blob: b963208f8c2d83eb676b914fcaf3255f1da00dc1 [file] [log] [blame]
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
Misha Brukman9d0919f2003-11-08 01:05:38 +00003<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
Reid Spencer3921c742004-08-26 20:44:00 +00006 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
Misha Brukman9d0919f2003-11-08 01:05:38 +000010 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
Chris Lattnerd7923912004-05-23 21:06:01 +000012
Misha Brukman9d0919f2003-11-08 01:05:38 +000013<body>
Chris Lattnerd7923912004-05-23 21:06:01 +000014
Chris Lattner261efe92003-11-25 01:02:51 +000015<div class="doc_title"> LLVM Language Reference Manual </div>
Chris Lattner00950542001-06-06 20:29:01 +000016<ol>
Misha Brukman9d0919f2003-11-08 01:05:38 +000017 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
Chris Lattnerfa730212004-12-09 16:11:40 +000020 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
Chris Lattnere5d947b2004-12-09 16:36:40 +000023 <li><a href="#linkage">Linkage Types</a></li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +000024 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattnerfa730212004-12-09 16:11:40 +000025 <li><a href="#globalvars">Global Variables</a></li>
Chris Lattner4e9aba72006-01-23 23:23:47 +000026 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
Chris Lattnerfa730212004-12-09 16:11:40 +000028 </ol>
29 </li>
Chris Lattner00950542001-06-06 20:29:01 +000030 <li><a href="#typesystem">Type System</a>
31 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +000032 <li><a href="#t_primitive">Primitive Types</a>
33 <ol>
Misha Brukman9d0919f2003-11-08 01:05:38 +000034 <li><a href="#t_classifications">Type Classifications</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000035 </ol>
36 </li>
Chris Lattner00950542001-06-06 20:29:01 +000037 <li><a href="#t_derived">Derived Types</a>
38 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +000039 <li><a href="#t_array">Array Type</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +000040 <li><a href="#t_function">Function Type</a></li>
41 <li><a href="#t_pointer">Pointer Type</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000042 <li><a href="#t_struct">Structure Type</a></li>
Chris Lattnera58561b2004-08-12 19:12:28 +000043 <li><a href="#t_packed">Packed Type</a></li>
Chris Lattner69c11bb2005-04-25 17:34:15 +000044 <li><a href="#t_opaque">Opaque Type</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000045 </ol>
46 </li>
47 </ol>
48 </li>
Chris Lattnerfa730212004-12-09 16:11:40 +000049 <li><a href="#constants">Constants</a>
Chris Lattnerc3f59762004-12-09 17:30:23 +000050 <ol>
51 <li><a href="#simpleconstants">Simple Constants</a>
52 <li><a href="#aggregateconstants">Aggregate Constants</a>
53 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
54 <li><a href="#undefvalues">Undefined Values</a>
55 <li><a href="#constantexprs">Constant Expressions</a>
56 </ol>
Chris Lattner261efe92003-11-25 01:02:51 +000057 </li>
Chris Lattnere87d6532006-01-25 23:47:57 +000058 <li><a href="#othervalues">Other Values</a>
59 <ol>
60 <li><a href="#inlineasm">Inline Assembler Expressions</a>
61 </ol>
62 </li>
Chris Lattner00950542001-06-06 20:29:01 +000063 <li><a href="#instref">Instruction Reference</a>
64 <ol>
65 <li><a href="#terminators">Terminator Instructions</a>
66 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +000067 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
68 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +000069 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
70 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000071 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
Chris Lattner35eca582004-10-16 18:04:13 +000072 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000073 </ol>
74 </li>
Chris Lattner00950542001-06-06 20:29:01 +000075 <li><a href="#binaryops">Binary Operations</a>
76 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +000077 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
78 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
79 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
80 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
81 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +000082 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000083 </ol>
84 </li>
Chris Lattner00950542001-06-06 20:29:01 +000085 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
86 <ol>
Misha Brukman9d0919f2003-11-08 01:05:38 +000087 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000088 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +000089 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
90 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
91 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000092 </ol>
93 </li>
Chris Lattner00950542001-06-06 20:29:01 +000094 <li><a href="#memoryops">Memory Access Operations</a>
95 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +000096 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
97 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
98 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
99 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
100 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
101 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
102 </ol>
103 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000104 <li><a href="#otherops">Other Operations</a>
105 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000106 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000107 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
Chris Lattnercc37aae2004-03-12 05:50:16 +0000108 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
Robert Bocchino3a558662006-01-05 17:37:02 +0000109 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
Robert Bocchino05ccd702006-01-15 20:48:27 +0000110 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000111 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
Chris Lattnerfb6977d2006-01-13 23:26:01 +0000112 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Chris Lattner00950542001-06-06 20:29:01 +0000113 </ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000114 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000115 </ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000116 </li>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +0000117 <li><a href="#intrinsics">Intrinsic Functions</a>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +0000118 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000119 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
120 <ol>
121 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
122 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
123 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
124 </ol>
125 </li>
Chris Lattnerd7923912004-05-23 21:06:01 +0000126 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
127 <ol>
128 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
129 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
130 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
131 </ol>
132 </li>
Chris Lattner10610642004-02-14 04:08:35 +0000133 <li><a href="#int_codegen">Code Generator Intrinsics</a>
134 <ol>
135 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
136 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
Chris Lattner57e1f392006-01-13 02:03:13 +0000137 <li><a href="#i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
138 <li><a href="#i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +0000139 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +0000140 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
Andrew Lenharth51b8d542005-11-11 16:47:30 +0000141 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
John Criswell7123e272004-04-09 16:43:20 +0000142 </ol>
143 </li>
144 <li><a href="#int_os">Operating System Intrinsics</a>
145 <ol>
Chris Lattner32006282004-06-11 02:28:03 +0000146 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
147 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
John Criswell183402a2004-04-12 15:02:16 +0000148 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
149 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
Chris Lattner10610642004-02-14 04:08:35 +0000150 </ol>
Chris Lattner33aec9e2004-02-12 17:01:32 +0000151 <li><a href="#int_libc">Standard C Library Intrinsics</a>
152 <ol>
153 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
Chris Lattner0eb51b42004-02-12 18:10:10 +0000154 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
Chris Lattner10610642004-02-14 04:08:35 +0000155 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
Chris Lattnerec6cb612006-01-16 22:38:59 +0000156 <li><a href="#i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a></li>
157 <li><a href="#i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
Chris Lattnera4d74142005-07-21 01:29:16 +0000158
Chris Lattner33aec9e2004-02-12 17:01:32 +0000159 </ol>
160 </li>
Nate Begeman7e36c472006-01-13 23:26:38 +0000161 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharthec370fd2005-05-03 18:01:48 +0000162 <ol>
Nate Begeman7e36c472006-01-13 23:26:38 +0000163 <li><a href="#i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
Chris Lattner8a886be2006-01-16 22:34:14 +0000164 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
165 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
166 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
Andrew Lenharthec370fd2005-05-03 18:01:48 +0000167 </ol>
168 </li>
Chris Lattnerd7923912004-05-23 21:06:01 +0000169 <li><a href="#int_debugger">Debugger intrinsics</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000170 </ol>
171 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000172</ol>
Chris Lattnerd7923912004-05-23 21:06:01 +0000173
174<div class="doc_author">
175 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
176 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000177</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000178
Chris Lattner00950542001-06-06 20:29:01 +0000179<!-- *********************************************************************** -->
Chris Lattner261efe92003-11-25 01:02:51 +0000180<div class="doc_section"> <a name="abstract">Abstract </a></div>
181<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000182
Misha Brukman9d0919f2003-11-08 01:05:38 +0000183<div class="doc_text">
Chris Lattner261efe92003-11-25 01:02:51 +0000184<p>This document is a reference manual for the LLVM assembly language.
185LLVM is an SSA based representation that provides type safety,
186low-level operations, flexibility, and the capability of representing
187'all' high-level languages cleanly. It is the common code
188representation used throughout all phases of the LLVM compilation
189strategy.</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000190</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000191
Chris Lattner00950542001-06-06 20:29:01 +0000192<!-- *********************************************************************** -->
Chris Lattner261efe92003-11-25 01:02:51 +0000193<div class="doc_section"> <a name="introduction">Introduction</a> </div>
194<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000195
Misha Brukman9d0919f2003-11-08 01:05:38 +0000196<div class="doc_text">
Chris Lattnerd7923912004-05-23 21:06:01 +0000197
Chris Lattner261efe92003-11-25 01:02:51 +0000198<p>The LLVM code representation is designed to be used in three
199different forms: as an in-memory compiler IR, as an on-disk bytecode
200representation (suitable for fast loading by a Just-In-Time compiler),
201and as a human readable assembly language representation. This allows
202LLVM to provide a powerful intermediate representation for efficient
203compiler transformations and analysis, while providing a natural means
204to debug and visualize the transformations. The three different forms
205of LLVM are all equivalent. This document describes the human readable
206representation and notation.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000207
John Criswellc1f786c2005-05-13 22:25:59 +0000208<p>The LLVM representation aims to be light-weight and low-level
Chris Lattner261efe92003-11-25 01:02:51 +0000209while being expressive, typed, and extensible at the same time. It
210aims to be a "universal IR" of sorts, by being at a low enough level
211that high-level ideas may be cleanly mapped to it (similar to how
212microprocessors are "universal IR's", allowing many source languages to
213be mapped to them). By providing type information, LLVM can be used as
214the target of optimizations: for example, through pointer analysis, it
215can be proven that a C automatic variable is never accessed outside of
216the current function... allowing it to be promoted to a simple SSA
217value instead of a memory location.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000218
Misha Brukman9d0919f2003-11-08 01:05:38 +0000219</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000220
Chris Lattner00950542001-06-06 20:29:01 +0000221<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +0000222<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000223
Misha Brukman9d0919f2003-11-08 01:05:38 +0000224<div class="doc_text">
Chris Lattnerd7923912004-05-23 21:06:01 +0000225
Chris Lattner261efe92003-11-25 01:02:51 +0000226<p>It is important to note that this document describes 'well formed'
227LLVM assembly language. There is a difference between what the parser
228accepts and what is considered 'well formed'. For example, the
229following instruction is syntactically okay, but not well formed:</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000230
231<pre>
232 %x = <a href="#i_add">add</a> int 1, %x
233</pre>
234
Chris Lattner261efe92003-11-25 01:02:51 +0000235<p>...because the definition of <tt>%x</tt> does not dominate all of
236its uses. The LLVM infrastructure provides a verification pass that may
237be used to verify that an LLVM module is well formed. This pass is
John Criswellc1f786c2005-05-13 22:25:59 +0000238automatically run by the parser after parsing input assembly and by
Chris Lattner261efe92003-11-25 01:02:51 +0000239the optimizer before it outputs bytecode. The violations pointed out
240by the verifier pass indicate bugs in transformation passes or input to
241the parser.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000242
Chris Lattner261efe92003-11-25 01:02:51 +0000243<!-- Describe the typesetting conventions here. --> </div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000244
Chris Lattner00950542001-06-06 20:29:01 +0000245<!-- *********************************************************************** -->
Chris Lattner261efe92003-11-25 01:02:51 +0000246<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
Chris Lattner00950542001-06-06 20:29:01 +0000247<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000248
Misha Brukman9d0919f2003-11-08 01:05:38 +0000249<div class="doc_text">
Chris Lattnerd7923912004-05-23 21:06:01 +0000250
Chris Lattner261efe92003-11-25 01:02:51 +0000251<p>LLVM uses three different forms of identifiers, for different
252purposes:</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000253
Chris Lattner00950542001-06-06 20:29:01 +0000254<ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000255 <li>Named values are represented as a string of characters with a '%' prefix.
256 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
257 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
258 Identifiers which require other characters in their names can be surrounded
259 with quotes. In this way, anything except a <tt>"</tt> character can be used
260 in a name.</li>
261
262 <li>Unnamed values are represented as an unsigned numeric value with a '%'
263 prefix. For example, %12, %2, %44.</li>
264
Reid Spencercc16dc32004-12-09 18:02:53 +0000265 <li>Constants, which are described in a <a href="#constants">section about
266 constants</a>, below.</li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000267</ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000268
269<p>LLVM requires that values start with a '%' sign for two reasons: Compilers
270don't need to worry about name clashes with reserved words, and the set of
271reserved words may be expanded in the future without penalty. Additionally,
272unnamed identifiers allow a compiler to quickly come up with a temporary
273variable without having to avoid symbol table conflicts.</p>
274
Chris Lattner261efe92003-11-25 01:02:51 +0000275<p>Reserved words in LLVM are very similar to reserved words in other
276languages. There are keywords for different opcodes ('<tt><a
Chris Lattnere5d947b2004-12-09 16:36:40 +0000277href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
278href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
279href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
280and others. These reserved words cannot conflict with variable names, because
281none of them start with a '%' character.</p>
282
283<p>Here is an example of LLVM code to multiply the integer variable
284'<tt>%X</tt>' by 8:</p>
285
Misha Brukman9d0919f2003-11-08 01:05:38 +0000286<p>The easy way:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000287
288<pre>
289 %result = <a href="#i_mul">mul</a> uint %X, 8
290</pre>
291
Misha Brukman9d0919f2003-11-08 01:05:38 +0000292<p>After strength reduction:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000293
294<pre>
295 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
296</pre>
297
Misha Brukman9d0919f2003-11-08 01:05:38 +0000298<p>And the hard way:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000299
300<pre>
301 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
302 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
303 %result = <a href="#i_add">add</a> uint %1, %1
304</pre>
305
Chris Lattner261efe92003-11-25 01:02:51 +0000306<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
307important lexical features of LLVM:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000308
Chris Lattner00950542001-06-06 20:29:01 +0000309<ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000310
311 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
312 line.</li>
313
314 <li>Unnamed temporaries are created when the result of a computation is not
315 assigned to a named value.</li>
316
Misha Brukman9d0919f2003-11-08 01:05:38 +0000317 <li>Unnamed temporaries are numbered sequentially</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000318
Misha Brukman9d0919f2003-11-08 01:05:38 +0000319</ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000320
John Criswelle4c57cc2005-05-12 16:52:32 +0000321<p>...and it also shows a convention that we follow in this document. When
Chris Lattnere5d947b2004-12-09 16:36:40 +0000322demonstrating instructions, we will follow an instruction with a comment that
323defines the type and name of value produced. Comments are shown in italic
324text.</p>
325
Misha Brukman9d0919f2003-11-08 01:05:38 +0000326</div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000327
328<!-- *********************************************************************** -->
329<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
330<!-- *********************************************************************** -->
331
332<!-- ======================================================================= -->
333<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
334</div>
335
336<div class="doc_text">
337
338<p>LLVM programs are composed of "Module"s, each of which is a
339translation unit of the input programs. Each module consists of
340functions, global variables, and symbol table entries. Modules may be
341combined together with the LLVM linker, which merges function (and
342global variable) definitions, resolves forward declarations, and merges
343symbol table entries. Here is an example of the "hello world" module:</p>
344
345<pre><i>; Declare the string constant as a global constant...</i>
346<a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
347 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
348
349<i>; External declaration of the puts function</i>
350<a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
351
352<i>; Definition of main function</i>
353int %main() { <i>; int()* </i>
354 <i>; Convert [13x sbyte]* to sbyte *...</i>
355 %cast210 = <a
356 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
357
358 <i>; Call puts function to write out the string to stdout...</i>
359 <a
360 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
361 <a
362 href="#i_ret">ret</a> int 0<br>}<br></pre>
363
364<p>This example is made up of a <a href="#globalvars">global variable</a>
365named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
366function, and a <a href="#functionstructure">function definition</a>
367for "<tt>main</tt>".</p>
368
Chris Lattnere5d947b2004-12-09 16:36:40 +0000369<p>In general, a module is made up of a list of global values,
370where both functions and global variables are global values. Global values are
371represented by a pointer to a memory location (in this case, a pointer to an
372array of char, and a pointer to a function), and have one of the following <a
373href="#linkage">linkage types</a>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000374
Chris Lattnere5d947b2004-12-09 16:36:40 +0000375</div>
376
377<!-- ======================================================================= -->
378<div class="doc_subsection">
379 <a name="linkage">Linkage Types</a>
380</div>
381
382<div class="doc_text">
383
384<p>
385All Global Variables and Functions have one of the following types of linkage:
386</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000387
388<dl>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000389
Chris Lattnerfa730212004-12-09 16:11:40 +0000390 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000391
392 <dd>Global values with internal linkage are only directly accessible by
393 objects in the current module. In particular, linking code into a module with
394 an internal global value may cause the internal to be renamed as necessary to
395 avoid collisions. Because the symbol is internal to the module, all
396 references can be updated. This corresponds to the notion of the
397 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
Chris Lattnerfa730212004-12-09 16:11:40 +0000398 </dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000399
Chris Lattnerfa730212004-12-09 16:11:40 +0000400 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000401
402 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
403 the twist that linking together two modules defining the same
404 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
405 is typically used to implement inline functions. Unreferenced
406 <tt>linkonce</tt> globals are allowed to be discarded.
Chris Lattnerfa730212004-12-09 16:11:40 +0000407 </dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000408
Chris Lattnerfa730212004-12-09 16:11:40 +0000409 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000410
411 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
412 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
413 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
Chris Lattnerfa730212004-12-09 16:11:40 +0000414 </dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000415
Chris Lattnerfa730212004-12-09 16:11:40 +0000416 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000417
418 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
419 pointer to array type. When two global variables with appending linkage are
420 linked together, the two global arrays are appended together. This is the
421 LLVM, typesafe, equivalent of having the system linker append together
422 "sections" with identical names when .o files are linked.
Chris Lattnerfa730212004-12-09 16:11:40 +0000423 </dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000424
Chris Lattnerfa730212004-12-09 16:11:40 +0000425 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000426
427 <dd>If none of the above identifiers are used, the global is externally
428 visible, meaning that it participates in linkage and can be used to resolve
429 external symbol references.
Chris Lattnerfa730212004-12-09 16:11:40 +0000430 </dd>
431</dl>
432
Chris Lattnerfa730212004-12-09 16:11:40 +0000433<p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
434variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
435variable and was linked with this one, one of the two would be renamed,
436preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
437external (i.e., lacking any linkage declarations), they are accessible
438outside of the current module. It is illegal for a function <i>declaration</i>
439to have any linkage type other than "externally visible".</a></p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000440
Chris Lattnerfa730212004-12-09 16:11:40 +0000441</div>
442
443<!-- ======================================================================= -->
444<div class="doc_subsection">
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000445 <a name="callingconv">Calling Conventions</a>
446</div>
447
448<div class="doc_text">
449
450<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
451and <a href="#i_invoke">invokes</a> can all have an optional calling convention
452specified for the call. The calling convention of any pair of dynamic
453caller/callee must match, or the behavior of the program is undefined. The
454following calling conventions are supported by LLVM, and more may be added in
455the future:</p>
456
457<dl>
458 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
459
460 <dd>This calling convention (the default if no other calling convention is
461 specified) matches the target C calling conventions. This calling convention
John Criswelle4c57cc2005-05-12 16:52:32 +0000462 supports varargs function calls and tolerates some mismatch in the declared
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000463 prototype and implemented declaration of the function (as does normal C).
464 </dd>
465
466 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
467
468 <dd>This calling convention attempts to make calls as fast as possible
469 (e.g. by passing things in registers). This calling convention allows the
470 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattner8cdc5bc2005-05-06 23:08:23 +0000471 without having to conform to an externally specified ABI. Implementations of
472 this convention should allow arbitrary tail call optimization to be supported.
473 This calling convention does not support varargs and requires the prototype of
474 all callees to exactly match the prototype of the function definition.
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000475 </dd>
476
477 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
478
479 <dd>This calling convention attempts to make code in the caller as efficient
480 as possible under the assumption that the call is not commonly executed. As
481 such, these calls often preserve all registers so that the call does not break
482 any live ranges in the caller side. This calling convention does not support
483 varargs and requires the prototype of all callees to exactly match the
484 prototype of the function definition.
485 </dd>
486
Chris Lattnercfe6b372005-05-07 01:46:40 +0000487 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000488
489 <dd>Any calling convention may be specified by number, allowing
490 target-specific calling conventions to be used. Target specific calling
491 conventions start at 64.
492 </dd>
Chris Lattnercfe6b372005-05-07 01:46:40 +0000493</dl>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000494
495<p>More calling conventions can be added/defined on an as-needed basis, to
496support pascal conventions or any other well-known target-independent
497convention.</p>
498
499</div>
500
501<!-- ======================================================================= -->
502<div class="doc_subsection">
Chris Lattnerfa730212004-12-09 16:11:40 +0000503 <a name="globalvars">Global Variables</a>
504</div>
505
506<div class="doc_text">
507
Chris Lattner3689a342005-02-12 19:30:21 +0000508<p>Global variables define regions of memory allocated at compilation time
Chris Lattner88f6c462005-11-12 00:45:07 +0000509instead of run-time. Global variables may optionally be initialized, may have
510an explicit section to be placed in, and may
Chris Lattner2cbdc452005-11-06 08:02:57 +0000511have an optional explicit alignment specified. A
John Criswell0ec250c2005-10-24 16:17:18 +0000512variable may be defined as a global "constant," which indicates that the
Chris Lattner3689a342005-02-12 19:30:21 +0000513contents of the variable will <b>never</b> be modified (enabling better
514optimization, allowing the global data to be placed in the read-only section of
515an executable, etc). Note that variables that need runtime initialization
John Criswell0ec250c2005-10-24 16:17:18 +0000516cannot be marked "constant" as there is a store to the variable.</p>
Chris Lattner3689a342005-02-12 19:30:21 +0000517
518<p>
519LLVM explicitly allows <em>declarations</em> of global variables to be marked
520constant, even if the final definition of the global is not. This capability
521can be used to enable slightly better optimization of the program, but requires
522the language definition to guarantee that optimizations based on the
523'constantness' are valid for the translation units that do not include the
524definition.
525</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000526
527<p>As SSA values, global variables define pointer values that are in
528scope (i.e. they dominate) all basic blocks in the program. Global
529variables always define a pointer to their "content" type because they
530describe a region of memory, and all memory objects in LLVM are
531accessed through pointers.</p>
532
Chris Lattner88f6c462005-11-12 00:45:07 +0000533<p>LLVM allows an explicit section to be specified for globals. If the target
534supports it, it will emit globals to the section specified.</p>
535
Chris Lattner2cbdc452005-11-06 08:02:57 +0000536<p>An explicit alignment may be specified for a global. If not present, or if
537the alignment is set to zero, the alignment of the global is set by the target
538to whatever it feels convenient. If an explicit alignment is specified, the
539global is forced to have at least that much alignment. All alignments must be
540a power of 2.</p>
541
Chris Lattnerfa730212004-12-09 16:11:40 +0000542</div>
543
544
545<!-- ======================================================================= -->
546<div class="doc_subsection">
547 <a name="functionstructure">Functions</a>
548</div>
549
550<div class="doc_text">
551
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000552<p>LLVM function definitions consist of an optional <a href="#linkage">linkage
553type</a>, an optional <a href="#callingconv">calling convention</a>, a return
Chris Lattner88f6c462005-11-12 00:45:07 +0000554type, a function name, a (possibly empty) argument list, an optional section,
555an optional alignment, an opening curly brace,
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000556a list of basic blocks, and a closing curly brace. LLVM function declarations
557are defined with the "<tt>declare</tt>" keyword, an optional <a
Chris Lattner2cbdc452005-11-06 08:02:57 +0000558href="#callingconv">calling convention</a>, a return type, a function name,
559a possibly empty list of arguments, and an optional alignment.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000560
561<p>A function definition contains a list of basic blocks, forming the CFG for
562the function. Each basic block may optionally start with a label (giving the
563basic block a symbol table entry), contains a list of instructions, and ends
564with a <a href="#terminators">terminator</a> instruction (such as a branch or
565function return).</p>
566
John Criswelle4c57cc2005-05-12 16:52:32 +0000567<p>The first basic block in a program is special in two ways: it is immediately
Chris Lattnerfa730212004-12-09 16:11:40 +0000568executed on entrance to the function, and it is not allowed to have predecessor
569basic blocks (i.e. there can not be any branches to the entry block of a
570function). Because the block can have no predecessors, it also cannot have any
571<a href="#i_phi">PHI nodes</a>.</p>
572
573<p>LLVM functions are identified by their name and type signature. Hence, two
574functions with the same name but different parameter lists or return values are
Chris Lattnerd4f6b172005-03-07 22:13:59 +0000575considered different functions, and LLVM will resolve references to each
Chris Lattnerfa730212004-12-09 16:11:40 +0000576appropriately.</p>
577
Chris Lattner88f6c462005-11-12 00:45:07 +0000578<p>LLVM allows an explicit section to be specified for functions. If the target
579supports it, it will emit functions to the section specified.</p>
580
Chris Lattner2cbdc452005-11-06 08:02:57 +0000581<p>An explicit alignment may be specified for a function. If not present, or if
582the alignment is set to zero, the alignment of the function is set by the target
583to whatever it feels convenient. If an explicit alignment is specified, the
584function is forced to have at least that much alignment. All alignments must be
585a power of 2.</p>
586
Chris Lattnerfa730212004-12-09 16:11:40 +0000587</div>
588
Chris Lattner4e9aba72006-01-23 23:23:47 +0000589<!-- ======================================================================= -->
590<div class="doc_subsection">
591 <a name="moduleasm">Module-Level Inline Assembly</a></li>
592</div>
593
594<div class="doc_text">
595<p>
596Modules may contain "module-level inline asm" blocks, which corresponds to the
597GCC "file scope inline asm" blocks. These blocks are internally concatenated by
598LLVM and treated as a single unit, but may be separated in the .ll file if
599desired. The syntax is very simple:
600</p>
601
602<div class="doc_code"><pre>
Chris Lattner52599e12006-01-24 00:37:20 +0000603 module asm "inline asm code goes here"
604 module asm "more can go here"
Chris Lattner4e9aba72006-01-23 23:23:47 +0000605</pre></div>
606
607<p>The strings can contain any character by escaping non-printable characters.
608 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
609 for the number.
610</p>
611
612<p>
613 The inline asm code is simply printed to the machine code .s file when
614 assembly code is generated.
615</p>
616</div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000617
618
Chris Lattner00950542001-06-06 20:29:01 +0000619<!-- *********************************************************************** -->
Chris Lattner261efe92003-11-25 01:02:51 +0000620<div class="doc_section"> <a name="typesystem">Type System</a> </div>
621<!-- *********************************************************************** -->
Chris Lattnerfa730212004-12-09 16:11:40 +0000622
Misha Brukman9d0919f2003-11-08 01:05:38 +0000623<div class="doc_text">
Chris Lattnerfa730212004-12-09 16:11:40 +0000624
Misha Brukman9d0919f2003-11-08 01:05:38 +0000625<p>The LLVM type system is one of the most important features of the
Chris Lattner261efe92003-11-25 01:02:51 +0000626intermediate representation. Being typed enables a number of
627optimizations to be performed on the IR directly, without having to do
628extra analyses on the side before the transformation. A strong type
629system makes it easier to read the generated code and enables novel
630analyses and transformations that are not feasible to perform on normal
631three address code representations.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000632
633</div>
634
Chris Lattner00950542001-06-06 20:29:01 +0000635<!-- ======================================================================= -->
Chris Lattner261efe92003-11-25 01:02:51 +0000636<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000637<div class="doc_text">
John Criswell4457dc92004-04-09 16:48:45 +0000638<p>The primitive types are the fundamental building blocks of the LLVM
Chris Lattnerd4f6b172005-03-07 22:13:59 +0000639system. The current set of primitive types is as follows:</p>
Misha Brukmandaa4cb02004-03-01 17:47:27 +0000640
Reid Spencerd3f876c2004-11-01 08:19:36 +0000641<table class="layout">
642 <tr class="layout">
643 <td class="left">
644 <table>
Chris Lattner261efe92003-11-25 01:02:51 +0000645 <tbody>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000646 <tr><th>Type</th><th>Description</th></tr>
647 <tr><td><tt>void</tt></td><td>No value</td></tr>
Misha Brukmancfa87bc2005-04-22 18:02:52 +0000648 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
649 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
650 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
651 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
652 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000653 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
Chris Lattner261efe92003-11-25 01:02:51 +0000654 </tbody>
655 </table>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000656 </td>
657 <td class="right">
658 <table>
Chris Lattner261efe92003-11-25 01:02:51 +0000659 <tbody>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000660 <tr><th>Type</th><th>Description</th></tr>
661 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
Misha Brukmancfa87bc2005-04-22 18:02:52 +0000662 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
663 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
664 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
665 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
666 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
Chris Lattner261efe92003-11-25 01:02:51 +0000667 </tbody>
668 </table>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000669 </td>
670 </tr>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000671</table>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000672</div>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000673
Chris Lattner00950542001-06-06 20:29:01 +0000674<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +0000675<div class="doc_subsubsection"> <a name="t_classifications">Type
676Classifications</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000677<div class="doc_text">
Chris Lattner261efe92003-11-25 01:02:51 +0000678<p>These different primitive types fall into a few useful
679classifications:</p>
Misha Brukmandaa4cb02004-03-01 17:47:27 +0000680
681<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner261efe92003-11-25 01:02:51 +0000682 <tbody>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000683 <tr><th>Classification</th><th>Types</th></tr>
Chris Lattner261efe92003-11-25 01:02:51 +0000684 <tr>
685 <td><a name="t_signed">signed</a></td>
686 <td><tt>sbyte, short, int, long, float, double</tt></td>
687 </tr>
688 <tr>
689 <td><a name="t_unsigned">unsigned</a></td>
690 <td><tt>ubyte, ushort, uint, ulong</tt></td>
691 </tr>
692 <tr>
693 <td><a name="t_integer">integer</a></td>
694 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
695 </tr>
696 <tr>
697 <td><a name="t_integral">integral</a></td>
Misha Brukmanc24b7582004-08-12 20:16:08 +0000698 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
699 </td>
Chris Lattner261efe92003-11-25 01:02:51 +0000700 </tr>
701 <tr>
702 <td><a name="t_floating">floating point</a></td>
703 <td><tt>float, double</tt></td>
704 </tr>
705 <tr>
706 <td><a name="t_firstclass">first class</a></td>
Misha Brukmanc24b7582004-08-12 20:16:08 +0000707 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
708 float, double, <a href="#t_pointer">pointer</a>,
709 <a href="#t_packed">packed</a></tt></td>
Chris Lattner261efe92003-11-25 01:02:51 +0000710 </tr>
711 </tbody>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000712</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +0000713
Chris Lattner261efe92003-11-25 01:02:51 +0000714<p>The <a href="#t_firstclass">first class</a> types are perhaps the
715most important. Values of these types are the only ones which can be
716produced by instructions, passed as arguments, or used as operands to
717instructions. This means that all structures and arrays must be
718manipulated either by pointer or by component.</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000719</div>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000720
Chris Lattner00950542001-06-06 20:29:01 +0000721<!-- ======================================================================= -->
Chris Lattner261efe92003-11-25 01:02:51 +0000722<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000723
Misha Brukman9d0919f2003-11-08 01:05:38 +0000724<div class="doc_text">
Chris Lattnerc3f59762004-12-09 17:30:23 +0000725
Chris Lattner261efe92003-11-25 01:02:51 +0000726<p>The real power in LLVM comes from the derived types in the system.
727This is what allows a programmer to represent arrays, functions,
728pointers, and other useful types. Note that these derived types may be
729recursive: For example, it is possible to have a two dimensional array.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000730
Misha Brukman9d0919f2003-11-08 01:05:38 +0000731</div>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000732
Chris Lattner00950542001-06-06 20:29:01 +0000733<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +0000734<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000735
Misha Brukman9d0919f2003-11-08 01:05:38 +0000736<div class="doc_text">
Chris Lattnerc3f59762004-12-09 17:30:23 +0000737
Chris Lattner00950542001-06-06 20:29:01 +0000738<h5>Overview:</h5>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000739
Misha Brukman9d0919f2003-11-08 01:05:38 +0000740<p>The array type is a very simple derived type that arranges elements
Chris Lattner261efe92003-11-25 01:02:51 +0000741sequentially in memory. The array type requires a size (number of
742elements) and an underlying data type.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000743
Chris Lattner7faa8832002-04-14 06:13:44 +0000744<h5>Syntax:</h5>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000745
746<pre>
747 [&lt;# elements&gt; x &lt;elementtype&gt;]
748</pre>
749
John Criswelle4c57cc2005-05-12 16:52:32 +0000750<p>The number of elements is a constant integer value; elementtype may
Chris Lattner261efe92003-11-25 01:02:51 +0000751be any type with a size.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000752
Chris Lattner7faa8832002-04-14 06:13:44 +0000753<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000754<table class="layout">
755 <tr class="layout">
756 <td class="left">
757 <tt>[40 x int ]</tt><br/>
758 <tt>[41 x int ]</tt><br/>
759 <tt>[40 x uint]</tt><br/>
760 </td>
761 <td class="left">
762 Array of 40 integer values.<br/>
763 Array of 41 integer values.<br/>
764 Array of 40 unsigned integer values.<br/>
765 </td>
766 </tr>
Chris Lattner00950542001-06-06 20:29:01 +0000767</table>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000768<p>Here are some examples of multidimensional arrays:</p>
769<table class="layout">
770 <tr class="layout">
771 <td class="left">
772 <tt>[3 x [4 x int]]</tt><br/>
773 <tt>[12 x [10 x float]]</tt><br/>
774 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
775 </td>
776 <td class="left">
John Criswellc1f786c2005-05-13 22:25:59 +0000777 3x4 array of integer values.<br/>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000778 12x10 array of single precision floating point values.<br/>
779 2x3x4 array of unsigned integer values.<br/>
780 </td>
781 </tr>
782</table>
Chris Lattnere67a9512005-06-24 17:22:57 +0000783
John Criswell0ec250c2005-10-24 16:17:18 +0000784<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
785length array. Normally, accesses past the end of an array are undefined in
Chris Lattnere67a9512005-06-24 17:22:57 +0000786LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
787As a special case, however, zero length arrays are recognized to be variable
788length. This allows implementation of 'pascal style arrays' with the LLVM
789type "{ int, [0 x float]}", for example.</p>
790
Misha Brukman9d0919f2003-11-08 01:05:38 +0000791</div>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000792
Chris Lattner00950542001-06-06 20:29:01 +0000793<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +0000794<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000795<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +0000796<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +0000797<p>The function type can be thought of as a function signature. It
798consists of a return type and a list of formal parameter types.
John Criswell009900b2003-11-25 21:45:46 +0000799Function types are usually used to build virtual function tables
Chris Lattner261efe92003-11-25 01:02:51 +0000800(which are structures of pointers to functions), for indirect function
801calls, and when defining a function.</p>
John Criswell009900b2003-11-25 21:45:46 +0000802<p>
803The return type of a function type cannot be an aggregate type.
804</p>
Chris Lattner00950542001-06-06 20:29:01 +0000805<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +0000806<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
John Criswell0ec250c2005-10-24 16:17:18 +0000807<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
Misha Brukmanc24b7582004-08-12 20:16:08 +0000808specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
Chris Lattner27f71f22003-09-03 00:41:47 +0000809which indicates that the function takes a variable number of arguments.
810Variable argument functions can access their arguments with the <a
Chris Lattner261efe92003-11-25 01:02:51 +0000811 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
Chris Lattner00950542001-06-06 20:29:01 +0000812<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000813<table class="layout">
814 <tr class="layout">
815 <td class="left">
816 <tt>int (int)</tt> <br/>
817 <tt>float (int, int *) *</tt><br/>
818 <tt>int (sbyte *, ...)</tt><br/>
819 </td>
820 <td class="left">
821 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
822 <a href="#t_pointer">Pointer</a> to a function that takes an
Misha Brukmanc24b7582004-08-12 20:16:08 +0000823 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
Reid Spencerd3f876c2004-11-01 08:19:36 +0000824 returning <tt>float</tt>.<br/>
825 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
826 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
827 the signature for <tt>printf</tt> in LLVM.<br/>
828 </td>
829 </tr>
Chris Lattner00950542001-06-06 20:29:01 +0000830</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +0000831
Misha Brukman9d0919f2003-11-08 01:05:38 +0000832</div>
Chris Lattner00950542001-06-06 20:29:01 +0000833<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +0000834<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000835<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +0000836<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +0000837<p>The structure type is used to represent a collection of data members
838together in memory. The packing of the field types is defined to match
839the ABI of the underlying processor. The elements of a structure may
840be any type that has a size.</p>
841<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
842and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
843field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
844instruction.</p>
Chris Lattner00950542001-06-06 20:29:01 +0000845<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +0000846<pre> { &lt;type list&gt; }<br></pre>
Chris Lattner00950542001-06-06 20:29:01 +0000847<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000848<table class="layout">
849 <tr class="layout">
850 <td class="left">
851 <tt>{ int, int, int }</tt><br/>
852 <tt>{ float, int (int) * }</tt><br/>
853 </td>
854 <td class="left">
855 a triple of three <tt>int</tt> values<br/>
856 A pair, where the first element is a <tt>float</tt> and the second element
857 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
858 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
859 </td>
860 </tr>
Chris Lattner00950542001-06-06 20:29:01 +0000861</table>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000862</div>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000863
Chris Lattner00950542001-06-06 20:29:01 +0000864<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +0000865<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000866<div class="doc_text">
Chris Lattner7faa8832002-04-14 06:13:44 +0000867<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +0000868<p>As in many languages, the pointer type represents a pointer or
869reference to another object, which must live in memory.</p>
Chris Lattner7faa8832002-04-14 06:13:44 +0000870<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +0000871<pre> &lt;type&gt; *<br></pre>
Chris Lattner7faa8832002-04-14 06:13:44 +0000872<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000873<table class="layout">
874 <tr class="layout">
875 <td class="left">
876 <tt>[4x int]*</tt><br/>
877 <tt>int (int *) *</tt><br/>
878 </td>
879 <td class="left">
880 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
881 four <tt>int</tt> values<br/>
882 A <a href="#t_pointer">pointer</a> to a <a
Chris Lattnera977c482005-02-19 02:22:14 +0000883 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
Reid Spencerd3f876c2004-11-01 08:19:36 +0000884 <tt>int</tt>.<br/>
885 </td>
886 </tr>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000887</table>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000888</div>
Reid Spencerd3f876c2004-11-01 08:19:36 +0000889
Chris Lattnera58561b2004-08-12 19:12:28 +0000890<!-- _______________________________________________________________________ -->
891<div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000892<div class="doc_text">
Chris Lattner69c11bb2005-04-25 17:34:15 +0000893
Chris Lattnera58561b2004-08-12 19:12:28 +0000894<h5>Overview:</h5>
Chris Lattner69c11bb2005-04-25 17:34:15 +0000895
Chris Lattnera58561b2004-08-12 19:12:28 +0000896<p>A packed type is a simple derived type that represents a vector
897of elements. Packed types are used when multiple primitive data
898are operated in parallel using a single instruction (SIMD).
899A packed type requires a size (number of
Chris Lattnerb8d172f2005-11-10 01:44:22 +0000900elements) and an underlying primitive data type. Vectors must have a power
901of two length (1, 2, 4, 8, 16 ...). Packed types are
Chris Lattnera58561b2004-08-12 19:12:28 +0000902considered <a href="#t_firstclass">first class</a>.</p>
Chris Lattner69c11bb2005-04-25 17:34:15 +0000903
Chris Lattnera58561b2004-08-12 19:12:28 +0000904<h5>Syntax:</h5>
Chris Lattner69c11bb2005-04-25 17:34:15 +0000905
906<pre>
907 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
908</pre>
909
John Criswellc1f786c2005-05-13 22:25:59 +0000910<p>The number of elements is a constant integer value; elementtype may
Chris Lattnera58561b2004-08-12 19:12:28 +0000911be any integral or floating point type.</p>
Chris Lattner69c11bb2005-04-25 17:34:15 +0000912
Chris Lattnera58561b2004-08-12 19:12:28 +0000913<h5>Examples:</h5>
Chris Lattner69c11bb2005-04-25 17:34:15 +0000914
Reid Spencerd3f876c2004-11-01 08:19:36 +0000915<table class="layout">
916 <tr class="layout">
917 <td class="left">
918 <tt>&lt;4 x int&gt;</tt><br/>
919 <tt>&lt;8 x float&gt;</tt><br/>
920 <tt>&lt;2 x uint&gt;</tt><br/>
921 </td>
922 <td class="left">
923 Packed vector of 4 integer values.<br/>
924 Packed vector of 8 floating-point values.<br/>
925 Packed vector of 2 unsigned integer values.<br/>
926 </td>
927 </tr>
928</table>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000929</div>
930
Chris Lattner69c11bb2005-04-25 17:34:15 +0000931<!-- _______________________________________________________________________ -->
932<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
933<div class="doc_text">
934
935<h5>Overview:</h5>
936
937<p>Opaque types are used to represent unknown types in the system. This
938corresponds (for example) to the C notion of a foward declared structure type.
939In LLVM, opaque types can eventually be resolved to any type (not just a
940structure type).</p>
941
942<h5>Syntax:</h5>
943
944<pre>
945 opaque
946</pre>
947
948<h5>Examples:</h5>
949
950<table class="layout">
951 <tr class="layout">
952 <td class="left">
953 <tt>opaque</tt>
954 </td>
955 <td class="left">
956 An opaque type.<br/>
957 </td>
958 </tr>
959</table>
960</div>
961
962
Chris Lattnerc3f59762004-12-09 17:30:23 +0000963<!-- *********************************************************************** -->
964<div class="doc_section"> <a name="constants">Constants</a> </div>
965<!-- *********************************************************************** -->
966
967<div class="doc_text">
968
969<p>LLVM has several different basic types of constants. This section describes
970them all and their syntax.</p>
971
972</div>
973
974<!-- ======================================================================= -->
Reid Spencercc16dc32004-12-09 18:02:53 +0000975<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
Chris Lattnerc3f59762004-12-09 17:30:23 +0000976
977<div class="doc_text">
978
979<dl>
980 <dt><b>Boolean constants</b></dt>
981
982 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
983 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
984 </dd>
985
986 <dt><b>Integer constants</b></dt>
987
Reid Spencercc16dc32004-12-09 18:02:53 +0000988 <dd>Standard integers (such as '4') are constants of the <a
Chris Lattnerc3f59762004-12-09 17:30:23 +0000989 href="#t_integer">integer</a> type. Negative numbers may be used with signed
990 integer types.
991 </dd>
992
993 <dt><b>Floating point constants</b></dt>
994
995 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
996 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerc3f59762004-12-09 17:30:23 +0000997 notation (see below). Floating point constants must have a <a
998 href="#t_floating">floating point</a> type. </dd>
999
1000 <dt><b>Null pointer constants</b></dt>
1001
John Criswell9e2485c2004-12-10 15:51:16 +00001002 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
Chris Lattnerc3f59762004-12-09 17:30:23 +00001003 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1004
1005</dl>
1006
John Criswell9e2485c2004-12-10 15:51:16 +00001007<p>The one non-intuitive notation for constants is the optional hexadecimal form
Chris Lattnerc3f59762004-12-09 17:30:23 +00001008of floating point constants. For example, the form '<tt>double
10090x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
10104.5e+15</tt>'. The only time hexadecimal floating point constants are required
Reid Spencercc16dc32004-12-09 18:02:53 +00001011(and the only time that they are generated by the disassembler) is when a
1012floating point constant must be emitted but it cannot be represented as a
1013decimal floating point number. For example, NaN's, infinities, and other
1014special values are represented in their IEEE hexadecimal format so that
1015assembly and disassembly do not cause any bits to change in the constants.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001016
1017</div>
1018
1019<!-- ======================================================================= -->
1020<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1021</div>
1022
1023<div class="doc_text">
Chris Lattnerd4f6b172005-03-07 22:13:59 +00001024<p>Aggregate constants arise from aggregation of simple constants
1025and smaller aggregate constants.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001026
1027<dl>
1028 <dt><b>Structure constants</b></dt>
1029
1030 <dd>Structure constants are represented with notation similar to structure
1031 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattnerd4f6b172005-03-07 22:13:59 +00001032 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
1033 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
1034 must have <a href="#t_struct">structure type</a>, and the number and
Chris Lattnerc3f59762004-12-09 17:30:23 +00001035 types of elements must match those specified by the type.
1036 </dd>
1037
1038 <dt><b>Array constants</b></dt>
1039
1040 <dd>Array constants are represented with notation similar to array type
1041 definitions (a comma separated list of elements, surrounded by square brackets
John Criswell9e2485c2004-12-10 15:51:16 +00001042 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
Chris Lattnerc3f59762004-12-09 17:30:23 +00001043 constants must have <a href="#t_array">array type</a>, and the number and
1044 types of elements must match those specified by the type.
1045 </dd>
1046
1047 <dt><b>Packed constants</b></dt>
1048
1049 <dd>Packed constants are represented with notation similar to packed type
1050 definitions (a comma separated list of elements, surrounded by
John Criswell9e2485c2004-12-10 15:51:16 +00001051 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; int 42,
Chris Lattnerc3f59762004-12-09 17:30:23 +00001052 int 11, int 74, int 100 &gt;</tt>". Packed constants must have <a
1053 href="#t_packed">packed type</a>, and the number and types of elements must
1054 match those specified by the type.
1055 </dd>
1056
1057 <dt><b>Zero initialization</b></dt>
1058
1059 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1060 value to zero of <em>any</em> type, including scalar and aggregate types.
1061 This is often used to avoid having to print large zero initializers (e.g. for
John Criswell0ec250c2005-10-24 16:17:18 +00001062 large arrays) and is always exactly equivalent to using explicit zero
Chris Lattnerc3f59762004-12-09 17:30:23 +00001063 initializers.
1064 </dd>
1065</dl>
1066
1067</div>
1068
1069<!-- ======================================================================= -->
1070<div class="doc_subsection">
1071 <a name="globalconstants">Global Variable and Function Addresses</a>
1072</div>
1073
1074<div class="doc_text">
1075
1076<p>The addresses of <a href="#globalvars">global variables</a> and <a
1077href="#functionstructure">functions</a> are always implicitly valid (link-time)
John Criswell9e2485c2004-12-10 15:51:16 +00001078constants. These constants are explicitly referenced when the <a
1079href="#identifiers">identifier for the global</a> is used and always have <a
Chris Lattnerc3f59762004-12-09 17:30:23 +00001080href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1081file:</p>
1082
1083<pre>
1084 %X = global int 17
1085 %Y = global int 42
1086 %Z = global [2 x int*] [ int* %X, int* %Y ]
1087</pre>
1088
1089</div>
1090
1091<!-- ======================================================================= -->
Reid Spencer2dc45b82004-12-09 18:13:12 +00001092<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001093<div class="doc_text">
Reid Spencer2dc45b82004-12-09 18:13:12 +00001094 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
John Criswellc1f786c2005-05-13 22:25:59 +00001095 no specific value. Undefined values may be of any type and be used anywhere
Reid Spencer2dc45b82004-12-09 18:13:12 +00001096 a constant is permitted.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001097
Reid Spencer2dc45b82004-12-09 18:13:12 +00001098 <p>Undefined values indicate to the compiler that the program is well defined
1099 no matter what value is used, giving the compiler more freedom to optimize.
1100 </p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001101</div>
1102
1103<!-- ======================================================================= -->
1104<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1105</div>
1106
1107<div class="doc_text">
1108
1109<p>Constant expressions are used to allow expressions involving other constants
1110to be used as constants. Constant expressions may be of any <a
John Criswellc1f786c2005-05-13 22:25:59 +00001111href="#t_firstclass">first class</a> type and may involve any LLVM operation
Chris Lattnerc3f59762004-12-09 17:30:23 +00001112that does not have side effects (e.g. load and call are not supported). The
1113following is the syntax for constant expressions:</p>
1114
1115<dl>
1116 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
1117
1118 <dd>Cast a constant to another type.</dd>
1119
1120 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1121
1122 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1123 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1124 instruction, the index list may have zero or more indexes, which are required
1125 to make sense for the type of "CSTPTR".</dd>
1126
Robert Bocchino9fbe1452006-01-10 19:31:34 +00001127 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1128
1129 <dd>Perform the <a href="#i_select">select operation</a> on
1130 constants.
1131
1132 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1133
1134 <dd>Perform the <a href="#i_extractelement">extractelement
1135 operation</a> on constants.
1136
Robert Bocchino05ccd702006-01-15 20:48:27 +00001137 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1138
1139 <dd>Perform the <a href="#i_insertelement">insertelement
1140 operation</a> on constants.
1141
Chris Lattnerc3f59762004-12-09 17:30:23 +00001142 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1143
Reid Spencer2dc45b82004-12-09 18:13:12 +00001144 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1145 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
Chris Lattnerc3f59762004-12-09 17:30:23 +00001146 binary</a> operations. The constraints on operands are the same as those for
1147 the corresponding instruction (e.g. no bitwise operations on floating point
John Criswelle4c57cc2005-05-12 16:52:32 +00001148 values are allowed).</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001149</dl>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001150</div>
Chris Lattner9ee5d222004-03-08 16:49:10 +00001151
Chris Lattner00950542001-06-06 20:29:01 +00001152<!-- *********************************************************************** -->
Chris Lattnere87d6532006-01-25 23:47:57 +00001153<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1154<!-- *********************************************************************** -->
1155
1156<!-- ======================================================================= -->
1157<div class="doc_subsection">
1158<a name="inlineasm">Inline Assembler Expressions</a>
1159</div>
1160
1161<div class="doc_text">
1162
1163<p>
1164LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1165Module-Level Inline Assembly</a>) through the use of a special value. This
1166value represents the inline assembler as a string (containing the instructions
1167to emit), a list of operand constraints (stored as a string), and a flag that
1168indicates whether or not the inline asm expression has side effects. An example
1169inline assembler expression is:
1170</p>
1171
1172<pre>
1173 int(int) asm "bswap $0", "=r,r"
1174</pre>
1175
1176<p>
1177Inline assembler expressions may <b>only</b> be used as the callee operand of
1178a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1179</p>
1180
1181<pre>
1182 %X = call int asm "<a href="#i_bswap">bswap</a> $0", "=r,r"(int %Y)
1183</pre>
1184
1185<p>
1186Inline asms with side effects not visible in the constraint list must be marked
1187as having side effects. This is done through the use of the
1188'<tt>sideeffect</tt>' keyword, like so:
1189</p>
1190
1191<pre>
1192 call void asm sideeffect "eieio", ""()
1193</pre>
1194
1195<p>TODO: The format of the asm and constraints string still need to be
1196documented here. Constraints on what can be done (e.g. duplication, moving, etc
1197need to be documented).
1198</p>
1199
1200</div>
1201
1202<!-- *********************************************************************** -->
Chris Lattner261efe92003-11-25 01:02:51 +00001203<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1204<!-- *********************************************************************** -->
Chris Lattnerc3f59762004-12-09 17:30:23 +00001205
Misha Brukman9d0919f2003-11-08 01:05:38 +00001206<div class="doc_text">
Chris Lattnerc3f59762004-12-09 17:30:23 +00001207
Chris Lattner261efe92003-11-25 01:02:51 +00001208<p>The LLVM instruction set consists of several different
1209classifications of instructions: <a href="#terminators">terminator
John Criswellc1f786c2005-05-13 22:25:59 +00001210instructions</a>, <a href="#binaryops">binary instructions</a>,
1211<a href="#bitwiseops">bitwise binary instructions</a>, <a
Chris Lattner261efe92003-11-25 01:02:51 +00001212 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1213instructions</a>.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001214
Misha Brukman9d0919f2003-11-08 01:05:38 +00001215</div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001216
Chris Lattner00950542001-06-06 20:29:01 +00001217<!-- ======================================================================= -->
Chris Lattner261efe92003-11-25 01:02:51 +00001218<div class="doc_subsection"> <a name="terminators">Terminator
1219Instructions</a> </div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001220
Misha Brukman9d0919f2003-11-08 01:05:38 +00001221<div class="doc_text">
Chris Lattnerc3f59762004-12-09 17:30:23 +00001222
Chris Lattner261efe92003-11-25 01:02:51 +00001223<p>As mentioned <a href="#functionstructure">previously</a>, every
1224basic block in a program ends with a "Terminator" instruction, which
1225indicates which block should be executed after the current block is
1226finished. These terminator instructions typically yield a '<tt>void</tt>'
1227value: they produce control flow, not values (the one exception being
1228the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
John Criswell9e2485c2004-12-10 15:51:16 +00001229<p>There are six different terminator instructions: the '<a
Chris Lattner261efe92003-11-25 01:02:51 +00001230 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1231instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
Chris Lattner35eca582004-10-16 18:04:13 +00001232the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1233 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1234 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001235
Misha Brukman9d0919f2003-11-08 01:05:38 +00001236</div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001237
Chris Lattner00950542001-06-06 20:29:01 +00001238<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001239<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1240Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001241<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001242<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001243<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Chris Lattner7faa8832002-04-14 06:13:44 +00001244 ret void <i>; Return from void function</i>
Chris Lattner00950542001-06-06 20:29:01 +00001245</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001246<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001247<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
John Criswellc1f786c2005-05-13 22:25:59 +00001248value) from a function back to the caller.</p>
John Criswell4457dc92004-04-09 16:48:45 +00001249<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Chris Lattner261efe92003-11-25 01:02:51 +00001250returns a value and then causes control flow, and one that just causes
1251control flow to occur.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001252<h5>Arguments:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001253<p>The '<tt>ret</tt>' instruction may return any '<a
1254 href="#t_firstclass">first class</a>' type. Notice that a function is
1255not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1256instruction inside of the function that returns a value that does not
1257match the return type of the function.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001258<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001259<p>When the '<tt>ret</tt>' instruction is executed, control flow
1260returns back to the calling function's context. If the caller is a "<a
John Criswellfa081872004-06-25 15:16:57 +00001261 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
Chris Lattner261efe92003-11-25 01:02:51 +00001262the instruction after the call. If the caller was an "<a
1263 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
John Criswelle4c57cc2005-05-12 16:52:32 +00001264at the beginning of the "normal" destination block. If the instruction
Chris Lattner261efe92003-11-25 01:02:51 +00001265returns a value, that value shall set the call or invoke instruction's
1266return value.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001267<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001268<pre> ret int 5 <i>; Return an integer value of 5</i>
Chris Lattner7faa8832002-04-14 06:13:44 +00001269 ret void <i>; Return from a void function</i>
Chris Lattner00950542001-06-06 20:29:01 +00001270</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001271</div>
Chris Lattner00950542001-06-06 20:29:01 +00001272<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001273<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001274<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001275<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001276<pre> br bool &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
Chris Lattner00950542001-06-06 20:29:01 +00001277</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001278<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001279<p>The '<tt>br</tt>' instruction is used to cause control flow to
1280transfer to a different basic block in the current function. There are
1281two forms of this instruction, corresponding to a conditional branch
1282and an unconditional branch.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001283<h5>Arguments:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001284<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1285single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1286unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1287value as a target.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001288<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001289<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1290argument is evaluated. If the value is <tt>true</tt>, control flows
1291to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1292control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001293<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001294<pre>Test:<br> %cond = <a href="#i_setcc">seteq</a> int %a, %b<br> br bool %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
1295 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001296</div>
Chris Lattner00950542001-06-06 20:29:01 +00001297<!-- _______________________________________________________________________ -->
Chris Lattnerc88c17b2004-02-24 04:54:45 +00001298<div class="doc_subsubsection">
1299 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1300</div>
1301
Misha Brukman9d0919f2003-11-08 01:05:38 +00001302<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001303<h5>Syntax:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00001304
1305<pre>
1306 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1307</pre>
1308
Chris Lattner00950542001-06-06 20:29:01 +00001309<h5>Overview:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00001310
1311<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1312several different places. It is a generalization of the '<tt>br</tt>'
Misha Brukman9d0919f2003-11-08 01:05:38 +00001313instruction, allowing a branch to occur to one of many possible
1314destinations.</p>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00001315
1316
Chris Lattner00950542001-06-06 20:29:01 +00001317<h5>Arguments:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00001318
1319<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1320comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1321an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1322table is not allowed to contain duplicate constant entries.</p>
1323
Chris Lattner00950542001-06-06 20:29:01 +00001324<h5>Semantics:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00001325
Chris Lattner261efe92003-11-25 01:02:51 +00001326<p>The <tt>switch</tt> instruction specifies a table of values and
1327destinations. When the '<tt>switch</tt>' instruction is executed, this
John Criswell84114752004-06-25 16:05:06 +00001328table is searched for the given value. If the value is found, control flow is
1329transfered to the corresponding destination; otherwise, control flow is
1330transfered to the default destination.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001331
Chris Lattnerc88c17b2004-02-24 04:54:45 +00001332<h5>Implementation:</h5>
1333
1334<p>Depending on properties of the target machine and the particular
1335<tt>switch</tt> instruction, this instruction may be code generated in different
John Criswell84114752004-06-25 16:05:06 +00001336ways. For example, it could be generated as a series of chained conditional
1337branches or with a lookup table.</p>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00001338
1339<h5>Example:</h5>
1340
1341<pre>
1342 <i>; Emulate a conditional br instruction</i>
1343 %Val = <a href="#i_cast">cast</a> bool %value to int
1344 switch int %Val, label %truedest [int 0, label %falsedest ]
1345
1346 <i>; Emulate an unconditional br instruction</i>
1347 switch uint 0, label %dest [ ]
1348
1349 <i>; Implement a jump table:</i>
1350 switch uint %val, label %otherwise [ uint 0, label %onzero
1351 uint 1, label %onone
1352 uint 2, label %ontwo ]
Chris Lattner00950542001-06-06 20:29:01 +00001353</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001354</div>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001355
Chris Lattner00950542001-06-06 20:29:01 +00001356<!-- _______________________________________________________________________ -->
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001357<div class="doc_subsubsection">
1358 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1359</div>
1360
Misha Brukman9d0919f2003-11-08 01:05:38 +00001361<div class="doc_text">
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001362
Chris Lattner00950542001-06-06 20:29:01 +00001363<h5>Syntax:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001364
1365<pre>
1366 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
1367 to label &lt;normal label&gt; except label &lt;exception label&gt;
1368</pre>
1369
Chris Lattner6536cfe2002-05-06 22:08:29 +00001370<h5>Overview:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001371
1372<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1373function, with the possibility of control flow transfer to either the
John Criswelle4c57cc2005-05-12 16:52:32 +00001374'<tt>normal</tt>' label or the
1375'<tt>exception</tt>' label. If the callee function returns with the
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001376"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1377"normal" label. If the callee (or any indirect callees) returns with the "<a
John Criswelle4c57cc2005-05-12 16:52:32 +00001378href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1379continued at the dynamically nearest "exception" label.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001380
Chris Lattner00950542001-06-06 20:29:01 +00001381<h5>Arguments:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001382
Misha Brukman9d0919f2003-11-08 01:05:38 +00001383<p>This instruction requires several arguments:</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001384
Chris Lattner00950542001-06-06 20:29:01 +00001385<ol>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001386 <li>
John Criswellc1f786c2005-05-13 22:25:59 +00001387 The optional "cconv" marker indicates which <a href="callingconv">calling
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001388 convention</a> the call should use. If none is specified, the call defaults
1389 to using C calling conventions.
1390 </li>
1391 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1392 function value being invoked. In most cases, this is a direct function
1393 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1394 an arbitrary pointer to function value.
1395 </li>
1396
1397 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1398 function to be invoked. </li>
1399
1400 <li>'<tt>function args</tt>': argument list whose types match the function
1401 signature argument types. If the function signature indicates the function
1402 accepts a variable number of arguments, the extra arguments can be
1403 specified. </li>
1404
1405 <li>'<tt>normal label</tt>': the label reached when the called function
1406 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1407
1408 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1409 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1410
Chris Lattner00950542001-06-06 20:29:01 +00001411</ol>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001412
Chris Lattner00950542001-06-06 20:29:01 +00001413<h5>Semantics:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001414
Misha Brukman9d0919f2003-11-08 01:05:38 +00001415<p>This instruction is designed to operate as a standard '<tt><a
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001416href="#i_call">call</a></tt>' instruction in most regards. The primary
1417difference is that it establishes an association with a label, which is used by
1418the runtime library to unwind the stack.</p>
1419
1420<p>This instruction is used in languages with destructors to ensure that proper
1421cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1422exception. Additionally, this is important for implementation of
1423'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1424
Chris Lattner00950542001-06-06 20:29:01 +00001425<h5>Example:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001426<pre>
1427 %retval = invoke int %Test(int 15) to label %Continue
1428 except label %TestCleanup <i>; {int}:retval set</i>
1429 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1430 except label %TestCleanup <i>; {int}:retval set</i>
Chris Lattner00950542001-06-06 20:29:01 +00001431</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001432</div>
Chris Lattner35eca582004-10-16 18:04:13 +00001433
1434
Chris Lattner27f71f22003-09-03 00:41:47 +00001435<!-- _______________________________________________________________________ -->
Chris Lattner35eca582004-10-16 18:04:13 +00001436
Chris Lattner261efe92003-11-25 01:02:51 +00001437<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1438Instruction</a> </div>
Chris Lattner35eca582004-10-16 18:04:13 +00001439
Misha Brukman9d0919f2003-11-08 01:05:38 +00001440<div class="doc_text">
Chris Lattner35eca582004-10-16 18:04:13 +00001441
Chris Lattner27f71f22003-09-03 00:41:47 +00001442<h5>Syntax:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00001443<pre>
1444 unwind
1445</pre>
1446
Chris Lattner27f71f22003-09-03 00:41:47 +00001447<h5>Overview:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00001448
1449<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1450at the first callee in the dynamic call stack which used an <a
1451href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1452primarily used to implement exception handling.</p>
1453
Chris Lattner27f71f22003-09-03 00:41:47 +00001454<h5>Semantics:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00001455
1456<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1457immediately halt. The dynamic call stack is then searched for the first <a
1458href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1459execution continues at the "exceptional" destination block specified by the
1460<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1461dynamic call chain, undefined behavior results.</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001462</div>
Chris Lattner35eca582004-10-16 18:04:13 +00001463
1464<!-- _______________________________________________________________________ -->
1465
1466<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1467Instruction</a> </div>
1468
1469<div class="doc_text">
1470
1471<h5>Syntax:</h5>
1472<pre>
1473 unreachable
1474</pre>
1475
1476<h5>Overview:</h5>
1477
1478<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1479instruction is used to inform the optimizer that a particular portion of the
1480code is not reachable. This can be used to indicate that the code after a
1481no-return function cannot be reached, and other facts.</p>
1482
1483<h5>Semantics:</h5>
1484
1485<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1486</div>
1487
1488
1489
Chris Lattner00950542001-06-06 20:29:01 +00001490<!-- ======================================================================= -->
Chris Lattner261efe92003-11-25 01:02:51 +00001491<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001492<div class="doc_text">
Chris Lattner261efe92003-11-25 01:02:51 +00001493<p>Binary operators are used to do most of the computation in a
1494program. They require two operands, execute an operation on them, and
John Criswell9e2485c2004-12-10 15:51:16 +00001495produce a single value. The operands might represent
Chris Lattnera58561b2004-08-12 19:12:28 +00001496multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1497The result value of a binary operator is not
Chris Lattner261efe92003-11-25 01:02:51 +00001498necessarily the same type as its operands.</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001499<p>There are several different binary operators:</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001500</div>
Chris Lattner00950542001-06-06 20:29:01 +00001501<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001502<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1503Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001504<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001505<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001506<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001507</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001508<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001509<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001510<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001511<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
Chris Lattnera58561b2004-08-12 19:12:28 +00001512 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1513 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1514Both arguments must have identical types.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001515<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001516<p>The value produced is the integer or floating point sum of the two
1517operands.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001518<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001519<pre> &lt;result&gt; = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00001520</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001521</div>
Chris Lattner00950542001-06-06 20:29:01 +00001522<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001523<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1524Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001525<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001526<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001527<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001528</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001529<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001530<p>The '<tt>sub</tt>' instruction returns the difference of its two
1531operands.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00001532<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1533instruction present in most other intermediate representations.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001534<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001535<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
Chris Lattner261efe92003-11-25 01:02:51 +00001536 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
Chris Lattnera58561b2004-08-12 19:12:28 +00001537values.
1538This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1539Both arguments must have identical types.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001540<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001541<p>The value produced is the integer or floating point difference of
1542the two operands.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001543<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001544<pre> &lt;result&gt; = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00001545 &lt;result&gt; = sub int 0, %val <i>; yields {int}:result = -%var</i>
1546</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001547</div>
Chris Lattner00950542001-06-06 20:29:01 +00001548<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001549<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1550Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001551<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001552<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001553<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001554</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001555<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001556<p>The '<tt>mul</tt>' instruction returns the product of its two
1557operands.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001558<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001559<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
Chris Lattner261efe92003-11-25 01:02:51 +00001560 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
Chris Lattnera58561b2004-08-12 19:12:28 +00001561values.
1562This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1563Both arguments must have identical types.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001564<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001565<p>The value produced is the integer or floating point product of the
Misha Brukman9d0919f2003-11-08 01:05:38 +00001566two operands.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00001567<p>There is no signed vs unsigned multiplication. The appropriate
1568action is taken based on the type of the operand.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001569<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001570<pre> &lt;result&gt; = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00001571</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001572</div>
Chris Lattner00950542001-06-06 20:29:01 +00001573<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001574<div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1575Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001576<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001577<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001578<pre> &lt;result&gt; = div &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
1579</pre>
1580<h5>Overview:</h5>
1581<p>The '<tt>div</tt>' instruction returns the quotient of its two
1582operands.</p>
1583<h5>Arguments:</h5>
1584<p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1585 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
Chris Lattnera58561b2004-08-12 19:12:28 +00001586values.
1587This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1588Both arguments must have identical types.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00001589<h5>Semantics:</h5>
1590<p>The value produced is the integer or floating point quotient of the
1591two operands.</p>
1592<h5>Example:</h5>
1593<pre> &lt;result&gt; = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1594</pre>
1595</div>
1596<!-- _______________________________________________________________________ -->
1597<div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1598Instruction</a> </div>
1599<div class="doc_text">
1600<h5>Syntax:</h5>
1601<pre> &lt;result&gt; = rem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
1602</pre>
1603<h5>Overview:</h5>
1604<p>The '<tt>rem</tt>' instruction returns the remainder from the
1605division of its two operands.</p>
1606<h5>Arguments:</h5>
1607<p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1608 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
Chris Lattnera58561b2004-08-12 19:12:28 +00001609values.
1610This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1611Both arguments must have identical types.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00001612<h5>Semantics:</h5>
1613<p>This returns the <i>remainder</i> of a division (where the result
1614has the same sign as the divisor), not the <i>modulus</i> (where the
1615result has the same sign as the dividend) of a value. For more
John Criswell0ec250c2005-10-24 16:17:18 +00001616information about the difference, see <a
Chris Lattner261efe92003-11-25 01:02:51 +00001617 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1618Math Forum</a>.</p>
1619<h5>Example:</h5>
1620<pre> &lt;result&gt; = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1621</pre>
1622</div>
1623<!-- _______________________________________________________________________ -->
1624<div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1625Instructions</a> </div>
1626<div class="doc_text">
1627<h5>Syntax:</h5>
1628<pre> &lt;result&gt; = seteq &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {bool}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001629 &lt;result&gt; = setne &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {bool}:result</i>
1630 &lt;result&gt; = setlt &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {bool}:result</i>
1631 &lt;result&gt; = setgt &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {bool}:result</i>
1632 &lt;result&gt; = setle &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {bool}:result</i>
1633 &lt;result&gt; = setge &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {bool}:result</i>
1634</pre>
Chris Lattner261efe92003-11-25 01:02:51 +00001635<h5>Overview:</h5>
1636<p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1637value based on a comparison of their two operands.</p>
1638<h5>Arguments:</h5>
1639<p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1640be of <a href="#t_firstclass">first class</a> type (it is not possible
1641to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1642or '<tt>void</tt>' values, etc...). Both arguments must have identical
1643types.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001644<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001645<p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1646value if both operands are equal.<br>
1647The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1648value if both operands are unequal.<br>
1649The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1650value if the first operand is less than the second operand.<br>
1651The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1652value if the first operand is greater than the second operand.<br>
1653The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1654value if the first operand is less than or equal to the second operand.<br>
1655The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1656value if the first operand is greater than or equal to the second
1657operand.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001658<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001659<pre> &lt;result&gt; = seteq int 4, 5 <i>; yields {bool}:result = false</i>
Chris Lattner00950542001-06-06 20:29:01 +00001660 &lt;result&gt; = setne float 4, 5 <i>; yields {bool}:result = true</i>
1661 &lt;result&gt; = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1662 &lt;result&gt; = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1663 &lt;result&gt; = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1664 &lt;result&gt; = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1665</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001666</div>
Chris Lattner00950542001-06-06 20:29:01 +00001667<!-- ======================================================================= -->
Chris Lattner261efe92003-11-25 01:02:51 +00001668<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1669Operations</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001670<div class="doc_text">
Chris Lattner261efe92003-11-25 01:02:51 +00001671<p>Bitwise binary operators are used to do various forms of
1672bit-twiddling in a program. They are generally very efficient
John Criswell9e2485c2004-12-10 15:51:16 +00001673instructions and can commonly be strength reduced from other
Chris Lattner261efe92003-11-25 01:02:51 +00001674instructions. They require two operands, execute an operation on them,
1675and produce a single value. The resulting value of the bitwise binary
1676operators is always the same type as its first operand.</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001677</div>
Chris Lattner00950542001-06-06 20:29:01 +00001678<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001679<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1680Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001681<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001682<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001683<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001684</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001685<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001686<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1687its two operands.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001688<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001689<p>The two arguments to the '<tt>and</tt>' instruction must be <a
Chris Lattner261efe92003-11-25 01:02:51 +00001690 href="#t_integral">integral</a> values. Both arguments must have
1691identical types.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001692<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001693<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
Chris Lattner261efe92003-11-25 01:02:51 +00001694<p> </p>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001695<div style="align: center">
Misha Brukman9d0919f2003-11-08 01:05:38 +00001696<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner261efe92003-11-25 01:02:51 +00001697 <tbody>
1698 <tr>
1699 <td>In0</td>
1700 <td>In1</td>
1701 <td>Out</td>
1702 </tr>
1703 <tr>
1704 <td>0</td>
1705 <td>0</td>
1706 <td>0</td>
1707 </tr>
1708 <tr>
1709 <td>0</td>
1710 <td>1</td>
1711 <td>0</td>
1712 </tr>
1713 <tr>
1714 <td>1</td>
1715 <td>0</td>
1716 <td>0</td>
1717 </tr>
1718 <tr>
1719 <td>1</td>
1720 <td>1</td>
1721 <td>1</td>
1722 </tr>
1723 </tbody>
1724</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001725</div>
Chris Lattner00950542001-06-06 20:29:01 +00001726<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001727<pre> &lt;result&gt; = and int 4, %var <i>; yields {int}:result = 4 &amp; %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00001728 &lt;result&gt; = and int 15, 40 <i>; yields {int}:result = 8</i>
1729 &lt;result&gt; = and int 4, 8 <i>; yields {int}:result = 0</i>
1730</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001731</div>
Chris Lattner00950542001-06-06 20:29:01 +00001732<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001733<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001734<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001735<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001736<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001737</pre>
Chris Lattner261efe92003-11-25 01:02:51 +00001738<h5>Overview:</h5>
1739<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1740or of its two operands.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001741<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001742<p>The two arguments to the '<tt>or</tt>' instruction must be <a
Chris Lattner261efe92003-11-25 01:02:51 +00001743 href="#t_integral">integral</a> values. Both arguments must have
1744identical types.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001745<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001746<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
Chris Lattner261efe92003-11-25 01:02:51 +00001747<p> </p>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001748<div style="align: center">
Chris Lattner261efe92003-11-25 01:02:51 +00001749<table border="1" cellspacing="0" cellpadding="4">
1750 <tbody>
1751 <tr>
1752 <td>In0</td>
1753 <td>In1</td>
1754 <td>Out</td>
1755 </tr>
1756 <tr>
1757 <td>0</td>
1758 <td>0</td>
1759 <td>0</td>
1760 </tr>
1761 <tr>
1762 <td>0</td>
1763 <td>1</td>
1764 <td>1</td>
1765 </tr>
1766 <tr>
1767 <td>1</td>
1768 <td>0</td>
1769 <td>1</td>
1770 </tr>
1771 <tr>
1772 <td>1</td>
1773 <td>1</td>
1774 <td>1</td>
1775 </tr>
1776 </tbody>
1777</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001778</div>
Chris Lattner00950542001-06-06 20:29:01 +00001779<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001780<pre> &lt;result&gt; = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00001781 &lt;result&gt; = or int 15, 40 <i>; yields {int}:result = 47</i>
1782 &lt;result&gt; = or int 4, 8 <i>; yields {int}:result = 12</i>
1783</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001784</div>
Chris Lattner00950542001-06-06 20:29:01 +00001785<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001786<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1787Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001788<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001789<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001790<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001791</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001792<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001793<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1794or of its two operands. The <tt>xor</tt> is used to implement the
1795"one's complement" operation, which is the "~" operator in C.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001796<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001797<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
Chris Lattner261efe92003-11-25 01:02:51 +00001798 href="#t_integral">integral</a> values. Both arguments must have
1799identical types.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001800<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001801<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
Chris Lattner261efe92003-11-25 01:02:51 +00001802<p> </p>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001803<div style="align: center">
Chris Lattner261efe92003-11-25 01:02:51 +00001804<table border="1" cellspacing="0" cellpadding="4">
1805 <tbody>
1806 <tr>
1807 <td>In0</td>
1808 <td>In1</td>
1809 <td>Out</td>
1810 </tr>
1811 <tr>
1812 <td>0</td>
1813 <td>0</td>
1814 <td>0</td>
1815 </tr>
1816 <tr>
1817 <td>0</td>
1818 <td>1</td>
1819 <td>1</td>
1820 </tr>
1821 <tr>
1822 <td>1</td>
1823 <td>0</td>
1824 <td>1</td>
1825 </tr>
1826 <tr>
1827 <td>1</td>
1828 <td>1</td>
1829 <td>0</td>
1830 </tr>
1831 </tbody>
1832</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001833</div>
Chris Lattner261efe92003-11-25 01:02:51 +00001834<p> </p>
Chris Lattner00950542001-06-06 20:29:01 +00001835<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001836<pre> &lt;result&gt; = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00001837 &lt;result&gt; = xor int 15, 40 <i>; yields {int}:result = 39</i>
1838 &lt;result&gt; = xor int 4, 8 <i>; yields {int}:result = 12</i>
Chris Lattner27f71f22003-09-03 00:41:47 +00001839 &lt;result&gt; = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
Chris Lattner00950542001-06-06 20:29:01 +00001840</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001841</div>
Chris Lattner00950542001-06-06 20:29:01 +00001842<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001843<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1844Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001845<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001846<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001847<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, ubyte &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001848</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001849<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001850<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1851the left a specified number of bits.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001852<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001853<p>The first argument to the '<tt>shl</tt>' instruction must be an <a
Chris Lattner261efe92003-11-25 01:02:51 +00001854 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1855type.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001856<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001857<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001858<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001859<pre> &lt;result&gt; = shl int 4, ubyte %var <i>; yields {int}:result = 4 &lt;&lt; %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00001860 &lt;result&gt; = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1861 &lt;result&gt; = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1862</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001863</div>
Chris Lattner00950542001-06-06 20:29:01 +00001864<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00001865<div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1866Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001867<div class="doc_text">
Chris Lattner00950542001-06-06 20:29:01 +00001868<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001869<pre> &lt;result&gt; = shr &lt;ty&gt; &lt;var1&gt;, ubyte &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001870</pre>
Chris Lattner00950542001-06-06 20:29:01 +00001871<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001872<p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1873the right a specified number of bits.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001874<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001875<p>The first argument to the '<tt>shr</tt>' instruction must be an <a
Chris Lattner261efe92003-11-25 01:02:51 +00001876 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1877type.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001878<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001879<p>If the first argument is a <a href="#t_signed">signed</a> type, the
1880most significant bit is duplicated in the newly free'd bit positions.
1881If the first argument is unsigned, zero bits shall fill the empty
1882positions.</p>
Chris Lattner00950542001-06-06 20:29:01 +00001883<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00001884<pre> &lt;result&gt; = shr int 4, ubyte %var <i>; yields {int}:result = 4 &gt;&gt; %var</i>
Chris Lattner8c6bb902003-06-18 21:30:51 +00001885 &lt;result&gt; = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
Chris Lattner00950542001-06-06 20:29:01 +00001886 &lt;result&gt; = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
Chris Lattner8c6bb902003-06-18 21:30:51 +00001887 &lt;result&gt; = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1888 &lt;result&gt; = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
Chris Lattner00950542001-06-06 20:29:01 +00001889</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001890</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001891
Chris Lattner00950542001-06-06 20:29:01 +00001892<!-- ======================================================================= -->
Chris Lattner2cbdc452005-11-06 08:02:57 +00001893<div class="doc_subsection">
1894 <a name="memoryops">Memory Access Operations</a>
1895</div>
1896
Misha Brukman9d0919f2003-11-08 01:05:38 +00001897<div class="doc_text">
Chris Lattner2cbdc452005-11-06 08:02:57 +00001898
Chris Lattner261efe92003-11-25 01:02:51 +00001899<p>A key design point of an SSA-based representation is how it
1900represents memory. In LLVM, no memory locations are in SSA form, which
1901makes things very simple. This section describes how to read, write,
John Criswell9e2485c2004-12-10 15:51:16 +00001902allocate, and free memory in LLVM.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001903
Misha Brukman9d0919f2003-11-08 01:05:38 +00001904</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001905
Chris Lattner00950542001-06-06 20:29:01 +00001906<!-- _______________________________________________________________________ -->
Chris Lattner2cbdc452005-11-06 08:02:57 +00001907<div class="doc_subsubsection">
1908 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
1909</div>
1910
Misha Brukman9d0919f2003-11-08 01:05:38 +00001911<div class="doc_text">
Chris Lattner2cbdc452005-11-06 08:02:57 +00001912
Chris Lattner00950542001-06-06 20:29:01 +00001913<h5>Syntax:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001914
1915<pre>
1916 &lt;result&gt; = malloc &lt;type&gt;[, uint &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00001917</pre>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001918
Chris Lattner00950542001-06-06 20:29:01 +00001919<h5>Overview:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001920
Chris Lattner261efe92003-11-25 01:02:51 +00001921<p>The '<tt>malloc</tt>' instruction allocates memory from the system
1922heap and returns a pointer to it.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001923
Chris Lattner00950542001-06-06 20:29:01 +00001924<h5>Arguments:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001925
1926<p>The '<tt>malloc</tt>' instruction allocates
1927<tt>sizeof(&lt;type&gt;)*NumElements</tt>
John Criswell6e4ca612004-02-24 16:13:56 +00001928bytes of memory from the operating system and returns a pointer of the
Chris Lattner2cbdc452005-11-06 08:02:57 +00001929appropriate type to the program. If "NumElements" is specified, it is the
1930number of elements allocated. If an alignment is specified, the value result
1931of the allocation is guaranteed to be aligned to at least that boundary. If
1932not specified, or if zero, the target can choose to align the allocation on any
1933convenient boundary.</p>
1934
Misha Brukman9d0919f2003-11-08 01:05:38 +00001935<p>'<tt>type</tt>' must be a sized type.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001936
Chris Lattner00950542001-06-06 20:29:01 +00001937<h5>Semantics:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001938
Chris Lattner261efe92003-11-25 01:02:51 +00001939<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1940a pointer is returned.</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001941
Chris Lattner2cbdc452005-11-06 08:02:57 +00001942<h5>Example:</h5>
1943
1944<pre>
1945 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1946
1947 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
Chris Lattner7faa8832002-04-14 06:13:44 +00001948 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1949 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001950 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
1951 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
Chris Lattner00950542001-06-06 20:29:01 +00001952</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001953</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001954
Chris Lattner00950542001-06-06 20:29:01 +00001955<!-- _______________________________________________________________________ -->
Chris Lattner2cbdc452005-11-06 08:02:57 +00001956<div class="doc_subsubsection">
1957 <a name="i_free">'<tt>free</tt>' Instruction</a>
1958</div>
1959
Misha Brukman9d0919f2003-11-08 01:05:38 +00001960<div class="doc_text">
Chris Lattner2cbdc452005-11-06 08:02:57 +00001961
Chris Lattner00950542001-06-06 20:29:01 +00001962<h5>Syntax:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001963
1964<pre>
1965 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
Chris Lattner00950542001-06-06 20:29:01 +00001966</pre>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001967
Chris Lattner00950542001-06-06 20:29:01 +00001968<h5>Overview:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001969
Chris Lattner261efe92003-11-25 01:02:51 +00001970<p>The '<tt>free</tt>' instruction returns memory back to the unused
John Criswellc1f786c2005-05-13 22:25:59 +00001971memory heap to be reallocated in the future.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001972
Chris Lattner00950542001-06-06 20:29:01 +00001973<h5>Arguments:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001974
Chris Lattner261efe92003-11-25 01:02:51 +00001975<p>'<tt>value</tt>' shall be a pointer value that points to a value
1976that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1977instruction.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001978
Chris Lattner00950542001-06-06 20:29:01 +00001979<h5>Semantics:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001980
John Criswell9e2485c2004-12-10 15:51:16 +00001981<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner261efe92003-11-25 01:02:51 +00001982after this instruction executes.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001983
Chris Lattner00950542001-06-06 20:29:01 +00001984<h5>Example:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001985
1986<pre>
1987 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
Chris Lattner00950542001-06-06 20:29:01 +00001988 free [4 x ubyte]* %array
1989</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001990</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00001991
Chris Lattner00950542001-06-06 20:29:01 +00001992<!-- _______________________________________________________________________ -->
Chris Lattner2cbdc452005-11-06 08:02:57 +00001993<div class="doc_subsubsection">
1994 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
1995</div>
1996
Misha Brukman9d0919f2003-11-08 01:05:38 +00001997<div class="doc_text">
Chris Lattner2cbdc452005-11-06 08:02:57 +00001998
Chris Lattner00950542001-06-06 20:29:01 +00001999<h5>Syntax:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002000
2001<pre>
2002 &lt;result&gt; = alloca &lt;type&gt;[, uint &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00002003</pre>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002004
Chris Lattner00950542001-06-06 20:29:01 +00002005<h5>Overview:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002006
Chris Lattner261efe92003-11-25 01:02:51 +00002007<p>The '<tt>alloca</tt>' instruction allocates memory on the current
2008stack frame of the procedure that is live until the current function
2009returns to its caller.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002010
Chris Lattner00950542001-06-06 20:29:01 +00002011<h5>Arguments:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002012
John Criswell9e2485c2004-12-10 15:51:16 +00002013<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
Chris Lattner261efe92003-11-25 01:02:51 +00002014bytes of memory on the runtime stack, returning a pointer of the
Chris Lattner2cbdc452005-11-06 08:02:57 +00002015appropriate type to the program. If "NumElements" is specified, it is the
2016number of elements allocated. If an alignment is specified, the value result
2017of the allocation is guaranteed to be aligned to at least that boundary. If
2018not specified, or if zero, the target can choose to align the allocation on any
2019convenient boundary.</p>
2020
Misha Brukman9d0919f2003-11-08 01:05:38 +00002021<p>'<tt>type</tt>' may be any sized type.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002022
Chris Lattner00950542001-06-06 20:29:01 +00002023<h5>Semantics:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002024
John Criswellc1f786c2005-05-13 22:25:59 +00002025<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
Chris Lattner261efe92003-11-25 01:02:51 +00002026memory is automatically released when the function returns. The '<tt>alloca</tt>'
2027instruction is commonly used to represent automatic variables that must
2028have an address available. When the function returns (either with the <tt><a
John Criswelldae2e932005-05-12 16:55:34 +00002029 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002030instructions), the memory is reclaimed.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002031
Chris Lattner00950542001-06-06 20:29:01 +00002032<h5>Example:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002033
2034<pre>
2035 %ptr = alloca int <i>; yields {int*}:ptr</i>
Chris Lattner7faa8832002-04-14 06:13:44 +00002036 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002037 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
2038 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
Chris Lattner00950542001-06-06 20:29:01 +00002039</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002040</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00002041
Chris Lattner00950542001-06-06 20:29:01 +00002042<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00002043<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2044Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002045<div class="doc_text">
Chris Lattner2b7d3202002-05-06 03:03:22 +00002046<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002047<pre> &lt;result&gt; = load &lt;ty&gt;* &lt;pointer&gt;<br> &lt;result&gt; = volatile load &lt;ty&gt;* &lt;pointer&gt;<br></pre>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002048<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002049<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002050<h5>Arguments:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002051<p>The argument to the '<tt>load</tt>' instruction specifies the memory
John Criswell0ec250c2005-10-24 16:17:18 +00002052address from which to load. The pointer must point to a <a
Chris Lattnere53e5082004-06-03 22:57:15 +00002053 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
John Criswell0ec250c2005-10-24 16:17:18 +00002054marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
Chris Lattner261efe92003-11-25 01:02:51 +00002055the number or order of execution of this <tt>load</tt> with other
2056volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2057instructions. </p>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002058<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002059<p>The location of memory pointed to is loaded.</p>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002060<h5>Examples:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002061<pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2062 <a
2063 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002064 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2065</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002066</div>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002067<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00002068<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2069Instruction</a> </div>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002070<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002071<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt; <i>; yields {void}</i>
Chris Lattnerf0651072003-09-08 18:27:49 +00002072 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt; <i>; yields {void}</i>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002073</pre>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002074<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002075<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002076<h5>Arguments:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002077<p>There are two arguments to the '<tt>store</tt>' instruction: a value
John Criswell0ec250c2005-10-24 16:17:18 +00002078to store and an address in which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
Chris Lattner261efe92003-11-25 01:02:51 +00002079operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
John Criswellc1f786c2005-05-13 22:25:59 +00002080operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
Chris Lattner261efe92003-11-25 01:02:51 +00002081optimizer is not allowed to modify the number or order of execution of
2082this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2083 href="#i_store">store</a></tt> instructions.</p>
2084<h5>Semantics:</h5>
2085<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2086at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002087<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002088<pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
2089 <a
2090 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002091 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2092</pre>
Chris Lattner2b7d3202002-05-06 03:03:22 +00002093<!-- _______________________________________________________________________ -->
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002094<div class="doc_subsubsection">
2095 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2096</div>
2097
Misha Brukman9d0919f2003-11-08 01:05:38 +00002098<div class="doc_text">
Chris Lattner7faa8832002-04-14 06:13:44 +00002099<h5>Syntax:</h5>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002100<pre>
2101 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2102</pre>
2103
Chris Lattner7faa8832002-04-14 06:13:44 +00002104<h5>Overview:</h5>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002105
2106<p>
2107The '<tt>getelementptr</tt>' instruction is used to get the address of a
2108subelement of an aggregate data structure.</p>
2109
Chris Lattner7faa8832002-04-14 06:13:44 +00002110<h5>Arguments:</h5>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002111
2112<p>This instruction takes a list of integer constants that indicate what
2113elements of the aggregate object to index to. The actual types of the arguments
2114provided depend on the type of the first pointer argument. The
2115'<tt>getelementptr</tt>' instruction is used to index down through the type
John Criswellfc6b8952005-05-16 16:17:45 +00002116levels of a structure or to a specific index in an array. When indexing into a
2117structure, only <tt>uint</tt>
John Criswellc1f786c2005-05-13 22:25:59 +00002118integer constants are allowed. When indexing into an array or pointer,
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002119<tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
2120
Chris Lattner261efe92003-11-25 01:02:51 +00002121<p>For example, let's consider a C code fragment and how it gets
2122compiled to LLVM:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002123
2124<pre>
2125 struct RT {
2126 char A;
2127 int B[10][20];
2128 char C;
2129 };
2130 struct ST {
2131 int X;
2132 double Y;
2133 struct RT Z;
2134 };
2135
2136 int *foo(struct ST *s) {
2137 return &amp;s[1].Z.B[5][13];
2138 }
2139</pre>
2140
Misha Brukman9d0919f2003-11-08 01:05:38 +00002141<p>The LLVM code generated by the GCC frontend is:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002142
2143<pre>
2144 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2145 %ST = type { int, double, %RT }
2146
Brian Gaeke7283e7c2004-07-02 21:08:14 +00002147 implementation
2148
2149 int* %foo(%ST* %s) {
2150 entry:
2151 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002152 ret int* %reg
2153 }
2154</pre>
2155
Chris Lattner7faa8832002-04-14 06:13:44 +00002156<h5>Semantics:</h5>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002157
2158<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
John Criswellc1f786c2005-05-13 22:25:59 +00002159on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
Chris Lattnere53e5082004-06-03 22:57:15 +00002160and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
2161<tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002162types require <tt>uint</tt> <b>constants</b>.</p>
2163
Misha Brukman9d0919f2003-11-08 01:05:38 +00002164<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002165type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2166}</tt>' type, a structure. The second index indexes into the third element of
2167the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2168sbyte }</tt>' type, another structure. The third index indexes into the second
2169element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2170array. The two dimensions of the array are subscripted into, yielding an
John Criswellfc6b8952005-05-16 16:17:45 +00002171'<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002172to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2173
Chris Lattner261efe92003-11-25 01:02:51 +00002174<p>Note that it is perfectly legal to index partially through a
2175structure, returning a pointer to an inner element. Because of this,
2176the LLVM code for the given testcase is equivalent to:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002177
2178<pre>
Chris Lattnerd4f6b172005-03-07 22:13:59 +00002179 int* %foo(%ST* %s) {
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002180 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2181 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2182 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2183 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2184 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2185 ret int* %t5
2186 }
Chris Lattner6536cfe2002-05-06 22:08:29 +00002187</pre>
Chris Lattnere67a9512005-06-24 17:22:57 +00002188
2189<p>Note that it is undefined to access an array out of bounds: array and
2190pointer indexes must always be within the defined bounds of the array type.
2191The one exception for this rules is zero length arrays. These arrays are
2192defined to be accessible as variable length arrays, which requires access
2193beyond the zero'th element.</p>
2194
Chris Lattner7faa8832002-04-14 06:13:44 +00002195<h5>Example:</h5>
Chris Lattnere67a9512005-06-24 17:22:57 +00002196
Chris Lattnerf74d5c72004-04-05 01:30:49 +00002197<pre>
2198 <i>; yields [12 x ubyte]*:aptr</i>
2199 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2200</pre>
2201
2202</div>
Chris Lattner00950542001-06-06 20:29:01 +00002203<!-- ======================================================================= -->
Chris Lattner261efe92003-11-25 01:02:51 +00002204<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002205<div class="doc_text">
John Criswell4457dc92004-04-09 16:48:45 +00002206<p>The instructions in this category are the "miscellaneous"
Chris Lattner261efe92003-11-25 01:02:51 +00002207instructions, which defy better classification.</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002208</div>
Chris Lattner00950542001-06-06 20:29:01 +00002209<!-- _______________________________________________________________________ -->
Chris Lattner261efe92003-11-25 01:02:51 +00002210<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
2211Instruction</a> </div>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002212<div class="doc_text">
Chris Lattner33ba0d92001-07-09 00:26:23 +00002213<h5>Syntax:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002214<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
Chris Lattner33ba0d92001-07-09 00:26:23 +00002215<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002216<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
2217the SSA graph representing the function.</p>
Chris Lattner33ba0d92001-07-09 00:26:23 +00002218<h5>Arguments:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002219<p>The type of the incoming values are specified with the first type
2220field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
2221as arguments, with one pair for each predecessor basic block of the
2222current block. Only values of <a href="#t_firstclass">first class</a>
2223type may be used as the value arguments to the PHI node. Only labels
2224may be used as the label arguments.</p>
2225<p>There must be no non-phi instructions between the start of a basic
2226block and the PHI instructions: i.e. PHI instructions must be first in
2227a basic block.</p>
Chris Lattner33ba0d92001-07-09 00:26:23 +00002228<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002229<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
2230value specified by the parameter, depending on which basic block we
2231came from in the last <a href="#terminators">terminator</a> instruction.</p>
Chris Lattner6536cfe2002-05-06 22:08:29 +00002232<h5>Example:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002233<pre>Loop: ; Infinite loop that counts from 0 on up...<br> %indvar = phi uint [ 0, %LoopHeader ], [ %nextindvar, %Loop ]<br> %nextindvar = add uint %indvar, 1<br> br label %Loop<br></pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002234</div>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002235
Chris Lattner6536cfe2002-05-06 22:08:29 +00002236<!-- _______________________________________________________________________ -->
Chris Lattnercc37aae2004-03-12 05:50:16 +00002237<div class="doc_subsubsection">
2238 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
2239</div>
2240
Misha Brukman9d0919f2003-11-08 01:05:38 +00002241<div class="doc_text">
Chris Lattnercc37aae2004-03-12 05:50:16 +00002242
Chris Lattner6536cfe2002-05-06 22:08:29 +00002243<h5>Syntax:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002244
2245<pre>
2246 &lt;result&gt; = cast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Chris Lattner6536cfe2002-05-06 22:08:29 +00002247</pre>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002248
Chris Lattner6536cfe2002-05-06 22:08:29 +00002249<h5>Overview:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002250
2251<p>
2252The '<tt>cast</tt>' instruction is used as the primitive means to convert
2253integers to floating point, change data type sizes, and break type safety (by
2254casting pointers).
2255</p>
2256
2257
Chris Lattner6536cfe2002-05-06 22:08:29 +00002258<h5>Arguments:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002259
2260<p>
2261The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
2262class value, and a type to cast it to, which must also be a <a
2263href="#t_firstclass">first class</a> type.
2264</p>
2265
Chris Lattner6536cfe2002-05-06 22:08:29 +00002266<h5>Semantics:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002267
2268<p>
2269This instruction follows the C rules for explicit casts when determining how the
2270data being cast must change to fit in its new container.
2271</p>
2272
2273<p>
2274When casting to bool, any value that would be considered true in the context of
2275a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
2276all else are '<tt>false</tt>'.
2277</p>
2278
2279<p>
2280When extending an integral value from a type of one signness to another (for
2281example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
2282<b>source</b> value is signed, and zero-extended if the source value is
2283unsigned. <tt>bool</tt> values are always zero extended into either zero or
2284one.
2285</p>
2286
Chris Lattner33ba0d92001-07-09 00:26:23 +00002287<h5>Example:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002288
2289<pre>
2290 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
Chris Lattner7bae3952002-06-25 18:03:17 +00002291 %Y = cast int 123 to bool <i>; yields bool:true</i>
Chris Lattner33ba0d92001-07-09 00:26:23 +00002292</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002293</div>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002294
2295<!-- _______________________________________________________________________ -->
2296<div class="doc_subsubsection">
2297 <a name="i_select">'<tt>select</tt>' Instruction</a>
2298</div>
2299
2300<div class="doc_text">
2301
2302<h5>Syntax:</h5>
2303
2304<pre>
2305 &lt;result&gt; = select bool &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
2306</pre>
2307
2308<h5>Overview:</h5>
2309
2310<p>
2311The '<tt>select</tt>' instruction is used to choose one value based on a
2312condition, without branching.
2313</p>
2314
2315
2316<h5>Arguments:</h5>
2317
2318<p>
2319The '<tt>select</tt>' instruction requires a boolean value indicating the condition, and two values of the same <a href="#t_firstclass">first class</a> type.
2320</p>
2321
2322<h5>Semantics:</h5>
2323
2324<p>
2325If the boolean condition evaluates to true, the instruction returns the first
John Criswellfc6b8952005-05-16 16:17:45 +00002326value argument; otherwise, it returns the second value argument.
Chris Lattnercc37aae2004-03-12 05:50:16 +00002327</p>
2328
2329<h5>Example:</h5>
2330
2331<pre>
2332 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2333</pre>
2334</div>
2335
2336
Robert Bocchino3a558662006-01-05 17:37:02 +00002337<!-- _______________________________________________________________________ -->
2338<div class="doc_subsubsection">
2339 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2340</div>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002341
Robert Bocchino3a558662006-01-05 17:37:02 +00002342<div class="doc_text">
2343
2344<h5>Syntax:</h5>
2345
2346<pre>
2347 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, uint &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2348</pre>
2349
2350<h5>Overview:</h5>
2351
2352<p>
2353The '<tt>extractelement</tt>' instruction extracts a single scalar
Robert Bocchino05ccd702006-01-15 20:48:27 +00002354element from a packed vector at a specified index.
Robert Bocchino3a558662006-01-05 17:37:02 +00002355</p>
2356
2357
2358<h5>Arguments:</h5>
2359
2360<p>
2361The first operand of an '<tt>extractelement</tt>' instruction is a
2362value of <a href="#t_packed">packed</a> type. The second operand is
2363an index indicating the position from which to extract the element.
2364The index may be a variable.</p>
2365
2366<h5>Semantics:</h5>
2367
2368<p>
2369The result is a scalar of the same type as the element type of
2370<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2371<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2372results are undefined.
2373</p>
2374
2375<h5>Example:</h5>
2376
2377<pre>
2378 %result = extractelement &lt;4 x int&gt; %vec, uint 0 <i>; yields int</i>
2379</pre>
2380</div>
Chris Lattnercc37aae2004-03-12 05:50:16 +00002381
2382
Chris Lattner33ba0d92001-07-09 00:26:23 +00002383<!-- _______________________________________________________________________ -->
Chris Lattner2bff5242005-05-06 05:47:36 +00002384<div class="doc_subsubsection">
Robert Bocchino05ccd702006-01-15 20:48:27 +00002385 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2386</div>
2387
2388<div class="doc_text">
2389
2390<h5>Syntax:</h5>
2391
2392<pre>
2393 &lt;result&gt; = insertelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, &lt;ty&gt; &lt;elt&gt, uint &lt;idx&gt; <i>; yields &lt;n x &lt;ty&gt;&gt;</i>
2394</pre>
2395
2396<h5>Overview:</h5>
2397
2398<p>
2399The '<tt>insertelement</tt>' instruction inserts a scalar
2400element into a packed vector at a specified index.
2401</p>
2402
2403
2404<h5>Arguments:</h5>
2405
2406<p>
2407The first operand of an '<tt>insertelement</tt>' instruction is a
2408value of <a href="#t_packed">packed</a> type. The second operand is a
2409scalar value whose type must equal the element type of the first
2410operand. The third operand is an index indicating the position at
2411which to insert the value. The index may be a variable.</p>
2412
2413<h5>Semantics:</h5>
2414
2415<p>
2416The result is a packed vector of the same type as <tt>val</tt>. Its
2417element values are those of <tt>val</tt> except at position
2418<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2419exceeds the length of <tt>val</tt>, the results are undefined.
2420</p>
2421
2422<h5>Example:</h5>
2423
2424<pre>
2425 %result = insertelement &lt;4 x int&gt; %vec, int 1, uint 0 <i>; yields &lt;4 x int&gt;</i>
2426</pre>
2427</div>
2428
2429
2430<!-- _______________________________________________________________________ -->
2431<div class="doc_subsubsection">
Chris Lattner2bff5242005-05-06 05:47:36 +00002432 <a name="i_call">'<tt>call</tt>' Instruction</a>
2433</div>
2434
Misha Brukman9d0919f2003-11-08 01:05:38 +00002435<div class="doc_text">
Chris Lattner2bff5242005-05-06 05:47:36 +00002436
Chris Lattner00950542001-06-06 20:29:01 +00002437<h5>Syntax:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00002438<pre>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00002439 &lt;result&gt; = [tail] call [<a href="#callingconv">cconv</a>] &lt;ty&gt;* &lt;fnptrval&gt;(&lt;param list&gt;)
Chris Lattner2bff5242005-05-06 05:47:36 +00002440</pre>
2441
Chris Lattner00950542001-06-06 20:29:01 +00002442<h5>Overview:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00002443
Misha Brukman9d0919f2003-11-08 01:05:38 +00002444<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00002445
Chris Lattner00950542001-06-06 20:29:01 +00002446<h5>Arguments:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00002447
Misha Brukman9d0919f2003-11-08 01:05:38 +00002448<p>This instruction requires several arguments:</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00002449
Chris Lattner6536cfe2002-05-06 22:08:29 +00002450<ol>
Chris Lattner261efe92003-11-25 01:02:51 +00002451 <li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00002452 <p>The optional "tail" marker indicates whether the callee function accesses
2453 any allocas or varargs in the caller. If the "tail" marker is present, the
Chris Lattner2bff5242005-05-06 05:47:36 +00002454 function call is eligible for tail call optimization. Note that calls may
2455 be marked "tail" even if they do not occur before a <a
2456 href="#i_ret"><tt>ret</tt></a> instruction.
Chris Lattner261efe92003-11-25 01:02:51 +00002457 </li>
2458 <li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00002459 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
2460 convention</a> the call should use. If none is specified, the call defaults
2461 to using C calling conventions.
2462 </li>
2463 <li>
Chris Lattner2bff5242005-05-06 05:47:36 +00002464 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
2465 being invoked. The argument types must match the types implied by this
John Criswellfc6b8952005-05-16 16:17:45 +00002466 signature. This type can be omitted if the function is not varargs and
2467 if the function type does not return a pointer to a function.</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00002468 </li>
2469 <li>
2470 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
2471 be invoked. In most cases, this is a direct function invocation, but
2472 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
John Criswellfc6b8952005-05-16 16:17:45 +00002473 to function value.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00002474 </li>
2475 <li>
2476 <p>'<tt>function args</tt>': argument list whose types match the
Reid Spencera7e302a2005-05-01 22:22:57 +00002477 function signature argument types. All arguments must be of
2478 <a href="#t_firstclass">first class</a> type. If the function signature
2479 indicates the function accepts a variable number of arguments, the extra
2480 arguments can be specified.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00002481 </li>
Chris Lattner6536cfe2002-05-06 22:08:29 +00002482</ol>
Chris Lattner2bff5242005-05-06 05:47:36 +00002483
Chris Lattner00950542001-06-06 20:29:01 +00002484<h5>Semantics:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00002485
Chris Lattner261efe92003-11-25 01:02:51 +00002486<p>The '<tt>call</tt>' instruction is used to cause control flow to
2487transfer to a specified function, with its incoming arguments bound to
2488the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2489instruction in the called function, control flow continues with the
2490instruction after the function call, and the return value of the
2491function is bound to the result argument. This is a simpler case of
2492the <a href="#i_invoke">invoke</a> instruction.</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00002493
Chris Lattner00950542001-06-06 20:29:01 +00002494<h5>Example:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00002495
2496<pre>
2497 %retval = call int %test(int %argc)
2498 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
2499 %X = tail call int %foo()
Chris Lattnerbad10ee2005-05-06 22:57:40 +00002500 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
Chris Lattner2bff5242005-05-06 05:47:36 +00002501</pre>
2502
Misha Brukman9d0919f2003-11-08 01:05:38 +00002503</div>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002504
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002505<!-- _______________________________________________________________________ -->
Chris Lattnere19d7a72004-09-27 21:51:25 +00002506<div class="doc_subsubsection">
Chris Lattnerfb6977d2006-01-13 23:26:01 +00002507 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002508</div>
2509
Misha Brukman9d0919f2003-11-08 01:05:38 +00002510<div class="doc_text">
Chris Lattnere19d7a72004-09-27 21:51:25 +00002511
Chris Lattner8d1a81d2003-10-18 05:51:36 +00002512<h5>Syntax:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002513
2514<pre>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002515 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
Chris Lattnere19d7a72004-09-27 21:51:25 +00002516</pre>
2517
Chris Lattner8d1a81d2003-10-18 05:51:36 +00002518<h5>Overview:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002519
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002520<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
Chris Lattnere19d7a72004-09-27 21:51:25 +00002521the "variable argument" area of a function call. It is used to implement the
2522<tt>va_arg</tt> macro in C.</p>
2523
Chris Lattner8d1a81d2003-10-18 05:51:36 +00002524<h5>Arguments:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002525
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002526<p>This instruction takes a <tt>va_list*</tt> value and the type of
2527the argument. It returns a value of the specified argument type and
Jeff Cohen25d4f7e2005-11-11 02:15:27 +00002528increments the <tt>va_list</tt> to point to the next argument. Again, the
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002529actual type of <tt>va_list</tt> is target specific.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002530
Chris Lattner8d1a81d2003-10-18 05:51:36 +00002531<h5>Semantics:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002532
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002533<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
2534type from the specified <tt>va_list</tt> and causes the
2535<tt>va_list</tt> to point to the next argument. For more information,
2536see the variable argument handling <a href="#int_varargs">Intrinsic
2537Functions</a>.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002538
2539<p>It is legal for this instruction to be called in a function which does not
2540take a variable number of arguments, for example, the <tt>vfprintf</tt>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002541function.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002542
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002543<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
John Criswellfc6b8952005-05-16 16:17:45 +00002544href="#intrinsics">intrinsic function</a> because it takes a type as an
Chris Lattnere19d7a72004-09-27 21:51:25 +00002545argument.</p>
2546
Chris Lattner8d1a81d2003-10-18 05:51:36 +00002547<h5>Example:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00002548
2549<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2550
Misha Brukman9d0919f2003-11-08 01:05:38 +00002551</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00002552
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002553<!-- *********************************************************************** -->
Chris Lattner261efe92003-11-25 01:02:51 +00002554<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2555<!-- *********************************************************************** -->
Chris Lattner8ff75902004-01-06 05:31:32 +00002556
Misha Brukman9d0919f2003-11-08 01:05:38 +00002557<div class="doc_text">
Chris Lattner33aec9e2004-02-12 17:01:32 +00002558
2559<p>LLVM supports the notion of an "intrinsic function". These functions have
John Criswellfc6b8952005-05-16 16:17:45 +00002560well known names and semantics and are required to follow certain
Chris Lattner33aec9e2004-02-12 17:01:32 +00002561restrictions. Overall, these instructions represent an extension mechanism for
2562the LLVM language that does not require changing all of the transformations in
2563LLVM to add to the language (or the bytecode reader/writer, the parser,
2564etc...).</p>
2565
John Criswellfc6b8952005-05-16 16:17:45 +00002566<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
2567prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
Chris Lattner33aec9e2004-02-12 17:01:32 +00002568this. Intrinsic functions must always be external functions: you cannot define
2569the body of intrinsic functions. Intrinsic functions may only be used in call
2570or invoke instructions: it is illegal to take the address of an intrinsic
2571function. Additionally, because intrinsic functions are part of the LLVM
2572language, it is required that they all be documented here if any are added.</p>
2573
2574
John Criswellfc6b8952005-05-16 16:17:45 +00002575<p>To learn how to add an intrinsic function, please see the <a
Chris Lattner590cff32005-05-11 03:35:57 +00002576href="ExtendingLLVM.html">Extending LLVM Guide</a>.
Chris Lattner33aec9e2004-02-12 17:01:32 +00002577</p>
2578
Misha Brukman9d0919f2003-11-08 01:05:38 +00002579</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00002580
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002581<!-- ======================================================================= -->
Chris Lattner8ff75902004-01-06 05:31:32 +00002582<div class="doc_subsection">
2583 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2584</div>
2585
Misha Brukman9d0919f2003-11-08 01:05:38 +00002586<div class="doc_text">
Chris Lattnerd7923912004-05-23 21:06:01 +00002587
Misha Brukman9d0919f2003-11-08 01:05:38 +00002588<p>Variable argument support is defined in LLVM with the <a
Chris Lattnerfb6977d2006-01-13 23:26:01 +00002589 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
Chris Lattner261efe92003-11-25 01:02:51 +00002590intrinsic functions. These functions are related to the similarly
2591named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00002592
Chris Lattner261efe92003-11-25 01:02:51 +00002593<p>All of these functions operate on arguments that use a
2594target-specific value type "<tt>va_list</tt>". The LLVM assembly
2595language reference manual does not define what this type is, so all
2596transformations should be prepared to handle intrinsics with any type
2597used.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00002598
Misha Brukman9d0919f2003-11-08 01:05:38 +00002599<p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
Chris Lattner261efe92003-11-25 01:02:51 +00002600instruction and the variable argument handling intrinsic functions are
2601used.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00002602
Chris Lattner33aec9e2004-02-12 17:01:32 +00002603<pre>
2604int %test(int %X, ...) {
2605 ; Initialize variable argument processing
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002606 %ap = alloca sbyte*
2607 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
Chris Lattner33aec9e2004-02-12 17:01:32 +00002608
2609 ; Read a single integer argument
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002610 %tmp = va_arg sbyte** %ap, int
Chris Lattner33aec9e2004-02-12 17:01:32 +00002611
2612 ; Demonstrate usage of llvm.va_copy and llvm.va_end
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002613 %aq = alloca sbyte*
Andrew Lenharthd0a4c622005-06-22 20:38:11 +00002614 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002615 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
Chris Lattner33aec9e2004-02-12 17:01:32 +00002616
2617 ; Stop processing of arguments.
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002618 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
Chris Lattner33aec9e2004-02-12 17:01:32 +00002619 ret int %tmp
2620}
2621</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002622</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00002623
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002624<!-- _______________________________________________________________________ -->
Chris Lattner8ff75902004-01-06 05:31:32 +00002625<div class="doc_subsubsection">
2626 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2627</div>
2628
2629
Misha Brukman9d0919f2003-11-08 01:05:38 +00002630<div class="doc_text">
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002631<h5>Syntax:</h5>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002632<pre> declare void %llvm.va_start(&lt;va_list&gt;* &lt;arglist&gt;)<br></pre>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002633<h5>Overview:</h5>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002634<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
2635<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
2636href="#i_va_arg">va_arg</a></tt>.</p>
2637
2638<h5>Arguments:</h5>
2639
2640<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
2641
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002642<h5>Semantics:</h5>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002643
2644<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2645macro available in C. In a target-dependent way, it initializes the
2646<tt>va_list</tt> element the argument points to, so that the next call to
2647<tt>va_arg</tt> will produce the first variable argument passed to the function.
2648Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2649last argument of the function, the compiler can figure that out.</p>
2650
Misha Brukman9d0919f2003-11-08 01:05:38 +00002651</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00002652
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002653<!-- _______________________________________________________________________ -->
Chris Lattner8ff75902004-01-06 05:31:32 +00002654<div class="doc_subsubsection">
2655 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2656</div>
2657
Misha Brukman9d0919f2003-11-08 01:05:38 +00002658<div class="doc_text">
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002659<h5>Syntax:</h5>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002660<pre> declare void %llvm.va_end(&lt;va_list*&gt; &lt;arglist&gt;)<br></pre>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002661<h5>Overview:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00002662<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>&lt;arglist&gt;</tt>
2663which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2664or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002665<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002666<p>The argument is a <tt>va_list</tt> to destroy.</p>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002667<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002668<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
Chris Lattner261efe92003-11-25 01:02:51 +00002669macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2670Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2671 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2672with calls to <tt>llvm.va_end</tt>.</p>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002673</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00002674
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002675<!-- _______________________________________________________________________ -->
Chris Lattner8ff75902004-01-06 05:31:32 +00002676<div class="doc_subsubsection">
2677 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2678</div>
2679
Misha Brukman9d0919f2003-11-08 01:05:38 +00002680<div class="doc_text">
Chris Lattnerd7923912004-05-23 21:06:01 +00002681
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002682<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00002683
2684<pre>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002685 declare void %llvm.va_copy(&lt;va_list&gt;* &lt;destarglist&gt;,
Andrew Lenharthd0a4c622005-06-22 20:38:11 +00002686 &lt;va_list&gt;* &lt;srcarglist&gt;)
Chris Lattnerd7923912004-05-23 21:06:01 +00002687</pre>
2688
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002689<h5>Overview:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00002690
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002691<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
2692the source argument list to the destination argument list.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00002693
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002694<h5>Arguments:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00002695
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002696<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
Andrew Lenharthd0a4c622005-06-22 20:38:11 +00002697The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002698
Chris Lattnerd7923912004-05-23 21:06:01 +00002699
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00002700<h5>Semantics:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00002701
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00002702<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
2703available in C. In a target-dependent way, it copies the source
2704<tt>va_list</tt> element into the destination list. This intrinsic is necessary
2705because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
Chris Lattnerd7923912004-05-23 21:06:01 +00002706arbitrarily complex and require memory allocation, for example.</p>
2707
Misha Brukman9d0919f2003-11-08 01:05:38 +00002708</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00002709
Chris Lattner33aec9e2004-02-12 17:01:32 +00002710<!-- ======================================================================= -->
2711<div class="doc_subsection">
Chris Lattnerd7923912004-05-23 21:06:01 +00002712 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2713</div>
2714
2715<div class="doc_text">
2716
2717<p>
2718LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2719Collection</a> requires the implementation and generation of these intrinsics.
2720These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2721stack</a>, as well as garbage collector implementations that require <a
2722href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2723Front-ends for type-safe garbage collected languages should generate these
2724intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2725href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2726</p>
2727</div>
2728
2729<!-- _______________________________________________________________________ -->
2730<div class="doc_subsubsection">
2731 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2732</div>
2733
2734<div class="doc_text">
2735
2736<h5>Syntax:</h5>
2737
2738<pre>
Reid Spencera8d451e2005-04-26 20:50:44 +00002739 declare void %llvm.gcroot(&lt;ty&gt;** %ptrloc, &lt;ty2&gt;* %metadata)
Chris Lattnerd7923912004-05-23 21:06:01 +00002740</pre>
2741
2742<h5>Overview:</h5>
2743
John Criswell9e2485c2004-12-10 15:51:16 +00002744<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
Chris Lattnerd7923912004-05-23 21:06:01 +00002745the code generator, and allows some metadata to be associated with it.</p>
2746
2747<h5>Arguments:</h5>
2748
2749<p>The first argument specifies the address of a stack object that contains the
2750root pointer. The second pointer (which must be either a constant or a global
2751value address) contains the meta-data to be associated with the root.</p>
2752
2753<h5>Semantics:</h5>
2754
2755<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2756location. At compile-time, the code generator generates information to allow
2757the runtime to find the pointer at GC safe points.
2758</p>
2759
2760</div>
2761
2762
2763<!-- _______________________________________________________________________ -->
2764<div class="doc_subsubsection">
2765 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2766</div>
2767
2768<div class="doc_text">
2769
2770<h5>Syntax:</h5>
2771
2772<pre>
Reid Spencera8d451e2005-04-26 20:50:44 +00002773 declare sbyte* %llvm.gcread(sbyte** %Ptr)
Chris Lattnerd7923912004-05-23 21:06:01 +00002774</pre>
2775
2776<h5>Overview:</h5>
2777
2778<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2779locations, allowing garbage collector implementations that require read
2780barriers.</p>
2781
2782<h5>Arguments:</h5>
2783
2784<p>The argument is the address to read from, which should be an address
2785allocated from the garbage collector.</p>
2786
2787<h5>Semantics:</h5>
2788
2789<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2790instruction, but may be replaced with substantially more complex code by the
2791garbage collector runtime, as needed.</p>
2792
2793</div>
2794
2795
2796<!-- _______________________________________________________________________ -->
2797<div class="doc_subsubsection">
2798 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2799</div>
2800
2801<div class="doc_text">
2802
2803<h5>Syntax:</h5>
2804
2805<pre>
Reid Spencera8d451e2005-04-26 20:50:44 +00002806 declare void %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
Chris Lattnerd7923912004-05-23 21:06:01 +00002807</pre>
2808
2809<h5>Overview:</h5>
2810
2811<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2812locations, allowing garbage collector implementations that require write
2813barriers (such as generational or reference counting collectors).</p>
2814
2815<h5>Arguments:</h5>
2816
2817<p>The first argument is the reference to store, and the second is the heap
2818location to store to.</p>
2819
2820<h5>Semantics:</h5>
2821
2822<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2823instruction, but may be replaced with substantially more complex code by the
2824garbage collector runtime, as needed.</p>
2825
2826</div>
2827
2828
2829
2830<!-- ======================================================================= -->
2831<div class="doc_subsection">
Chris Lattner10610642004-02-14 04:08:35 +00002832 <a name="int_codegen">Code Generator Intrinsics</a>
2833</div>
2834
2835<div class="doc_text">
2836<p>
2837These intrinsics are provided by LLVM to expose special features that may only
2838be implemented with code generator support.
2839</p>
2840
2841</div>
2842
2843<!-- _______________________________________________________________________ -->
2844<div class="doc_subsubsection">
2845 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2846</div>
2847
2848<div class="doc_text">
2849
2850<h5>Syntax:</h5>
2851<pre>
Chris Lattnerfcf39d42006-01-13 01:20:27 +00002852 declare sbyte *%llvm.returnaddress(uint &lt;level&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00002853</pre>
2854
2855<h5>Overview:</h5>
2856
2857<p>
2858The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2859indicating the return address of the current function or one of its callers.
2860</p>
2861
2862<h5>Arguments:</h5>
2863
2864<p>
2865The argument to this intrinsic indicates which function to return the address
2866for. Zero indicates the calling function, one indicates its caller, etc. The
2867argument is <b>required</b> to be a constant integer value.
2868</p>
2869
2870<h5>Semantics:</h5>
2871
2872<p>
2873The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2874the return address of the specified call frame, or zero if it cannot be
2875identified. The value returned by this intrinsic is likely to be incorrect or 0
2876for arguments other than zero, so it should only be used for debugging purposes.
2877</p>
2878
2879<p>
2880Note that calling this intrinsic does not prevent function inlining or other
Chris Lattnerb40bb382005-03-07 20:30:51 +00002881aggressive transformations, so the value returned may not be that of the obvious
Chris Lattner10610642004-02-14 04:08:35 +00002882source-language caller.
2883</p>
2884</div>
2885
2886
2887<!-- _______________________________________________________________________ -->
2888<div class="doc_subsubsection">
2889 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2890</div>
2891
2892<div class="doc_text">
2893
2894<h5>Syntax:</h5>
2895<pre>
Chris Lattnerfcf39d42006-01-13 01:20:27 +00002896 declare sbyte *%llvm.frameaddress(uint &lt;level&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00002897</pre>
2898
2899<h5>Overview:</h5>
2900
2901<p>
2902The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2903pointer value for the specified stack frame.
2904</p>
2905
2906<h5>Arguments:</h5>
2907
2908<p>
2909The argument to this intrinsic indicates which function to return the frame
2910pointer for. Zero indicates the calling function, one indicates its caller,
2911etc. The argument is <b>required</b> to be a constant integer value.
2912</p>
2913
2914<h5>Semantics:</h5>
2915
2916<p>
2917The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2918the frame address of the specified call frame, or zero if it cannot be
2919identified. The value returned by this intrinsic is likely to be incorrect or 0
2920for arguments other than zero, so it should only be used for debugging purposes.
2921</p>
2922
2923<p>
2924Note that calling this intrinsic does not prevent function inlining or other
Chris Lattnerb40bb382005-03-07 20:30:51 +00002925aggressive transformations, so the value returned may not be that of the obvious
Chris Lattner10610642004-02-14 04:08:35 +00002926source-language caller.
2927</p>
2928</div>
2929
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00002930<!-- _______________________________________________________________________ -->
2931<div class="doc_subsubsection">
Chris Lattner57e1f392006-01-13 02:03:13 +00002932 <a name="i_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
2933</div>
2934
2935<div class="doc_text">
2936
2937<h5>Syntax:</h5>
2938<pre>
2939 declare sbyte *%llvm.stacksave()
2940</pre>
2941
2942<h5>Overview:</h5>
2943
2944<p>
2945The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
2946the function stack, for use with <a href="#i_stackrestore">
2947<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
2948features like scoped automatic variable sized arrays in C99.
2949</p>
2950
2951<h5>Semantics:</h5>
2952
2953<p>
2954This intrinsic returns a opaque pointer value that can be passed to <a
2955href="#i_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
2956<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
2957<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
2958state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
2959practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
2960that were allocated after the <tt>llvm.stacksave</tt> was executed.
2961</p>
2962
2963</div>
2964
2965<!-- _______________________________________________________________________ -->
2966<div class="doc_subsubsection">
2967 <a name="i_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
2968</div>
2969
2970<div class="doc_text">
2971
2972<h5>Syntax:</h5>
2973<pre>
2974 declare void %llvm.stackrestore(sbyte* %ptr)
2975</pre>
2976
2977<h5>Overview:</h5>
2978
2979<p>
2980The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
2981the function stack to the state it was in when the corresponding <a
2982href="#llvm.stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
2983useful for implementing language features like scoped automatic variable sized
2984arrays in C99.
2985</p>
2986
2987<h5>Semantics:</h5>
2988
2989<p>
2990See the description for <a href="#i_stacksave"><tt>llvm.stacksave</tt></a>.
2991</p>
2992
2993</div>
2994
2995
2996<!-- _______________________________________________________________________ -->
2997<div class="doc_subsubsection">
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00002998 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
2999</div>
3000
3001<div class="doc_text">
3002
3003<h5>Syntax:</h5>
3004<pre>
Reid Spencera8d451e2005-04-26 20:50:44 +00003005 declare void %llvm.prefetch(sbyte * &lt;address&gt;,
3006 uint &lt;rw&gt;, uint &lt;locality&gt;)
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00003007</pre>
3008
3009<h5>Overview:</h5>
3010
3011
3012<p>
3013The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
John Criswellfc6b8952005-05-16 16:17:45 +00003014a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
3015no
3016effect on the behavior of the program but can change its performance
Chris Lattner2a615362005-02-28 19:47:14 +00003017characteristics.
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00003018</p>
3019
3020<h5>Arguments:</h5>
3021
3022<p>
3023<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
3024determining if the fetch should be for a read (0) or write (1), and
3025<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
Chris Lattneraeffb4a2005-03-07 20:31:38 +00003026locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00003027<tt>locality</tt> arguments must be constant integers.
3028</p>
3029
3030<h5>Semantics:</h5>
3031
3032<p>
3033This intrinsic does not modify the behavior of the program. In particular,
3034prefetches cannot trap and do not produce a value. On targets that support this
3035intrinsic, the prefetch can provide hints to the processor cache for better
3036performance.
3037</p>
3038
3039</div>
3040
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00003041<!-- _______________________________________________________________________ -->
3042<div class="doc_subsubsection">
3043 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
3044</div>
3045
3046<div class="doc_text">
3047
3048<h5>Syntax:</h5>
3049<pre>
Reid Spencera8d451e2005-04-26 20:50:44 +00003050 declare void %llvm.pcmarker( uint &lt;id&gt; )
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00003051</pre>
3052
3053<h5>Overview:</h5>
3054
3055
3056<p>
John Criswellfc6b8952005-05-16 16:17:45 +00003057The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
3058(PC) in a region of
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00003059code to simulators and other tools. The method is target specific, but it is
3060expected that the marker will use exported symbols to transmit the PC of the marker.
Jeff Cohen25d4f7e2005-11-11 02:15:27 +00003061The marker makes no guarantees that it will remain with any specific instruction
Chris Lattnerd07c3f42005-11-15 06:07:55 +00003062after optimizations. It is possible that the presence of a marker will inhibit
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00003063optimizations. The intended use is to be inserted after optmizations to allow
John Criswellfc6b8952005-05-16 16:17:45 +00003064correlations of simulation runs.
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00003065</p>
3066
3067<h5>Arguments:</h5>
3068
3069<p>
3070<tt>id</tt> is a numerical id identifying the marker.
3071</p>
3072
3073<h5>Semantics:</h5>
3074
3075<p>
3076This intrinsic does not modify the behavior of the program. Backends that do not
3077support this intrinisic may ignore it.
3078</p>
3079
3080</div>
3081
Andrew Lenharth51b8d542005-11-11 16:47:30 +00003082<!-- _______________________________________________________________________ -->
3083<div class="doc_subsubsection">
3084 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
3085</div>
3086
3087<div class="doc_text">
3088
3089<h5>Syntax:</h5>
3090<pre>
3091 declare ulong %llvm.readcyclecounter( )
3092</pre>
3093
3094<h5>Overview:</h5>
3095
3096
3097<p>
3098The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
3099counter register (or similar low latency, high accuracy clocks) on those targets
3100that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
3101As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
3102should only be used for small timings.
3103</p>
3104
3105<h5>Semantics:</h5>
3106
3107<p>
3108When directly supported, reading the cycle counter should not modify any memory.
3109Implementations are allowed to either return a application specific value or a
3110system wide value. On backends without support, this is lowered to a constant 0.
3111</p>
3112
3113</div>
3114
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00003115
John Criswell7123e272004-04-09 16:43:20 +00003116<!-- ======================================================================= -->
3117<div class="doc_subsection">
3118 <a name="int_os">Operating System Intrinsics</a>
3119</div>
3120
3121<div class="doc_text">
3122<p>
3123These intrinsics are provided by LLVM to support the implementation of
3124operating system level code.
3125</p>
3126
3127</div>
John Criswell183402a2004-04-12 15:02:16 +00003128
John Criswellcfd3bac2004-04-09 15:23:37 +00003129<!-- _______________________________________________________________________ -->
3130<div class="doc_subsubsection">
3131 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
3132</div>
3133
3134<div class="doc_text">
3135
3136<h5>Syntax:</h5>
3137<pre>
Reid Spencera8d451e2005-04-26 20:50:44 +00003138 declare &lt;integer type&gt; %llvm.readport (&lt;integer type&gt; &lt;address&gt;)
John Criswellcfd3bac2004-04-09 15:23:37 +00003139</pre>
3140
3141<h5>Overview:</h5>
3142
3143<p>
John Criswell7123e272004-04-09 16:43:20 +00003144The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
3145I/O port.
John Criswellcfd3bac2004-04-09 15:23:37 +00003146</p>
3147
3148<h5>Arguments:</h5>
3149
3150<p>
John Criswell7123e272004-04-09 16:43:20 +00003151The argument to this intrinsic indicates the hardware I/O address from which
3152to read the data. The address is in the hardware I/O address namespace (as
3153opposed to being a memory location for memory mapped I/O).
John Criswellcfd3bac2004-04-09 15:23:37 +00003154</p>
3155
3156<h5>Semantics:</h5>
3157
3158<p>
John Criswell7123e272004-04-09 16:43:20 +00003159The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
3160specified by <i>address</i> and returns the value. The address and return
3161value must be integers, but the size is dependent upon the platform upon which
3162the program is code generated. For example, on x86, the address must be an
Misha Brukmancfa87bc2005-04-22 18:02:52 +00003163unsigned 16-bit value, and the return value must be 8, 16, or 32 bits.
John Criswellcfd3bac2004-04-09 15:23:37 +00003164</p>
3165
3166</div>
3167
3168<!-- _______________________________________________________________________ -->
3169<div class="doc_subsubsection">
3170 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
3171</div>
3172
3173<div class="doc_text">
3174
3175<h5>Syntax:</h5>
3176<pre>
Chris Lattnerc3f59762004-12-09 17:30:23 +00003177 call void (&lt;integer type&gt;, &lt;integer type&gt;)*
3178 %llvm.writeport (&lt;integer type&gt; &lt;value&gt;,
3179 &lt;integer type&gt; &lt;address&gt;)
John Criswellcfd3bac2004-04-09 15:23:37 +00003180</pre>
3181
3182<h5>Overview:</h5>
3183
3184<p>
John Criswell7123e272004-04-09 16:43:20 +00003185The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
3186I/O port.
John Criswellcfd3bac2004-04-09 15:23:37 +00003187</p>
3188
3189<h5>Arguments:</h5>
3190
3191<p>
John Criswell96db6fc2004-04-12 16:33:19 +00003192The first argument is the value to write to the I/O port.
John Criswellcfd3bac2004-04-09 15:23:37 +00003193</p>
3194
3195<p>
John Criswell96db6fc2004-04-12 16:33:19 +00003196The second argument indicates the hardware I/O address to which data should be
3197written. The address is in the hardware I/O address namespace (as opposed to
3198being a memory location for memory mapped I/O).
John Criswellcfd3bac2004-04-09 15:23:37 +00003199</p>
3200
3201<h5>Semantics:</h5>
3202
3203<p>
3204The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
3205specified by <i>address</i>. The address and value must be integers, but the
3206size is dependent upon the platform upon which the program is code generated.
Misha Brukmancfa87bc2005-04-22 18:02:52 +00003207For example, on x86, the address must be an unsigned 16-bit value, and the
John Criswell7123e272004-04-09 16:43:20 +00003208value written must be 8, 16, or 32 bits in length.
John Criswellcfd3bac2004-04-09 15:23:37 +00003209</p>
3210
3211</div>
Chris Lattner10610642004-02-14 04:08:35 +00003212
John Criswell183402a2004-04-12 15:02:16 +00003213<!-- _______________________________________________________________________ -->
3214<div class="doc_subsubsection">
3215 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
3216</div>
3217
3218<div class="doc_text">
3219
3220<h5>Syntax:</h5>
3221<pre>
Reid Spencera8d451e2005-04-26 20:50:44 +00003222 declare &lt;result&gt; %llvm.readio (&lt;ty&gt; * &lt;pointer&gt;)
John Criswell183402a2004-04-12 15:02:16 +00003223</pre>
3224
3225<h5>Overview:</h5>
3226
3227<p>
3228The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
3229address.
3230</p>
3231
3232<h5>Arguments:</h5>
3233
3234<p>
John Criswell96db6fc2004-04-12 16:33:19 +00003235The argument to this intrinsic is a pointer indicating the memory address from
3236which to read the data. The data must be a
3237<a href="#t_firstclass">first class</a> type.
John Criswell183402a2004-04-12 15:02:16 +00003238</p>
3239
3240<h5>Semantics:</h5>
3241
3242<p>
3243The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
John Criswell96db6fc2004-04-12 16:33:19 +00003244location specified by <i>pointer</i> and returns the value. The argument must
3245be a pointer, and the return value must be a
3246<a href="#t_firstclass">first class</a> type. However, certain architectures
Misha Brukmancfa87bc2005-04-22 18:02:52 +00003247may not support I/O on all first class types. For example, 32-bit processors
John Criswell96db6fc2004-04-12 16:33:19 +00003248may only support I/O on data types that are 32 bits or less.
John Criswell183402a2004-04-12 15:02:16 +00003249</p>
3250
3251<p>
John Criswell96db6fc2004-04-12 16:33:19 +00003252This intrinsic enforces an in-order memory model for llvm.readio and
3253llvm.writeio calls on machines that use dynamic scheduling. Dynamically
3254scheduled processors may execute loads and stores out of order, re-ordering at
3255run time accesses to memory mapped I/O registers. Using these intrinsics
3256ensures that accesses to memory mapped I/O registers occur in program order.
John Criswell183402a2004-04-12 15:02:16 +00003257</p>
3258
3259</div>
3260
3261<!-- _______________________________________________________________________ -->
3262<div class="doc_subsubsection">
3263 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
3264</div>
3265
3266<div class="doc_text">
3267
3268<h5>Syntax:</h5>
3269<pre>
Reid Spencera8d451e2005-04-26 20:50:44 +00003270 declare void %llvm.writeio (&lt;ty1&gt; &lt;value&gt;, &lt;ty2&gt; * &lt;pointer&gt;)
John Criswell183402a2004-04-12 15:02:16 +00003271</pre>
3272
3273<h5>Overview:</h5>
3274
3275<p>
3276The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
3277mapped I/O address.
3278</p>
3279
3280<h5>Arguments:</h5>
3281
3282<p>
John Criswell96db6fc2004-04-12 16:33:19 +00003283The first argument is the value to write to the memory mapped I/O location.
3284The second argument is a pointer indicating the memory address to which the
3285data should be written.
John Criswell183402a2004-04-12 15:02:16 +00003286</p>
3287
3288<h5>Semantics:</h5>
3289
3290<p>
3291The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
John Criswell96db6fc2004-04-12 16:33:19 +00003292I/O address specified by <i>pointer</i>. The value must be a
3293<a href="#t_firstclass">first class</a> type. However, certain architectures
Misha Brukmancfa87bc2005-04-22 18:02:52 +00003294may not support I/O on all first class types. For example, 32-bit processors
John Criswell96db6fc2004-04-12 16:33:19 +00003295may only support I/O on data types that are 32 bits or less.
John Criswell183402a2004-04-12 15:02:16 +00003296</p>
3297
3298<p>
John Criswell96db6fc2004-04-12 16:33:19 +00003299This intrinsic enforces an in-order memory model for llvm.readio and
3300llvm.writeio calls on machines that use dynamic scheduling. Dynamically
3301scheduled processors may execute loads and stores out of order, re-ordering at
3302run time accesses to memory mapped I/O registers. Using these intrinsics
3303ensures that accesses to memory mapped I/O registers occur in program order.
John Criswell183402a2004-04-12 15:02:16 +00003304</p>
3305
3306</div>
3307
Chris Lattner10610642004-02-14 04:08:35 +00003308<!-- ======================================================================= -->
3309<div class="doc_subsection">
Chris Lattner33aec9e2004-02-12 17:01:32 +00003310 <a name="int_libc">Standard C Library Intrinsics</a>
3311</div>
3312
3313<div class="doc_text">
3314<p>
Chris Lattner10610642004-02-14 04:08:35 +00003315LLVM provides intrinsics for a few important standard C library functions.
3316These intrinsics allow source-language front-ends to pass information about the
3317alignment of the pointer arguments to the code generator, providing opportunity
3318for more efficient code generation.
Chris Lattner33aec9e2004-02-12 17:01:32 +00003319</p>
3320
3321</div>
3322
3323<!-- _______________________________________________________________________ -->
3324<div class="doc_subsubsection">
3325 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3326</div>
3327
3328<div class="doc_text">
3329
3330<h5>Syntax:</h5>
3331<pre>
Reid Spencerd4622352005-04-26 20:41:16 +00003332 declare void %llvm.memcpy(sbyte* &lt;dest&gt;, sbyte* &lt;src&gt;,
3333 uint &lt;len&gt;, uint &lt;align&gt;)
Chris Lattner33aec9e2004-02-12 17:01:32 +00003334</pre>
3335
3336<h5>Overview:</h5>
3337
3338<p>
3339The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
3340location to the destination location.
3341</p>
3342
3343<p>
3344Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
3345does not return a value, and takes an extra alignment argument.
3346</p>
3347
3348<h5>Arguments:</h5>
3349
3350<p>
3351The first argument is a pointer to the destination, the second is a pointer to
3352the source. The third argument is an (arbitrarily sized) integer argument
3353specifying the number of bytes to copy, and the fourth argument is the alignment
3354of the source and destination locations.
3355</p>
3356
Chris Lattner3301ced2004-02-12 21:18:15 +00003357<p>
3358If the call to this intrinisic has an alignment value that is not 0 or 1, then
3359the caller guarantees that the size of the copy is a multiple of the alignment
3360and that both the source and destination pointers are aligned to that boundary.
3361</p>
3362
Chris Lattner33aec9e2004-02-12 17:01:32 +00003363<h5>Semantics:</h5>
3364
3365<p>
3366The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
3367location to the destination location, which are not allowed to overlap. It
3368copies "len" bytes of memory over. If the argument is known to be aligned to
3369some boundary, this can be specified as the fourth argument, otherwise it should
3370be set to 0 or 1.
3371</p>
3372</div>
3373
3374
Chris Lattner0eb51b42004-02-12 18:10:10 +00003375<!-- _______________________________________________________________________ -->
3376<div class="doc_subsubsection">
3377 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
3378</div>
3379
3380<div class="doc_text">
3381
3382<h5>Syntax:</h5>
3383<pre>
Reid Spencerd4622352005-04-26 20:41:16 +00003384 declare void %llvm.memmove(sbyte* &lt;dest&gt;, sbyte* &lt;src&gt;,
3385 uint &lt;len&gt;, uint &lt;align&gt;)
Chris Lattner0eb51b42004-02-12 18:10:10 +00003386</pre>
3387
3388<h5>Overview:</h5>
3389
3390<p>
3391The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
3392location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
3393intrinsic but allows the two memory locations to overlap.
3394</p>
3395
3396<p>
3397Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
3398does not return a value, and takes an extra alignment argument.
3399</p>
3400
3401<h5>Arguments:</h5>
3402
3403<p>
3404The first argument is a pointer to the destination, the second is a pointer to
3405the source. The third argument is an (arbitrarily sized) integer argument
3406specifying the number of bytes to copy, and the fourth argument is the alignment
3407of the source and destination locations.
3408</p>
3409
Chris Lattner3301ced2004-02-12 21:18:15 +00003410<p>
3411If the call to this intrinisic has an alignment value that is not 0 or 1, then
3412the caller guarantees that the size of the copy is a multiple of the alignment
3413and that both the source and destination pointers are aligned to that boundary.
3414</p>
3415
Chris Lattner0eb51b42004-02-12 18:10:10 +00003416<h5>Semantics:</h5>
3417
3418<p>
3419The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
3420location to the destination location, which may overlap. It
3421copies "len" bytes of memory over. If the argument is known to be aligned to
3422some boundary, this can be specified as the fourth argument, otherwise it should
3423be set to 0 or 1.
3424</p>
3425</div>
3426
Chris Lattner8ff75902004-01-06 05:31:32 +00003427
Chris Lattner10610642004-02-14 04:08:35 +00003428<!-- _______________________________________________________________________ -->
3429<div class="doc_subsubsection">
3430 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
3431</div>
3432
3433<div class="doc_text">
3434
3435<h5>Syntax:</h5>
3436<pre>
Reid Spencerd4622352005-04-26 20:41:16 +00003437 declare void %llvm.memset(sbyte* &lt;dest&gt;, ubyte &lt;val&gt;,
3438 uint &lt;len&gt;, uint &lt;align&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00003439</pre>
3440
3441<h5>Overview:</h5>
3442
3443<p>
3444The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
3445byte value.
3446</p>
3447
3448<p>
3449Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
3450does not return a value, and takes an extra alignment argument.
3451</p>
3452
3453<h5>Arguments:</h5>
3454
3455<p>
3456The first argument is a pointer to the destination to fill, the second is the
3457byte value to fill it with, the third argument is an (arbitrarily sized) integer
3458argument specifying the number of bytes to fill, and the fourth argument is the
3459known alignment of destination location.
3460</p>
3461
3462<p>
3463If the call to this intrinisic has an alignment value that is not 0 or 1, then
3464the caller guarantees that the size of the copy is a multiple of the alignment
3465and that the destination pointer is aligned to that boundary.
3466</p>
3467
3468<h5>Semantics:</h5>
3469
3470<p>
3471The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
3472destination location. If the argument is known to be aligned to some boundary,
3473this can be specified as the fourth argument, otherwise it should be set to 0 or
34741.
3475</p>
3476</div>
3477
3478
Chris Lattner32006282004-06-11 02:28:03 +00003479<!-- _______________________________________________________________________ -->
3480<div class="doc_subsubsection">
Reid Spencer0b118202006-01-16 21:12:35 +00003481 <a name="i_isunordered">'<tt>llvm.isunordered.*</tt>' Intrinsic</a>
Alkis Evlogimenos26bbe932004-06-13 01:16:15 +00003482</div>
3483
3484<div class="doc_text">
3485
3486<h5>Syntax:</h5>
3487<pre>
Reid Spencer0b118202006-01-16 21:12:35 +00003488 declare bool %llvm.isunordered.f32(float Val1, float Val2)
3489 declare bool %llvm.isunordered.f64(double Val1, double Val2)
Alkis Evlogimenos26bbe932004-06-13 01:16:15 +00003490</pre>
3491
3492<h5>Overview:</h5>
3493
3494<p>
Reid Spencer0b118202006-01-16 21:12:35 +00003495The '<tt>llvm.isunordered</tt>' intrinsics return true if either or both of the
Alkis Evlogimenos26bbe932004-06-13 01:16:15 +00003496specified floating point values is a NAN.
3497</p>
3498
3499<h5>Arguments:</h5>
3500
3501<p>
3502The arguments are floating point numbers of the same type.
3503</p>
3504
3505<h5>Semantics:</h5>
3506
3507<p>
3508If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3509false.
3510</p>
3511</div>
3512
3513
Chris Lattnera4d74142005-07-21 01:29:16 +00003514<!-- _______________________________________________________________________ -->
3515<div class="doc_subsubsection">
Chris Lattnerec6cb612006-01-16 22:38:59 +00003516 <a name="i_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
Chris Lattnera4d74142005-07-21 01:29:16 +00003517</div>
3518
3519<div class="doc_text">
3520
3521<h5>Syntax:</h5>
3522<pre>
Reid Spencer0b118202006-01-16 21:12:35 +00003523 declare double %llvm.sqrt.f32(float Val)
3524 declare double %llvm.sqrt.f64(double Val)
Chris Lattnera4d74142005-07-21 01:29:16 +00003525</pre>
3526
3527<h5>Overview:</h5>
3528
3529<p>
Reid Spencer0b118202006-01-16 21:12:35 +00003530The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Chris Lattnera4d74142005-07-21 01:29:16 +00003531returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
3532<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
3533negative numbers (which allows for better optimization).
3534</p>
3535
3536<h5>Arguments:</h5>
3537
3538<p>
3539The argument and return value are floating point numbers of the same type.
3540</p>
3541
3542<h5>Semantics:</h5>
3543
3544<p>
3545This function returns the sqrt of the specified operand if it is a positive
3546floating point number.
3547</p>
3548</div>
3549
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003550<!-- ======================================================================= -->
3551<div class="doc_subsection">
Nate Begeman7e36c472006-01-13 23:26:38 +00003552 <a name="int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003553</div>
3554
3555<div class="doc_text">
3556<p>
Nate Begeman7e36c472006-01-13 23:26:38 +00003557LLVM provides intrinsics for a few important bit manipulation operations.
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003558These allow efficient code generation for some algorithms.
3559</p>
3560
3561</div>
3562
3563<!-- _______________________________________________________________________ -->
3564<div class="doc_subsubsection">
Nate Begeman7e36c472006-01-13 23:26:38 +00003565 <a name="i_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
3566</div>
3567
3568<div class="doc_text">
3569
3570<h5>Syntax:</h5>
3571<pre>
Chris Lattnerec6cb612006-01-16 22:38:59 +00003572 declare ushort %llvm.bswap.i16(ushort &lt;id&gt;)
3573 declare uint %llvm.bswap.i32(uint &lt;id&gt;)
3574 declare ulong %llvm.bswap.i64(ulong &lt;id&gt;)
Nate Begeman7e36c472006-01-13 23:26:38 +00003575</pre>
3576
3577<h5>Overview:</h5>
3578
3579<p>
3580The '<tt>llvm.bwsap</tt>' family of intrinsics is used to byteswap a 16, 32 or
358164 bit quantity. These are useful for performing operations on data that is not
3582in the target's native byte order.
3583</p>
3584
3585<h5>Semantics:</h5>
3586
3587<p>
Chris Lattnerec6cb612006-01-16 22:38:59 +00003588The <tt>llvm.bswap.16</tt> intrinsic returns a ushort value that has the high and low
3589byte of the input ushort swapped. Similarly, the <tt>llvm.bswap.i32</tt> intrinsic
Nate Begeman7e36c472006-01-13 23:26:38 +00003590returns a uint value that has the four bytes of the input uint swapped, so that
3591if the input bytes are numbered 0, 1, 2, 3 then the returned uint will have its
Chris Lattnerec6cb612006-01-16 22:38:59 +00003592bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i64</tt> intrinsic extends this concept
Nate Begeman7e36c472006-01-13 23:26:38 +00003593to 64 bits.
3594</p>
3595
3596</div>
3597
3598<!-- _______________________________________________________________________ -->
3599<div class="doc_subsubsection">
Reid Spencer0b118202006-01-16 21:12:35 +00003600 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003601</div>
3602
3603<div class="doc_text">
3604
3605<h5>Syntax:</h5>
3606<pre>
Chris Lattnerec6cb612006-01-16 22:38:59 +00003607 declare ubyte %llvm.ctpop.i8 (ubyte &lt;src&gt;)
3608 declare ushort %llvm.ctpop.i16(ushort &lt;src&gt;)
3609 declare uint %llvm.ctpop.i32(uint &lt;src&gt;)
3610 declare ulong %llvm.ctpop.i64(ulong &lt;src&gt;)
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003611</pre>
3612
3613<h5>Overview:</h5>
3614
3615<p>
Chris Lattnerec6cb612006-01-16 22:38:59 +00003616The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
3617value.
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003618</p>
3619
3620<h5>Arguments:</h5>
3621
3622<p>
Chris Lattnercfe6b372005-05-07 01:46:40 +00003623The only argument is the value to be counted. The argument may be of any
Chris Lattnerec6cb612006-01-16 22:38:59 +00003624unsigned integer type. The return type must match the argument type.
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003625</p>
3626
3627<h5>Semantics:</h5>
3628
3629<p>
3630The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
3631</p>
3632</div>
3633
3634<!-- _______________________________________________________________________ -->
3635<div class="doc_subsubsection">
Chris Lattner8a886be2006-01-16 22:34:14 +00003636 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003637</div>
3638
3639<div class="doc_text">
3640
3641<h5>Syntax:</h5>
3642<pre>
Chris Lattnerec6cb612006-01-16 22:38:59 +00003643 declare ubyte %llvm.ctlz.i8 (ubyte &lt;src&gt;)
3644 declare ushort %llvm.ctlz.i16(ushort &lt;src&gt;)
3645 declare uint %llvm.ctlz.i32(uint &lt;src&gt;)
3646 declare ulong %llvm.ctlz.i64(ulong &lt;src&gt;)
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003647</pre>
3648
3649<h5>Overview:</h5>
3650
3651<p>
Reid Spencer0b118202006-01-16 21:12:35 +00003652The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
3653leading zeros in a variable.
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003654</p>
3655
3656<h5>Arguments:</h5>
3657
3658<p>
Chris Lattnercfe6b372005-05-07 01:46:40 +00003659The only argument is the value to be counted. The argument may be of any
Chris Lattnerec6cb612006-01-16 22:38:59 +00003660unsigned integer type. The return type must match the argument type.
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003661</p>
3662
3663<h5>Semantics:</h5>
3664
3665<p>
Chris Lattnereff29ab2005-05-15 19:39:26 +00003666The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
3667in a variable. If the src == 0 then the result is the size in bits of the type
3668of src. For example, <tt>llvm.cttz(int 2) = 30</tt>.
Andrew Lenharthec370fd2005-05-03 18:01:48 +00003669</p>
3670</div>
Chris Lattner32006282004-06-11 02:28:03 +00003671
3672
Chris Lattnereff29ab2005-05-15 19:39:26 +00003673
3674<!-- _______________________________________________________________________ -->
3675<div class="doc_subsubsection">
Chris Lattner8a886be2006-01-16 22:34:14 +00003676 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
Chris Lattnereff29ab2005-05-15 19:39:26 +00003677</div>
3678
3679<div class="doc_text">
3680
3681<h5>Syntax:</h5>
3682<pre>
Chris Lattnerec6cb612006-01-16 22:38:59 +00003683 declare ubyte %llvm.cttz.i8 (ubyte &lt;src&gt;)
3684 declare ushort %llvm.cttz.i16(ushort &lt;src&gt;)
3685 declare uint %llvm.cttz.i32(uint &lt;src&gt;)
3686 declare ulong %llvm.cttz.i64(ulong &lt;src&gt;)
Chris Lattnereff29ab2005-05-15 19:39:26 +00003687</pre>
3688
3689<h5>Overview:</h5>
3690
3691<p>
Reid Spencer0b118202006-01-16 21:12:35 +00003692The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
3693trailing zeros.
Chris Lattnereff29ab2005-05-15 19:39:26 +00003694</p>
3695
3696<h5>Arguments:</h5>
3697
3698<p>
3699The only argument is the value to be counted. The argument may be of any
Chris Lattnerec6cb612006-01-16 22:38:59 +00003700unsigned integer type. The return type must match the argument type.
Chris Lattnereff29ab2005-05-15 19:39:26 +00003701</p>
3702
3703<h5>Semantics:</h5>
3704
3705<p>
3706The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
3707in a variable. If the src == 0 then the result is the size in bits of the type
3708of src. For example, <tt>llvm.cttz(2) = 1</tt>.
3709</p>
3710</div>
3711
Chris Lattner8ff75902004-01-06 05:31:32 +00003712<!-- ======================================================================= -->
3713<div class="doc_subsection">
3714 <a name="int_debugger">Debugger Intrinsics</a>
3715</div>
3716
3717<div class="doc_text">
3718<p>
3719The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3720are described in the <a
3721href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3722Debugging</a> document.
3723</p>
3724</div>
3725
3726
Chris Lattner00950542001-06-06 20:29:01 +00003727<!-- *********************************************************************** -->
Chris Lattner00950542001-06-06 20:29:01 +00003728<hr>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00003729<address>
3730 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3731 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3732 <a href="http://validator.w3.org/check/referer"><img
3733 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3734
3735 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3736 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
3737 Last modified: $Date$
3738</address>
Misha Brukman9d0919f2003-11-08 01:05:38 +00003739</body>
3740</html>