimes useful to be able to force an object to be
+released at a precise time, even if that object does not appear to be used.
+This is likely to be uncommon enough that the syntactic weight of explicitly
+requesting these semantics will not be burdensome, and may even make the code
+clearer.</p>
+</div>
+</div>
+</div>
+<div class="section" id="miscellaneous">
+<span id="arc-misc"></span><h2><a class="toc-backref" href="#id43">Miscellaneous</a><a class="headerlink" href="#miscellaneous" title="Permalink to this headline">¶</a></h2>
+<div class="section" id="special-methods">
+<span id="arc-misc-special-methods"></span><h3><a class="toc-backref" href="#id44">Special methods</a><a class="headerlink" href="#special-methods" title="Permalink to this headline">¶</a></h3>
+<div class="section" id="memory-management-methods">
+<span id="arc-misc-special-methods-retain"></span><h4><a class="toc-backref" href="#id45">Memory management methods</a><a class="headerlink" href="#memory-management-methods" title="Permalink to this headline">¶</a></h4>
+<p>A program is ill-formed if it contains a method definition, message send, or
+<code class="docutils literal notranslate"><span class="pre">@selector</span></code> expression for any of the following selectors:</p>
+<ul class="simple">
+<li><code class="docutils literal notranslate"><span class="pre">autorelease</span></code></li>
+<li><code class="docutils literal notranslate"><span class="pre">release</span></code></li>
+<li><code class="docutils literal notranslate"><span class="pre">retain</span></code></li>
+<li><code class="docutils literal notranslate"><span class="pre">retainCount</span></code></li>
+</ul>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p><code class="docutils literal notranslate"><span class="pre">retainCount</span></code> is banned because ARC robs it of consistent semantics. The
+others were banned after weighing three options for how to deal with message
+sends:</p>
+<p><strong>Honoring</strong> them would work out very poorly if a programmer naively or
+accidentally tried to incorporate code written for manual retain/release code
+into an ARC program. At best, such code would do twice as much work as
+necessary; quite frequently, however, ARC and the explicit code would both
+try to balance the same retain, leading to crashes. The cost is losing the
+ability to perform âunrootedâ retains, i.e. retains not logically
+corresponding to a strong reference in the object graph.</p>
+<p><strong>Ignoring</strong> them would badly violate user expectations about their code.
+While it <em>would</em> make it easier to develop code simultaneously for ARC and
+non-ARC, there is very little reason to do so except for certain library
+developers. ARC and non-ARC translation units share an execution model and
+can seamlessly interoperate. Within a translation unit, a developer who
+faithfully maintains their code in non-ARC mode is suffering all the
+restrictions of ARC for zero benefit, while a developer who isnât testing the
+non-ARC mode is likely to be unpleasantly surprised if they try to go back to
+it.</p>
+<p><strong>Banning</strong> them has the disadvantage of making it very awkward to migrate
+existing code to ARC. The best answer to that, given a number of other
+changes and restrictions in ARC, is to provide a specialized tool to assist
+users in that migration.</p>
+<p class="last">Implementing these methods was banned because they are too integral to the
+semantics of ARC; many tricks which worked tolerably under manual reference
+counting will misbehave if ARC performs an ephemeral extra retain or two. If
+absolutely required, it is still possible to implement them in non-ARC code,
+for example in a category; the implementations must obey the <a class="reference internal" href="#arc-objects-retains"><span class="std std-ref">semantics</span></a> laid out elsewhere in this document.</p>
+</div>
+</div>
+<div class="section" id="dealloc">
+<span id="arc-misc-special-methods-dealloc"></span><h4><a class="toc-backref" href="#id46"><code class="docutils literal notranslate"><span class="pre">dealloc</span></code></a><a class="headerlink" href="#dealloc" title="Permalink to this headline">¶</a></h4>
+<p>A program is ill-formed if it contains a message send or <code class="docutils literal notranslate"><span class="pre">@selector</span></code>
+expression for the selector <code class="docutils literal notranslate"><span class="pre">dealloc</span></code>.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">There are no legitimate reasons to call <code class="docutils literal notranslate"><span class="pre">dealloc</span></code> directly.</p>
+</div>
+<p>A class may provide a method definition for an instance method named
+<code class="docutils literal notranslate"><span class="pre">dealloc</span></code>. This method will be called after the final <code class="docutils literal notranslate"><span class="pre">release</span></code> of the
+object but before it is deallocated or any of its instance variables are
+destroyed. The superclassâs implementation of <code class="docutils literal notranslate"><span class="pre">dealloc</span></code> will be called
+automatically when the method returns.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Even though ARC destroys instance variables automatically, there are still
+legitimate reasons to write a <code class="docutils literal notranslate"><span class="pre">dealloc</span></code> method, such as freeing
+non-retainable resources. Failing to call <code class="docutils literal notranslate"><span class="pre">[super</span> <span class="pre">dealloc]</span></code> in such a
+method is nearly always a bug. Sometimes, the object is simply trying to
+prevent itself from being destroyed, but <code class="docutils literal notranslate"><span class="pre">dealloc</span></code> is really far too late
+for the object to be raising such objections. Somewhat more legitimately, an
+object may have been pool-allocated and should not be deallocated with
+<code class="docutils literal notranslate"><span class="pre">free</span></code>; for now, this can only be supported with a <code class="docutils literal notranslate"><span class="pre">dealloc</span></code>
+implementation outside of ARC. Such an implementation must be very careful
+to do all the other work that <code class="docutils literal notranslate"><span class="pre">NSObject</span></code>âs <code class="docutils literal notranslate"><span class="pre">dealloc</span></code> would, which is
+outside the scope of this document to describe.</p>
+</div>
+<p>The instance variables for an ARC-compiled class will be destroyed at some
+point after control enters the <code class="docutils literal notranslate"><span class="pre">dealloc</span></code> method for the root class of the
+class. The ordering of the destruction of instance variables is unspecified,
+both within a single class and between subclasses and superclasses.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>The traditional, non-ARC pattern for destroying instance variables is to
+destroy them immediately before calling <code class="docutils literal notranslate"><span class="pre">[super</span> <span class="pre">dealloc]</span></code>. Unfortunately,
+message sends from the superclass are quite capable of reaching methods in
+the subclass, and those methods may well read or write to those instance
+variables. Making such message sends from dealloc is generally discouraged,
+since the subclass may well rely on other invariants that were broken during
+<code class="docutils literal notranslate"><span class="pre">dealloc</span></code>, but itâs not so inescapably dangerous that we felt comfortable
+calling it undefined behavior. Therefore we chose to delay destroying the
+instance variables to a point at which message sends are clearly disallowed:
+the point at which the root classâs deallocation routines take over.</p>
+<p class="last">In most code, the difference is not observable. It can, however, be observed
+if an instance variable holds a strong reference to an object whose
+deallocation will trigger a side-effect which must be carefully ordered with
+respect to the destruction of the super class. Such code violates the design
+principle that semantically important behavior should be explicit. A simple
+fix is to clear the instance variable manually during <code class="docutils literal notranslate"><span class="pre">dealloc</span></code>; a more
+holistic solution is to move semantically important side-effects out of
+<code class="docutils literal notranslate"><span class="pre">dealloc</span></code> and into a separate teardown phase which can rely on working with
+well-formed objects.</p>
+</div>
+</div>
+</div>
+<div class="section" id="autoreleasepool">
+<span id="arc-misc-autoreleasepool"></span><h3><a class="toc-backref" href="#id47"><code class="docutils literal notranslate"><span class="pre">@autoreleasepool</span></code></a><a class="headerlink" href="#autoreleasepool" title="Permalink to this headline">¶</a></h3>
+<p>To simplify the use of autorelease pools, and to bring them under the control
+of the compiler, a new kind of statement is available in Objective-C. It is
+written <code class="docutils literal notranslate"><span class="pre">@autoreleasepool</span></code> followed by a <em>compound-statement</em>, i.e. by a new
+scope delimited by curly braces. Upon entry to this block, the current state
+of the autorelease pool is captured. When the block is exited normally,
+whether by fallthrough or directed control flow (such as <code class="docutils literal notranslate"><span class="pre">return</span></code> or
+<code class="docutils literal notranslate"><span class="pre">break</span></code>), the autorelease pool is restored to the saved state, releasing all
+the objects in it. When the block is exited with an exception, the pool is not
+drained.</p>
+<p><code class="docutils literal notranslate"><span class="pre">@autoreleasepool</span></code> may be used in non-ARC translation units, with equivalent
+semantics.</p>
+<p>A program is ill-formed if it refers to the <code class="docutils literal notranslate"><span class="pre">NSAutoreleasePool</span></code> class.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Autorelease pools are clearly important for the compiler to reason about, but
+it is far too much to expect the compiler to accurately reason about control
+dependencies between two calls. It is also very easy to accidentally forget
+to drain an autorelease pool when using the manual API, and this can
+significantly inflate the processâs high-water-mark. The introduction of a
+new scope is unfortunate but basically required for sane interaction with the
+rest of the language. Not draining the pool during an unwind is apparently
+required by the Objective-C exceptions implementation.</p>
+</div>
+</div>
+<div class="section" id="self">
+<span id="arc-misc-self"></span><h3><a class="toc-backref" href="#id48"><code class="docutils literal notranslate"><span class="pre">self</span></code></a><a class="headerlink" href="#self" title="Permalink to this headline">¶</a></h3>
+<p>The <code class="docutils literal notranslate"><span class="pre">self</span></code> parameter variable of an Objective-C method is never actually
+retained by the implementation. It is undefined behavior, or at least
+dangerous, to cause an object to be deallocated during a message send to that
+object.</p>
+<p>To make this safe, for Objective-C instance methods <code class="docutils literal notranslate"><span class="pre">self</span></code> is implicitly
+<code class="docutils literal notranslate"><span class="pre">const</span></code> unless the method is in the <a class="reference internal" href="#arc-family-semantics-init"><span class="std std-ref">init family</span></a>. Further, <code class="docutils literal notranslate"><span class="pre">self</span></code> is <strong>always</strong> implicitly
+<code class="docutils literal notranslate"><span class="pre">const</span></code> within a class method.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">The cost of retaining <code class="docutils literal notranslate"><span class="pre">self</span></code> in all methods was found to be prohibitive, as
+it tends to be live across calls, preventing the optimizer from proving that
+the retain and release are unnecessary â for good reason, as itâs quite
+possible in theory to cause an object to be deallocated during its execution
+without this retain and release. Since itâs extremely uncommon to actually
+do so, even unintentionally, and since thereâs no natural way for the
+programmer to remove this retain/release pair otherwise (as there is for
+other parameters by, say, making the variable <code class="docutils literal notranslate"><span class="pre">__unsafe_unretained</span></code>), we
+chose to make this optimizing assumption and shift some amount of risk to the
+user.</p>
+</div>
+</div>
+<div class="section" id="fast-enumeration-iteration-variables">
+<span id="arc-misc-enumeration"></span><h3><a class="toc-backref" href="#id49">Fast enumeration iteration variables</a><a class="headerlink" href="#fast-enumeration-iteration-variables" title="Permalink to this headline">¶</a></h3>
+<p>If a variable is declared in the condition of an Objective-C fast enumeration
+loop, and the variable has no explicit ownership qualifier, then it is
+qualified with <code class="docutils literal notranslate"><span class="pre">const</span> <span class="pre">__strong</span></code> and objects encountered during the
+enumeration are not actually retained.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">This is an optimization made possible because fast enumeration loops promise
+to keep the objects retained during enumeration, and the collection itself
+cannot be synchronously modified. It can be overridden by explicitly
+qualifying the variable with <code class="docutils literal notranslate"><span class="pre">__strong</span></code>, which will make the variable
+mutable again and cause the loop to retain the objects it encounters.</p>
+</div>
+</div>
+<div class="section" id="blocks">
+<span id="arc-misc-blocks"></span><h3><a class="toc-backref" href="#id50">Blocks</a><a class="headerlink" href="#blocks" title="Permalink to this headline">¶</a></h3>
+<p>The implicit <code class="docutils literal notranslate"><span class="pre">const</span></code> capture variables created when evaluating a block
+literal expression have the same ownership semantics as the local variables
+they capture. The capture is performed by reading from the captured variable
+and initializing the capture variable with that value; the capture variable is
+destroyed when the block literal is, i.e. at the end of the enclosing scope.</p>
+<p>The <a class="reference internal" href="#arc-ownership-inference"><span class="std std-ref">inference</span></a> rules apply equally to
+<code class="docutils literal notranslate"><span class="pre">__block</span></code> variables, which is a shift in semantics from non-ARC, where
+<code class="docutils literal notranslate"><span class="pre">__block</span></code> variables did not implicitly retain during capture.</p>
+<p><code class="docutils literal notranslate"><span class="pre">__block</span></code> variables of retainable object owner type are moved off the stack
+by initializing the heap copy with the result of moving from the stack copy.</p>
+<p>With the exception of retains done as part of initializing a <code class="docutils literal notranslate"><span class="pre">__strong</span></code>
+parameter variable or reading a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> variable, whenever these semantics
+call for retaining a value of block-pointer type, it has the effect of a
+<code class="docutils literal notranslate"><span class="pre">Block_copy</span></code>. The optimizer may remove such copies when it sees that the
+result is used only as an argument to a call.</p>
+</div>
+<div class="section" id="exceptions">
+<span id="arc-misc-exceptions"></span><h3><a class="toc-backref" href="#id51">Exceptions</a><a class="headerlink" href="#exceptions" title="Permalink to this headline">¶</a></h3>
+<p>By default in Objective C, ARC is not exception-safe for normal releases:</p>
+<ul class="simple">
+<li>It does not end the lifetime of <code class="docutils literal notranslate"><span class="pre">__strong</span></code> variables when their scopes are
+abnormally terminated by an exception.</li>
+<li>It does not perform releases which would occur at the end of a
+full-expression if that full-expression throws an exception.</li>
+</ul>
+<p>A program may be compiled with the option <code class="docutils literal notranslate"><span class="pre">-fobjc-arc-exceptions</span></code> in order to
+enable these, or with the option <code class="docutils literal notranslate"><span class="pre">-fno-objc-arc-exceptions</span></code> to explicitly
+disable them, with the last such argument âwinningâ.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">The standard Cocoa convention is that exceptions signal programmer error and
+are not intended to be recovered from. Making code exceptions-safe by
+default would impose severe runtime and code size penalties on code that
+typically does not actually care about exceptions safety. Therefore,
+ARC-generated code leaks by default on exceptions, which is just fine if the
+process is going to be immediately terminated anyway. Programs which do care
+about recovering from exceptions should enable the option.</p>
+</div>
+<p>In Objective-C++, <code class="docutils literal notranslate"><span class="pre">-fobjc-arc-exceptions</span></code> is enabled by default.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">C++ already introduces pervasive exceptions-cleanup code of the sort that ARC
+introduces. C++ programmers who have not already disabled exceptions are
+much more likely to actual require exception-safety.</p>
+</div>
+<p>ARC does end the lifetimes of <code class="docutils literal notranslate"><span class="pre">__weak</span></code> objects when an exception terminates
+their scope unless exceptions are disabled in the compiler.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">The consequence of a local <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object not being destroyed is very
+likely to be corruption of the Objective-C runtime, so we want to be safer
+here. Of course, potentially massive leaks are about as likely to take down
+the process as this corruption is if the program does try to recover from
+exceptions.</p>
+</div>
+</div>
+<div class="section" id="interior-pointers">
+<span id="arc-misc-interior"></span><h3><a class="toc-backref" href="#id52">Interior pointers</a><a class="headerlink" href="#interior-pointers" title="Permalink to this headline">¶</a></h3>
+<p>An Objective-C method returning a non-retainable pointer may be annotated with
+the <code class="docutils literal notranslate"><span class="pre">objc_returns_inner_pointer</span></code> attribute to indicate that it returns a
+handle to the internal data of an object, and that this reference will be
+invalidated if the object is destroyed. When such a message is sent to an
+object, the objectâs lifetime will be extended until at least the earliest of:</p>
+<ul class="simple">
+<li>the last use of the returned pointer, or any pointer derived from it, in the
+calling function or</li>
+<li>the autorelease pool is restored to a previous state.</li>
+</ul>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>Rationale: not all memory and resources are managed with reference counts; it
+is common for objects to manage private resources in their own, private way.
+Typically these resources are completely encapsulated within the object, but
+some classes offer their users direct access for efficiency. If ARC is not
+aware of methods that return such âinteriorâ pointers, its optimizations can
+cause the owning object to be reclaimed too soon. This attribute informs ARC
+that it must tread lightly.</p>
+<p class="last">The extension rules are somewhat intentionally vague. The autorelease pool
+limit is there to permit a simple implementation to simply retain and
+autorelease the receiver. The other limit permits some amount of
+optimization. The phrase âderived fromâ is intended to encompass the results
+both of pointer transformations, such as casts and arithmetic, and of loading
+from such derived pointers; furthermore, it applies whether or not such
+derivations are applied directly in the calling code or by other utility code
+(for example, the C library routine <code class="docutils literal notranslate"><span class="pre">strchr</span></code>). However, the implementation
+never need account for uses after a return from the code which calls the
+method returning an interior pointer.</p>
+</div>
+<p>As an exception, no extension is required if the receiver is loaded directly
+from a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object with <a class="reference internal" href="#arc-optimization-precise"><span class="std std-ref">precise lifetime semantics</span></a>.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Implicit autoreleases carry the risk of significantly inflating memory use,
+so itâs important to provide users a way of avoiding these autoreleases.
+Tying this to precise lifetime semantics is ideal, as for local variables
+this requires a very explicit annotation, which allows ARC to trust the user
+with good cheer.</p>
+</div>
+</div>
+<div class="section" id="c-retainable-pointer-types">
+<span id="arc-misc-c-retainable"></span><h3><a class="toc-backref" href="#id53">C retainable pointer types</a><a class="headerlink" href="#c-retainable-pointer-types" title="Permalink to this headline">¶</a></h3>
+<p>A type is a <span class="arc-term">C retainable pointer type</span> if it is a pointer to
+(possibly qualified) <code class="docutils literal notranslate"><span class="pre">void</span></code> or a pointer to a (possibly qualifier) <code class="docutils literal notranslate"><span class="pre">struct</span></code>
+or <code class="docutils literal notranslate"><span class="pre">class</span></code> type.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">ARC does not manage pointers of CoreFoundation type (or any of the related
+families of retainable C pointers which interoperate with Objective-C for
+retain/release operation). In fact, ARC does not even know how to
+distinguish these types from arbitrary C pointer types. The intent of this
+concept is to filter out some obviously non-object types while leaving a hook
+for later tightening if a means of exhaustively marking CF types is made
+available.</p>
+</div>
+<div class="section" id="auditing-of-c-retainable-pointer-interfaces">
+<span id="arc-misc-c-retainable-audit"></span><h4><a class="toc-backref" href="#id54">Auditing of C retainable pointer interfaces</a><a class="headerlink" href="#auditing-of-c-retainable-pointer-interfaces" title="Permalink to this headline">¶</a></h4>
+<p><span class="when-revised">[beginning Apple 4.0, LLVM 3.1]</span></p>
+<p>A C function may be marked with the <code class="docutils literal notranslate"><span class="pre">cf_audited_transfer</span></code> attribute to
+express that, except as otherwise marked with attributes, it obeys the
+parameter (consuming vs. non-consuming) and return (retained vs. non-retained)
+conventions for a C function of its name, namely:</p>
+<ul class="simple">
+<li>A parameter of C retainable pointer type is assumed to not be consumed
+unless it is marked with the <code class="docutils literal notranslate"><span class="pre">cf_consumed</span></code> attribute, and</li>
+<li>A result of C retainable pointer type is assumed to not be returned retained
+unless the function is either marked <code class="docutils literal notranslate"><span class="pre">cf_returns_retained</span></code> or it follows
+the create/copy naming convention and is not marked
+<code class="docutils literal notranslate"><span class="pre">cf_returns_not_retained</span></code>.</li>
+</ul>
+<p>A function obeys the <span class="arc-term">create/copy</span> naming convention if its name
+contains as a substring:</p>
+<ul class="simple">
+<li>either âCreateâ or âCopyâ not followed by a lowercase letter, or</li>
+<li>either âcreateâ or âcopyâ not followed by a lowercase letter and
+not preceded by any letter, whether uppercase or lowercase.</li>
+</ul>
+<p>A second attribute, <code class="docutils literal notranslate"><span class="pre">cf_unknown_transfer</span></code>, signifies that a functionâs
+transfer semantics cannot be accurately captured using any of these
+annotations. A program is ill-formed if it annotates the same function with
+both <code class="docutils literal notranslate"><span class="pre">cf_audited_transfer</span></code> and <code class="docutils literal notranslate"><span class="pre">cf_unknown_transfer</span></code>.</p>
+<p>A pragma is provided to facilitate the mass annotation of interfaces:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="cp">#pragma clang arc_cf_code_audited begin</span>
+<span class="p">...</span>
+<span class="cp">#pragma clang arc_cf_code_audited end</span>
+</pre></div>
+</div>
+<p>All C functions declared within the extent of this pragma are treated as if
+annotated with the <code class="docutils literal notranslate"><span class="pre">cf_audited_transfer</span></code> attribute unless they otherwise have
+the <code class="docutils literal notranslate"><span class="pre">cf_unknown_transfer</span></code> attribute. The pragma is accepted in all language
+modes. A program is ill-formed if it attempts to change files, whether by
+including a file or ending the current file, within the extent of this pragma.</p>
+<p>It is possible to test for all the features in this section with
+<code class="docutils literal notranslate"><span class="pre">__has_feature(arc_cf_code_audited)</span></code>.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">A significant inconvenience in ARC programming is the necessity of
+interacting with APIs based around C retainable pointers. These features are
+designed to make it relatively easy for API authors to quickly review and
+annotate their interfaces, in turn improving the fidelity of tools such as
+the static analyzer and ARC. The single-file restriction on the pragma is
+designed to eliminate the risk of accidentally annotating some other headerâs
+interfaces.</p>
+</div>
+</div>
+</div>
+</div>
+<div class="section" id="runtime-support">
+<span id="arc-runtime"></span><h2><a class="toc-backref" href="#id55">Runtime support</a><a class="headerlink" href="#runtime-support" title="Permalink to this headline">¶</a></h2>
+<p>This section describes the interaction between the ARC runtime and the code
+generated by the ARC compiler. This is not part of the ARC language
+specification; instead, it is effectively a language-specific ABI supplement,
+akin to the âItaniumâ generic ABI for C++.</p>
+<p>Ownership qualification does not alter the storage requirements for objects,
+except that it is undefined behavior if a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object is inadequately
+aligned for an object of type <code class="docutils literal notranslate"><span class="pre">id</span></code>. The other qualifiers may be used on
+explicitly under-aligned memory.</p>
+<p>The runtime tracks <code class="docutils literal notranslate"><span class="pre">__weak</span></code> objects which holds non-null values. It is
+undefined behavior to direct modify a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object which is being tracked
+by the runtime except through an
+<a class="reference internal" href="#arc-runtime-objc-storeweak"><span class="std std-ref">objc_storeWeak</span></a>,
+<a class="reference internal" href="#arc-runtime-objc-destroyweak"><span class="std std-ref">objc_destroyWeak</span></a>, or
+<a class="reference internal" href="#arc-runtime-objc-moveweak"><span class="std std-ref">objc_moveWeak</span></a> call.</p>
+<p>The runtime must provide a number of new entrypoints which the compiler may
+emit, which are described in the remainder of this section.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p>Several of these functions are semantically equivalent to a message send; we
+emit calls to C functions instead because:</p>
+<ul class="simple">
+<li>the machine code to do so is significantly smaller,</li>
+<li>it is much easier to recognize the C functions in the ARC optimizer, and</li>
+<li>a sufficient sophisticated runtime may be able to avoid the message send in
+common cases.</li>
+</ul>
+<p class="last">Several other of these functions are âfusedâ operations which can be
+described entirely in terms of other operations. We use the fused operations
+primarily as a code-size optimization, although in some cases there is also a
+real potential for avoiding redundant operations in the runtime.</p>
+</div>
+<div class="section" id="arc-runtime-objc-autorelease">
+<span id="id-objc-autorelease-id-value"></span><h3><a class="toc-backref" href="#id56"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_autorelease(id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-autorelease" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is null, this call has no effect. Otherwise, it adds the object
+to the innermost autorelease pool exactly as if the object had been sent the
+<code class="docutils literal notranslate"><span class="pre">autorelease</span></code> message.</p>
+<p>Always returns <code class="docutils literal notranslate"><span class="pre">value</span></code>.</p>
+</div>
+<div class="section" id="void-objc-autoreleasepoolpop-void-pool">
+<span id="arc-runtime-objc-autoreleasepoolpop"></span><h3><a class="toc-backref" href="#id57"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_autoreleasePoolPop(void</span> <span class="pre">*pool);</span></code></a><a class="headerlink" href="#void-objc-autoreleasepoolpop-void-pool" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">pool</span></code> is the result of a previous call to
+<a class="reference internal" href="#arc-runtime-objc-autoreleasepoolpush"><span class="std std-ref">objc_autoreleasePoolPush</span></a> on the
+current thread, where neither <code class="docutils literal notranslate"><span class="pre">pool</span></code> nor any enclosing pool have previously
+been popped.</p>
+<p>Releases all the objects added to the given autorelease pool and any
+autorelease pools it encloses, then sets the current autorelease pool to the
+pool directly enclosing <code class="docutils literal notranslate"><span class="pre">pool</span></code>.</p>
+</div>
+<div class="section" id="void-objc-autoreleasepoolpush-void">
+<span id="arc-runtime-objc-autoreleasepoolpush"></span><h3><a class="toc-backref" href="#id58"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">*objc_autoreleasePoolPush(void);</span></code></a><a class="headerlink" href="#void-objc-autoreleasepoolpush-void" title="Permalink to this headline">¶</a></h3>
+<p>Creates a new autorelease pool that is enclosed by the current pool, makes that
+the current pool, and returns an opaque âhandleâ to it.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">While the interface is described as an explicit hierarchy of pools, the rules
+allow the implementation to just keep a stack of objects, using the stack
+depth as the opaque pool handle.</p>
+</div>
+</div>
+<div class="section" id="arc-runtime-objc-autoreleasereturnvalue">
+<span id="id-objc-autoreleasereturnvalue-id-value"></span><h3><a class="toc-backref" href="#id59"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_autoreleaseReturnValue(id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-autoreleasereturnvalue" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is null, this call has no effect. Otherwise, it makes a best
+effort to hand off ownership of a retain count on the object to a call to
+<a class="reference internal" href="#arc-runtime-objc-retainautoreleasedreturnvalue"><span class="std std-ref">objc_retainAutoreleasedReturnValue</span></a> for the same object in an
+enclosing call frame. If this is not possible, the object is autoreleased as
+above.</p>
+<p>Always returns <code class="docutils literal notranslate"><span class="pre">value</span></code>.</p>
+</div>
+<div class="section" id="void-objc-copyweak-id-dest-id-src">
+<span id="arc-runtime-objc-copyweak"></span><h3><a class="toc-backref" href="#id60"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_copyWeak(id</span> <span class="pre">*dest,</span> <span class="pre">id</span> <span class="pre">*src);</span></code></a><a class="headerlink" href="#void-objc-copyweak-id-dest-id-src" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">src</span></code> is a valid pointer which either contains a null pointer
+or has been registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object. <code class="docutils literal notranslate"><span class="pre">dest</span></code> is a valid pointer
+which has not been registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object.</p>
+<p><code class="docutils literal notranslate"><span class="pre">dest</span></code> is initialized to be equivalent to <code class="docutils literal notranslate"><span class="pre">src</span></code>, potentially registering it
+with the runtime. Equivalent to the following code:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">objc_copyWeak</span><span class="p">(</span><span class="kt">id</span> <span class="o">*</span><span class="n">dest</span><span class="p">,</span> <span class="kt">id</span> <span class="o">*</span><span class="n">src</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">objc_release</span><span class="p">(</span><span class="n">objc_initWeak</span><span class="p">(</span><span class="n">dest</span><span class="p">,</span> <span class="n">objc_loadWeakRetained</span><span class="p">(</span><span class="n">src</span><span class="p">)));</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Must be atomic with respect to calls to <code class="docutils literal notranslate"><span class="pre">objc_storeWeak</span></code> on <code class="docutils literal notranslate"><span class="pre">src</span></code>.</p>
+</div>
+<div class="section" id="void-objc-destroyweak-id-object">
+<span id="arc-runtime-objc-destroyweak"></span><h3><a class="toc-backref" href="#id61"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_destroyWeak(id</span> <span class="pre">*object);</span></code></a><a class="headerlink" href="#void-objc-destroyweak-id-object" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">object</span></code> is a valid pointer which either contains a null
+pointer or has been registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object.</p>
+<p><code class="docutils literal notranslate"><span class="pre">object</span></code> is unregistered as a weak object, if it ever was. The current value
+of <code class="docutils literal notranslate"><span class="pre">object</span></code> is left unspecified; otherwise, equivalent to the following code:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">objc_destroyWeak</span><span class="p">(</span><span class="kt">id</span> <span class="o">*</span><span class="n">object</span><span class="p">)</span> <span class="p">{</span>
+ <span class="n">objc_storeWeak</span><span class="p">(</span><span class="n">object</span><span class="p">,</span> <span class="nb">nil</span><span class="p">);</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Does not need to be atomic with respect to calls to <code class="docutils literal notranslate"><span class="pre">objc_storeWeak</span></code> on
+<code class="docutils literal notranslate"><span class="pre">object</span></code>.</p>
+</div>
+<div class="section" id="arc-runtime-objc-initweak">
+<span id="id-objc-initweak-id-object-id-value"></span><h3><a class="toc-backref" href="#id62"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_initWeak(id</span> <span class="pre">*object,</span> <span class="pre">id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-initweak" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">object</span></code> is a valid pointer which has not been registered as
+a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object. <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is a null pointer or the object to which it points has begun
+deallocation, <code class="docutils literal notranslate"><span class="pre">object</span></code> is zero-initialized. Otherwise, <code class="docutils literal notranslate"><span class="pre">object</span></code> is
+registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object pointing to <code class="docutils literal notranslate"><span class="pre">value</span></code>. Equivalent to the
+following code:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">id</span> <span class="nf">objc_initWeak</span><span class="p">(</span><span class="kt">id</span> <span class="o">*</span><span class="n">object</span><span class="p">,</span> <span class="kt">id</span> <span class="n">value</span><span class="p">)</span> <span class="p">{</span>
+ <span class="o">*</span><span class="n">object</span> <span class="o">=</span> <span class="nb">nil</span><span class="p">;</span>
+ <span class="k">return</span> <span class="n">objc_storeWeak</span><span class="p">(</span><span class="n">object</span><span class="p">,</span> <span class="n">value</span><span class="p">);</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Returns the value of <code class="docutils literal notranslate"><span class="pre">object</span></code> after the call.</p>
+<p>Does not need to be atomic with respect to calls to <code class="docutils literal notranslate"><span class="pre">objc_storeWeak</span></code> on
+<code class="docutils literal notranslate"><span class="pre">object</span></code>.</p>
+</div>
+<div class="section" id="arc-runtime-objc-loadweak">
+<span id="id-objc-loadweak-id-object"></span><h3><a class="toc-backref" href="#id63"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_loadWeak(id</span> <span class="pre">*object);</span></code></a><a class="headerlink" href="#arc-runtime-objc-loadweak" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">object</span></code> is a valid pointer which either contains a null
+pointer or has been registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">object</span></code> is registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object, and the last value stored
+into <code class="docutils literal notranslate"><span class="pre">object</span></code> has not yet been deallocated or begun deallocation, retains and
+autoreleases that value and returns it. Otherwise returns null. Equivalent to
+the following code:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">id</span> <span class="nf">objc_loadWeak</span><span class="p">(</span><span class="kt">id</span> <span class="o">*</span><span class="n">object</span><span class="p">)</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="n">objc_autorelease</span><span class="p">(</span><span class="n">objc_loadWeakRetained</span><span class="p">(</span><span class="n">object</span><span class="p">));</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Must be atomic with respect to calls to <code class="docutils literal notranslate"><span class="pre">objc_storeWeak</span></code> on <code class="docutils literal notranslate"><span class="pre">object</span></code>.</p>
+<div class="admonition-rationale admonition">
+<p class="first admonition-title">Rationale</p>
+<p class="last">Loading weak references would be inherently prone to race conditions without
+the retain.</p>
+</div>
+</div>
+<div class="section" id="arc-runtime-objc-loadweakretained">
+<span id="id-objc-loadweakretained-id-object"></span><h3><a class="toc-backref" href="#id64"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_loadWeakRetained(id</span> <span class="pre">*object);</span></code></a><a class="headerlink" href="#arc-runtime-objc-loadweakretained" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">object</span></code> is a valid pointer which either contains a null
+pointer or has been registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">object</span></code> is registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object, and the last value stored
+into <code class="docutils literal notranslate"><span class="pre">object</span></code> has not yet been deallocated or begun deallocation, retains
+that value and returns it. Otherwise returns null.</p>
+<p>Must be atomic with respect to calls to <code class="docutils literal notranslate"><span class="pre">objc_storeWeak</span></code> on <code class="docutils literal notranslate"><span class="pre">object</span></code>.</p>
+</div>
+<div class="section" id="void-objc-moveweak-id-dest-id-src">
+<span id="arc-runtime-objc-moveweak"></span><h3><a class="toc-backref" href="#id65"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_moveWeak(id</span> <span class="pre">*dest,</span> <span class="pre">id</span> <span class="pre">*src);</span></code></a><a class="headerlink" href="#void-objc-moveweak-id-dest-id-src" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">src</span></code> is a valid pointer which either contains a null pointer
+or has been registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object. <code class="docutils literal notranslate"><span class="pre">dest</span></code> is a valid pointer
+which has not been registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object.</p>
+<p><code class="docutils literal notranslate"><span class="pre">dest</span></code> is initialized to be equivalent to <code class="docutils literal notranslate"><span class="pre">src</span></code>, potentially registering it
+with the runtime. <code class="docutils literal notranslate"><span class="pre">src</span></code> may then be left in its original state, in which
+case this call is equivalent to <a class="reference internal" href="#arc-runtime-objc-copyweak"><span class="std std-ref">objc_copyWeak</span></a>, or it may be left as null.</p>
+<p>Must be atomic with respect to calls to <code class="docutils literal notranslate"><span class="pre">objc_storeWeak</span></code> on <code class="docutils literal notranslate"><span class="pre">src</span></code>.</p>
+</div>
+<div class="section" id="void-objc-release-id-value">
+<span id="arc-runtime-objc-release"></span><h3><a class="toc-backref" href="#id66"><code class="docutils literal notranslate"><span class="pre">void</span> <span class="pre">objc_release(id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#void-objc-release-id-value" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is null, this call has no effect. Otherwise, it performs a
+release operation exactly as if the object had been sent the <code class="docutils literal notranslate"><span class="pre">release</span></code>
+message.</p>
+</div>
+<div class="section" id="arc-runtime-objc-retain">
+<span id="id-objc-retain-id-value"></span><h3><a class="toc-backref" href="#id67"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retain(id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-retain" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is null, this call has no effect. Otherwise, it performs a retain
+operation exactly as if the object had been sent the <code class="docutils literal notranslate"><span class="pre">retain</span></code> message.</p>
+<p>Always returns <code class="docutils literal notranslate"><span class="pre">value</span></code>.</p>
+</div>
+<div class="section" id="arc-runtime-objc-retainautorelease">
+<span id="id-objc-retainautorelease-id-value"></span><h3><a class="toc-backref" href="#id68"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retainAutorelease(id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-retainautorelease" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is null, this call has no effect. Otherwise, it performs a retain
+operation followed by an autorelease operation. Equivalent to the following
+code:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">id</span> <span class="nf">objc_retainAutorelease</span><span class="p">(</span><span class="kt">id</span> <span class="n">value</span><span class="p">)</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="n">objc_autorelease</span><span class="p">(</span><span class="n">objc_retain</span><span class="p">(</span><span class="n">value</span><span class="p">));</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Always returns <code class="docutils literal notranslate"><span class="pre">value</span></code>.</p>
+</div>
+<div class="section" id="arc-runtime-objc-retainautoreleasereturnvalue">
+<span id="id-objc-retainautoreleasereturnvalue-id-value"></span><h3><a class="toc-backref" href="#id69"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retainAutoreleaseReturnValue(id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-retainautoreleasereturnvalue" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is null, this call has no effect. Otherwise, it performs a retain
+operation followed by the operation described in
+<a class="reference internal" href="#arc-runtime-objc-autoreleasereturnvalue"><span class="std std-ref">objc_autoreleaseReturnValue</span></a>.
+Equivalent to the following code:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">id</span> <span class="nf">objc_retainAutoreleaseReturnValue</span><span class="p">(</span><span class="kt">id</span> <span class="n">value</span><span class="p">)</span> <span class="p">{</span>
+ <span class="k">return</span> <span class="n">objc_autoreleaseReturnValue</span><span class="p">(</span><span class="n">objc_retain</span><span class="p">(</span><span class="n">value</span><span class="p">));</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<p>Always returns <code class="docutils literal notranslate"><span class="pre">value</span></code>.</p>
+</div>
+<div class="section" id="arc-runtime-objc-retainautoreleasedreturnvalue">
+<span id="id-objc-retainautoreleasedreturnvalue-id-value"></span><h3><a class="toc-backref" href="#id70"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retainAutoreleasedReturnValue(id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-retainautoreleasedreturnvalue" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is null, this call has no effect. Otherwise, it attempts to
+accept a hand off of a retain count from a call to
+<a class="reference internal" href="#arc-runtime-objc-autoreleasereturnvalue"><span class="std std-ref">objc_autoreleaseReturnValue</span></a> on
+<code class="docutils literal notranslate"><span class="pre">value</span></code> in a recently-called function or something it calls. If that fails,
+it performs a retain operation exactly like <a class="reference internal" href="#arc-runtime-objc-retain"><span class="std std-ref">objc_retain</span></a>.</p>
+<p>Always returns <code class="docutils literal notranslate"><span class="pre">value</span></code>.</p>
+</div>
+<div class="section" id="arc-runtime-objc-retainblock">
+<span id="id-objc-retainblock-id-value"></span><h3><a class="toc-backref" href="#id71"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_retainBlock(id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-retainblock" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid block object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is null, this call has no effect. Otherwise, if the block pointed
+to by <code class="docutils literal notranslate"><span class="pre">value</span></code> is still on the stack, it is copied to the heap and the address
+of the copy is returned. Otherwise a retain operation is performed on the
+block exactly as if it had been sent the <code class="docutils literal notranslate"><span class="pre">retain</span></code> message.</p>
+</div>
+<div class="section" id="arc-runtime-objc-storestrong">
+<span id="id-objc-storestrong-id-object-id-value"></span><h3><a class="toc-backref" href="#id72"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_storeStrong(id</span> <span class="pre">*object,</span> <span class="pre">id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-storestrong" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">object</span></code> is a valid pointer to a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object which is
+adequately aligned for a pointer. <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a pointer to a valid
+object.</p>
+<p>Performs the complete sequence for assigning to a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object of
+non-block type <a class="footnote-reference" href="#id3" id="id2">[*]</a>. Equivalent to the following code:</p>
+<div class="highlight-objc notranslate"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">objc_storeStrong</span><span class="p">(</span><span class="kt">id</span> <span class="o">*</span><span class="n">object</span><span class="p">,</span> <span class="kt">id</span> <span class="n">value</span><span class="p">)</span> <span class="p">{</span>
+ <span class="kt">id</span> <span class="n">oldValue</span> <span class="o">=</span> <span class="o">*</span><span class="n">object</span><span class="p">;</span>
+ <span class="n">value</span> <span class="o">=</span> <span class="p">[</span><span class="n">value</span> <span class="k">retain</span><span class="p">];</span>
+ <span class="o">*</span><span class="n">object</span> <span class="o">=</span> <span class="n">value</span><span class="p">;</span>
+ <span class="p">[</span><span class="n">oldValue</span> <span class="k">release</span><span class="p">];</span>
+<span class="p">}</span>
+</pre></div>
+</div>
+<table class="docutils footnote" frame="void" id="id3" rules="none">
+<colgroup><col class="label" /><col /></colgroup>
+<tbody valign="top">
+<tr><td class="label"><a class="fn-backref" href="#id2">[*]</a></td><td>This does not imply that a <code class="docutils literal notranslate"><span class="pre">__strong</span></code> object of block type is an
+invalid argument to this function. Rather it implies that an <code class="docutils literal notranslate"><span class="pre">objc_retain</span></code>
+and not an <code class="docutils literal notranslate"><span class="pre">objc_retainBlock</span></code> operation will be emitted if the argument is
+a block.</td></tr>
+</tbody>
+</table>
+</div>
+<div class="section" id="arc-runtime-objc-storeweak">
+<span id="id-objc-storeweak-id-object-id-value"></span><h3><a class="toc-backref" href="#id73"><code class="docutils literal notranslate"><span class="pre">id</span> <span class="pre">objc_storeWeak(id</span> <span class="pre">*object,</span> <span class="pre">id</span> <span class="pre">value);</span></code></a><a class="headerlink" href="#arc-runtime-objc-storeweak" title="Permalink to this headline">¶</a></h3>
+<p><em>Precondition:</em> <code class="docutils literal notranslate"><span class="pre">object</span></code> is a valid pointer which either contains a null
+pointer or has been registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object. <code class="docutils literal notranslate"><span class="pre">value</span></code> is null or a
+pointer to a valid object.</p>
+<p>If <code class="docutils literal notranslate"><span class="pre">value</span></code> is a null pointer or the object to which it points has begun
+deallocation, <code class="docutils literal notranslate"><span class="pre">object</span></code> is assigned null and unregistered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code>
+object. Otherwise, <code class="docutils literal notranslate"><span class="pre">object</span></code> is registered as a <code class="docutils literal notranslate"><span class="pre">__weak</span></code> object or has its
+registration updated to point to <code class="docutils literal notranslate"><span class="pre">value</span></code>.</p>
+<p>Returns the value of <code class="docutils literal notranslate"><span class="pre">object</span></code> after the call.</p>
+</div>
+</div>
+</div>
+
+
+ </div>
+ <div class="bottomnav" role="navigation" aria-label="bottom navigation">
+
+ <p>
+ « <a href="Block-ABI-Apple.html">Block Implementation Specification</a>
+ ::
+ <a class="uplink" href="index.html">Contents</a>
+ ::
+ <a href="ClangCommandLineReference.html">Clang command line argument reference</a> »
+ </p>
+
+ </div>
+
+ <div class="footer" role="contentinfo">
+ © Copyright 2007-2018, The Clang Team.
+ Created using <a href="http://sphinx-doc.org/">Sphinx</a> 1.7.6.
+ </div>
+ </body>
+</html>
\ No newline at end of file
More information about the llvm-commits
mailing list