blob: 1f9f80d6d43468b6d70a8b7294c9bf1bccb62959 [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
6 <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.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
12
13<body>
14
15<div class="doc_title"> LLVM Language Reference Manual </div>
16<ol>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
Bill Wendling41a07852009-07-20 01:03:30 +000023 <li><a href="#linkage">Linkage Types</a>
24 <ol>
Bill Wendlinge2753242009-07-20 02:41:50 +000025 <li><a href="#linkage_private">'<tt>private</tt>' Linkage</a></li>
26 <li><a href="#linkage_linker_private">'<tt>linker_private</tt>' Linkage</a></li>
27 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
28 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
29 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
30 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
31 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
32 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
33 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_linkonce">'<tt>linkonce_odr</tt>' Linkage</a></li>
35 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
36 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
37 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
38 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
Bill Wendling41a07852009-07-20 01:03:30 +000039 </ol>
40 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000041 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +000042 <li><a href="#namedtypes">Named Types</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000043 <li><a href="#globalvars">Global Variables</a></li>
44 <li><a href="#functionstructure">Functions</a></li>
Dan Gohman2672f3e2008-10-14 16:51:45 +000045 <li><a href="#aliasstructure">Aliases</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000046 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel008cd3e2008-09-26 23:51:19 +000047 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen13fe5e32007-12-10 03:18:06 +000048 <li><a href="#gc">Garbage Collector Names</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000049 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
50 <li><a href="#datalayout">Data Layout</a></li>
Dan Gohman27b47012009-07-27 18:07:55 +000051 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000052 </ol>
53 </li>
54 <li><a href="#typesystem">Type System</a>
55 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +000056 <li><a href="#t_classifications">Type Classifications</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000057 <li><a href="#t_primitive">Primitive Types</a>
58 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +000059 <li><a href="#t_floating">Floating Point Types</a></li>
60 <li><a href="#t_void">Void Type</a></li>
61 <li><a href="#t_label">Label Type</a></li>
Nick Lewycky29aaef82009-05-30 05:06:04 +000062 <li><a href="#t_metadata">Metadata Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000063 </ol>
64 </li>
65 <li><a href="#t_derived">Derived Types</a>
66 <ol>
Chris Lattner251ab812007-12-18 06:18:21 +000067 <li><a href="#t_integer">Integer Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000068 <li><a href="#t_array">Array Type</a></li>
69 <li><a href="#t_function">Function Type</a></li>
70 <li><a href="#t_pointer">Pointer Type</a></li>
71 <li><a href="#t_struct">Structure Type</a></li>
72 <li><a href="#t_pstruct">Packed Structure Type</a></li>
73 <li><a href="#t_vector">Vector Type</a></li>
74 <li><a href="#t_opaque">Opaque Type</a></li>
75 </ol>
76 </li>
Chris Lattner515195a2009-02-02 07:32:36 +000077 <li><a href="#t_uprefs">Type Up-references</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000078 </ol>
79 </li>
80 <li><a href="#constants">Constants</a>
81 <ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +000082 <li><a href="#simpleconstants">Simple Constants</a></li>
Chris Lattner97063852009-02-28 18:32:25 +000083 <li><a href="#complexconstants">Complex Constants</a></li>
Dan Gohman2672f3e2008-10-14 16:51:45 +000084 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
85 <li><a href="#undefvalues">Undefined Values</a></li>
86 <li><a href="#constantexprs">Constant Expressions</a></li>
Nick Lewycky4dcf8102009-04-04 07:22:01 +000087 <li><a href="#metadata">Embedded Metadata</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000088 </ol>
89 </li>
90 <li><a href="#othervalues">Other Values</a>
91 <ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +000092 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +000093 </ol>
94 </li>
Chris Lattner75c24e02009-07-20 05:55:19 +000095 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
96 <ol>
97 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
Chris Lattner1e0e0d12009-07-20 06:14:25 +000098 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
99 Global Variable</a></li>
Chris Lattner75c24e02009-07-20 05:55:19 +0000100 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
101 Global Variable</a></li>
102 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
103 Global Variable</a></li>
104 </ol>
105 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000106 <li><a href="#instref">Instruction Reference</a>
107 <ol>
108 <li><a href="#terminators">Terminator Instructions</a>
109 <ol>
110 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
111 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
112 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
113 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
114 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
115 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
116 </ol>
117 </li>
118 <li><a href="#binaryops">Binary Operations</a>
119 <ol>
120 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
Dan Gohman7ce405e2009-06-04 22:49:04 +0000121 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000122 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
Dan Gohman7ce405e2009-06-04 22:49:04 +0000123 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000124 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
Dan Gohman7ce405e2009-06-04 22:49:04 +0000125 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000126 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
127 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
128 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
129 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
130 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
131 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
132 </ol>
133 </li>
134 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
135 <ol>
136 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
137 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
138 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
139 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
140 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
141 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
142 </ol>
143 </li>
144 <li><a href="#vectorops">Vector Operations</a>
145 <ol>
146 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
147 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
148 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
149 </ol>
150 </li>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000151 <li><a href="#aggregateops">Aggregate Operations</a>
152 <ol>
153 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
154 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
155 </ol>
156 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000157 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
158 <ol>
159 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
160 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
161 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
162 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
163 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
164 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
165 </ol>
166 </li>
167 <li><a href="#convertops">Conversion Operations</a>
168 <ol>
169 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
170 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
171 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
172 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
173 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
174 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
175 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
176 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
177 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
178 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
179 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
180 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
181 </ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000182 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000183 <li><a href="#otherops">Other Operations</a>
184 <ol>
185 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
186 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
187 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
188 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
189 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
190 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
191 </ol>
192 </li>
193 </ol>
194 </li>
195 <li><a href="#intrinsics">Intrinsic Functions</a>
196 <ol>
197 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
198 <ol>
199 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
200 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
201 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
202 </ol>
203 </li>
204 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
205 <ol>
206 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
207 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
208 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
209 </ol>
210 </li>
211 <li><a href="#int_codegen">Code Generator Intrinsics</a>
212 <ol>
213 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
214 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
215 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
216 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
217 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
218 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
219 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
220 </ol>
221 </li>
222 <li><a href="#int_libc">Standard C Library Intrinsics</a>
223 <ol>
224 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
225 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
226 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
227 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
228 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000229 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
230 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
231 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000232 </ol>
233 </li>
234 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
235 <ol>
236 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
237 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
238 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
239 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000240 </ol>
241 </li>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000242 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
243 <ol>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000244 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
245 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
246 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
247 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
248 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingbda98b62009-02-08 23:00:09 +0000249 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000250 </ol>
251 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000252 <li><a href="#int_debugger">Debugger intrinsics</a></li>
253 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000254 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000255 <ol>
256 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000257 </ol>
258 </li>
Bill Wendling9127adb2008-11-18 22:10:53 +0000259 <li><a href="#int_atomics">Atomic intrinsics</a>
260 <ol>
261 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
262 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
263 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
264 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
265 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
266 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
267 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
268 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
269 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
270 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
271 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
272 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
273 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
274 </ol>
275 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000276 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000277 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000278 <li><a href="#int_var_annotation">
Bill Wendlinge4164592008-11-19 05:56:17 +0000279 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000280 <li><a href="#int_annotation">
Bill Wendlinge4164592008-11-19 05:56:17 +0000281 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000282 <li><a href="#int_trap">
Bill Wendlinge4164592008-11-19 05:56:17 +0000283 '<tt>llvm.trap</tt>' Intrinsic</a></li>
284 <li><a href="#int_stackprotector">
285 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000286 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000287 </li>
288 </ol>
289 </li>
290</ol>
291
292<div class="doc_author">
293 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
294 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
295</div>
296
297<!-- *********************************************************************** -->
298<div class="doc_section"> <a name="abstract">Abstract </a></div>
299<!-- *********************************************************************** -->
300
301<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +0000302
303<p>This document is a reference manual for the LLVM assembly language. LLVM is
304 a Static Single Assignment (SSA) based representation that provides type
305 safety, low-level operations, flexibility, and the capability of representing
306 'all' high-level languages cleanly. It is the common code representation
307 used throughout all phases of the LLVM compilation strategy.</p>
308
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000309</div>
310
311<!-- *********************************************************************** -->
312<div class="doc_section"> <a name="introduction">Introduction</a> </div>
313<!-- *********************************************************************** -->
314
315<div class="doc_text">
316
Bill Wendlingf85859d2009-07-20 02:29:24 +0000317<p>The LLVM code representation is designed to be used in three different forms:
318 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
319 for fast loading by a Just-In-Time compiler), and as a human readable
320 assembly language representation. This allows LLVM to provide a powerful
321 intermediate representation for efficient compiler transformations and
322 analysis, while providing a natural means to debug and visualize the
323 transformations. The three different forms of LLVM are all equivalent. This
324 document describes the human readable representation and notation.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000325
Bill Wendlingf85859d2009-07-20 02:29:24 +0000326<p>The LLVM representation aims to be light-weight and low-level while being
327 expressive, typed, and extensible at the same time. It aims to be a
328 "universal IR" of sorts, by being at a low enough level that high-level ideas
329 may be cleanly mapped to it (similar to how microprocessors are "universal
330 IR's", allowing many source languages to be mapped to them). By providing
331 type information, LLVM can be used as the target of optimizations: for
332 example, through pointer analysis, it can be proven that a C automatic
333 variable is never accessed outside of the current function... allowing it to
334 be promoted to a simple SSA value instead of a memory location.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000335
336</div>
337
338<!-- _______________________________________________________________________ -->
339<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
340
341<div class="doc_text">
342
Bill Wendlingf85859d2009-07-20 02:29:24 +0000343<p>It is important to note that this document describes 'well formed' LLVM
344 assembly language. There is a difference between what the parser accepts and
345 what is considered 'well formed'. For example, the following instruction is
346 syntactically okay, but not well formed:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000347
348<div class="doc_code">
349<pre>
350%x = <a href="#i_add">add</a> i32 1, %x
351</pre>
352</div>
353
Bill Wendlingf85859d2009-07-20 02:29:24 +0000354<p>...because the definition of <tt>%x</tt> does not dominate all of its
355 uses. The LLVM infrastructure provides a verification pass that may be used
356 to verify that an LLVM module is well formed. This pass is automatically run
357 by the parser after parsing input assembly and by the optimizer before it
358 outputs bitcode. The violations pointed out by the verifier pass indicate
359 bugs in transformation passes or input to the parser.</p>
360
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000361</div>
362
Chris Lattnera83fdc02007-10-03 17:34:29 +0000363<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000364
365<!-- *********************************************************************** -->
366<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
367<!-- *********************************************************************** -->
368
369<div class="doc_text">
370
Bill Wendlingf85859d2009-07-20 02:29:24 +0000371<p>LLVM identifiers come in two basic types: global and local. Global
372 identifiers (functions, global variables) begin with the <tt>'@'</tt>
373 character. Local identifiers (register names, types) begin with
374 the <tt>'%'</tt> character. Additionally, there are three different formats
375 for identifiers, for different purposes:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000376
377<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000378 <li>Named values are represented as a string of characters with their prefix.
Bill Wendlingf85859d2009-07-20 02:29:24 +0000379 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
380 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
381 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
382 other characters in their names can be surrounded with quotes. Special
383 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
384 ASCII code for the character in hexadecimal. In this way, any character
385 can be used in a name value, even quotes themselves.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000386
Reid Spencerc8245b02007-08-07 14:34:28 +0000387 <li>Unnamed values are represented as an unsigned numeric value with their
Bill Wendlingf85859d2009-07-20 02:29:24 +0000388 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000389
390 <li>Constants, which are described in a <a href="#constants">section about
Bill Wendlingf85859d2009-07-20 02:29:24 +0000391 constants</a>, below.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000392</ol>
393
Reid Spencerc8245b02007-08-07 14:34:28 +0000394<p>LLVM requires that values start with a prefix for two reasons: Compilers
Bill Wendlingf85859d2009-07-20 02:29:24 +0000395 don't need to worry about name clashes with reserved words, and the set of
396 reserved words may be expanded in the future without penalty. Additionally,
397 unnamed identifiers allow a compiler to quickly come up with a temporary
398 variable without having to avoid symbol table conflicts.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000399
400<p>Reserved words in LLVM are very similar to reserved words in other
Bill Wendlingf85859d2009-07-20 02:29:24 +0000401 languages. There are keywords for different opcodes
402 ('<tt><a href="#i_add">add</a></tt>',
403 '<tt><a href="#i_bitcast">bitcast</a></tt>',
404 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
405 ('<tt><a href="#t_void">void</a></tt>',
406 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
407 reserved words cannot conflict with variable names, because none of them
408 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000409
410<p>Here is an example of LLVM code to multiply the integer variable
Bill Wendlingf85859d2009-07-20 02:29:24 +0000411 '<tt>%X</tt>' by 8:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000412
413<p>The easy way:</p>
414
415<div class="doc_code">
416<pre>
417%result = <a href="#i_mul">mul</a> i32 %X, 8
418</pre>
419</div>
420
421<p>After strength reduction:</p>
422
423<div class="doc_code">
424<pre>
425%result = <a href="#i_shl">shl</a> i32 %X, i8 3
426</pre>
427</div>
428
429<p>And the hard way:</p>
430
431<div class="doc_code">
432<pre>
433<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
434<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
435%result = <a href="#i_add">add</a> i32 %1, %1
436</pre>
437</div>
438
Bill Wendlingf85859d2009-07-20 02:29:24 +0000439<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
440 lexical features of LLVM:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000441
442<ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000443 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
Bill Wendlingf85859d2009-07-20 02:29:24 +0000444 line.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000445
446 <li>Unnamed temporaries are created when the result of a computation is not
Bill Wendlingf85859d2009-07-20 02:29:24 +0000447 assigned to a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000448
449 <li>Unnamed temporaries are numbered sequentially</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000450</ol>
451
452<p>...and it also shows a convention that we follow in this document. When
Bill Wendlingf85859d2009-07-20 02:29:24 +0000453 demonstrating instructions, we will follow an instruction with a comment that
454 defines the type and name of value produced. Comments are shown in italic
455 text.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000456
457</div>
458
459<!-- *********************************************************************** -->
460<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
461<!-- *********************************************************************** -->
462
463<!-- ======================================================================= -->
464<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
465</div>
466
467<div class="doc_text">
468
Bill Wendlingf85859d2009-07-20 02:29:24 +0000469<p>LLVM programs are composed of "Module"s, each of which is a translation unit
470 of the input programs. Each module consists of functions, global variables,
471 and symbol table entries. Modules may be combined together with the LLVM
472 linker, which merges function (and global variable) definitions, resolves
473 forward declarations, and merges symbol table entries. Here is an example of
474 the "hello world" module:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000475
476<div class="doc_code">
477<pre><i>; Declare the string constant as a global constant...</i>
478<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
479 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
480
481<i>; External declaration of the puts function</i>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000482<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000483
484<i>; Definition of main function</i>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000485define i32 @main() { <i>; i32()* </i>
Dan Gohman01852382009-01-04 23:44:43 +0000486 <i>; Convert [13 x i8]* to i8 *...</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000487 %cast210 = <a
Bill Wendlingf85859d2009-07-20 02:29:24 +0000488 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000489
490 <i>; Call puts function to write out the string to stdout...</i>
491 <a
Bill Wendlingf85859d2009-07-20 02:29:24 +0000492 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000493 <a
494 href="#i_ret">ret</a> i32 0<br>}<br>
495</pre>
496</div>
497
Bill Wendlingf85859d2009-07-20 02:29:24 +0000498<p>This example is made up of a <a href="#globalvars">global variable</a> named
499 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function, and
500 a <a href="#functionstructure">function definition</a> for
501 "<tt>main</tt>".</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000502
Bill Wendlingf85859d2009-07-20 02:29:24 +0000503<p>In general, a module is made up of a list of global values, where both
504 functions and global variables are global values. Global values are
505 represented by a pointer to a memory location (in this case, a pointer to an
506 array of char, and a pointer to a function), and have one of the
507 following <a href="#linkage">linkage types</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000508
509</div>
510
511<!-- ======================================================================= -->
512<div class="doc_subsection">
513 <a name="linkage">Linkage Types</a>
514</div>
515
516<div class="doc_text">
517
Bill Wendlingf85859d2009-07-20 02:29:24 +0000518<p>All Global Variables and Functions have one of the following types of
519 linkage:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000520
521<dl>
Rafael Espindolaa168fc92009-01-15 20:18:42 +0000522 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000523 <dd>Global values with private linkage are only directly accessible by objects
524 in the current module. In particular, linking code into a module with an
525 private global value may cause the private to be renamed as necessary to
526 avoid collisions. Because the symbol is private to the module, all
527 references can be updated. This doesn't show up in any symbol table in the
528 object file.</dd>
Rafael Espindolaa168fc92009-01-15 20:18:42 +0000529
Bill Wendling41a07852009-07-20 01:03:30 +0000530 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt>: </dt>
Bill Wendling41a07852009-07-20 01:03:30 +0000531 <dd>Similar to private, but the symbol is passed through the assembler and
532 removed by the linker after evaluation.</dd>
533
Dale Johannesen96e7e092008-05-23 23:13:41 +0000534 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000535 <dd>Similar to private, but the value shows as a local symbol
536 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
537 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000538
Bill Wendlingf85859d2009-07-20 02:29:24 +0000539 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt>: </dt>
Chris Lattner68433442009-04-13 05:44:34 +0000540 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
Bill Wendlingf85859d2009-07-20 02:29:24 +0000541 into the object file corresponding to the LLVM module. They exist to
542 allow inlining and other optimizations to take place given knowledge of
543 the definition of the global, which is known to be somewhere outside the
544 module. Globals with <tt>available_externally</tt> linkage are allowed to
545 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
546 This linkage type is only allowed on definitions, not declarations.</dd>
Chris Lattner68433442009-04-13 05:44:34 +0000547
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000548 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000549 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
Bill Wendlingf85859d2009-07-20 02:29:24 +0000550 the same name when linkage occurs. This is typically used to implement
551 inline functions, templates, or other code which must be generated in each
552 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
553 allowed to be discarded.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000554
Dale Johannesen96e7e092008-05-23 23:13:41 +0000555 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000556 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
557 linkage, except that unreferenced <tt>common</tt> globals may not be
558 discarded. This is used for globals that may be emitted in multiple
559 translation units, but that are not guaranteed to be emitted into every
560 translation unit that uses them. One example of this is tentative
561 definitions in C, such as "<tt>int X;</tt>" at global scope.</dd>
Dale Johannesen96e7e092008-05-23 23:13:41 +0000562
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000563 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
Dale Johannesen96e7e092008-05-23 23:13:41 +0000564 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
Bill Wendlingf85859d2009-07-20 02:29:24 +0000565 that some targets may choose to emit different assembly sequences for them
566 for target-dependent reasons. This is used for globals that are declared
567 "weak" in C source code.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000568
569 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000570 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
Bill Wendlingf85859d2009-07-20 02:29:24 +0000571 pointer to array type. When two global variables with appending linkage
572 are linked together, the two global arrays are appended together. This is
573 the LLVM, typesafe, equivalent of having the system linker append together
574 "sections" with identical names when .o files are linked.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000575
576 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000577 <dd>The semantics of this linkage follow the ELF object file model: the symbol
578 is weak until linked, if not linked, the symbol becomes null instead of
579 being an undefined reference.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000580
Duncan Sands19d161f2009-03-07 15:45:40 +0000581 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
Duncan Sands19d161f2009-03-07 15:45:40 +0000582 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000583 <dd>Some languages allow differing globals to be merged, such as two functions
584 with different semantics. Other languages, such as <tt>C++</tt>, ensure
585 that only equivalent globals are ever merged (the "one definition rule" -
586 "ODR"). Such languages can use the <tt>linkonce_odr</tt>
587 and <tt>weak_odr</tt> linkage types to indicate that the global will only
588 be merged with equivalent globals. These linkage types are otherwise the
589 same as their non-<tt>odr</tt> versions.</dd>
Duncan Sands19d161f2009-03-07 15:45:40 +0000590
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000591 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000592 <dd>If none of the above identifiers are used, the global is externally
Bill Wendlingf85859d2009-07-20 02:29:24 +0000593 visible, meaning that it participates in linkage and can be used to
594 resolve external symbol references.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000595</dl>
596
Bill Wendlingf85859d2009-07-20 02:29:24 +0000597<p>The next two types of linkage are targeted for Microsoft Windows platform
598 only. They are designed to support importing (exporting) symbols from (to)
599 DLLs (Dynamic Link Libraries).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000600
Bill Wendlingf85859d2009-07-20 02:29:24 +0000601<dl>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000602 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000603 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
Bill Wendlingf85859d2009-07-20 02:29:24 +0000604 or variable via a global pointer to a pointer that is set up by the DLL
605 exporting the symbol. On Microsoft Windows targets, the pointer name is
606 formed by combining <code>__imp_</code> and the function or variable
607 name.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000608
609 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000610 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
Bill Wendlingf85859d2009-07-20 02:29:24 +0000611 pointer to a pointer in a DLL, so that it can be referenced with the
612 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
613 name is formed by combining <code>__imp_</code> and the function or
614 variable name.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000615</dl>
616
Bill Wendlingf85859d2009-07-20 02:29:24 +0000617<p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
618 another module defined a "<tt>.LC0</tt>" variable and was linked with this
619 one, one of the two would be renamed, preventing a collision. Since
620 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
621 declarations), they are accessible outside of the current module.</p>
622
623<p>It is illegal for a function <i>declaration</i> to have any linkage type
624 other than "externally visible", <tt>dllimport</tt>
625 or <tt>extern_weak</tt>.</p>
626
Duncan Sands19d161f2009-03-07 15:45:40 +0000627<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000628 or <tt>weak_odr</tt> linkages.</p>
629
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000630</div>
631
632<!-- ======================================================================= -->
633<div class="doc_subsection">
634 <a name="callingconv">Calling Conventions</a>
635</div>
636
637<div class="doc_text">
638
639<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000640 and <a href="#i_invoke">invokes</a> can all have an optional calling
641 convention specified for the call. The calling convention of any pair of
642 dynamic caller/callee must match, or the behavior of the program is
643 undefined. The following calling conventions are supported by LLVM, and more
644 may be added in the future:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000645
646<dl>
647 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000648 <dd>This calling convention (the default if no other calling convention is
Bill Wendlingf85859d2009-07-20 02:29:24 +0000649 specified) matches the target C calling conventions. This calling
650 convention supports varargs function calls and tolerates some mismatch in
651 the declared prototype and implemented declaration of the function (as
652 does normal C).</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000653
654 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000655 <dd>This calling convention attempts to make calls as fast as possible
Bill Wendlingf85859d2009-07-20 02:29:24 +0000656 (e.g. by passing things in registers). This calling convention allows the
657 target to use whatever tricks it wants to produce fast code for the
658 target, without having to conform to an externally specified ABI
659 (Application Binary Interface). Implementations of this convention should
660 allow arbitrary <a href="CodeGenerator.html#tailcallopt">tail call
661 optimization</a> to be supported. This calling convention does not
662 support varargs and requires the prototype of all callees to exactly match
663 the prototype of the function definition.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000664
665 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000666 <dd>This calling convention attempts to make code in the caller as efficient
Bill Wendlingf85859d2009-07-20 02:29:24 +0000667 as possible under the assumption that the call is not commonly executed.
668 As such, these calls often preserve all registers so that the call does
669 not break any live ranges in the caller side. This calling convention
670 does not support varargs and requires the prototype of all callees to
671 exactly match the prototype of the function definition.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000672
673 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000674 <dd>Any calling convention may be specified by number, allowing
Bill Wendlingf85859d2009-07-20 02:29:24 +0000675 target-specific calling conventions to be used. Target specific calling
676 conventions start at 64.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000677</dl>
678
679<p>More calling conventions can be added/defined on an as-needed basis, to
Bill Wendlingf85859d2009-07-20 02:29:24 +0000680 support Pascal conventions or any other well-known target-independent
681 convention.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000682
683</div>
684
685<!-- ======================================================================= -->
686<div class="doc_subsection">
687 <a name="visibility">Visibility Styles</a>
688</div>
689
690<div class="doc_text">
691
Bill Wendlingf85859d2009-07-20 02:29:24 +0000692<p>All Global Variables and Functions have one of the following visibility
693 styles:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000694
695<dl>
696 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
Chris Lattner96451482008-08-05 18:29:16 +0000697 <dd>On targets that use the ELF object file format, default visibility means
Bill Wendlingf85859d2009-07-20 02:29:24 +0000698 that the declaration is visible to other modules and, in shared libraries,
699 means that the declared entity may be overridden. On Darwin, default
700 visibility means that the declaration is visible to other modules. Default
701 visibility corresponds to "external linkage" in the language.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000702
703 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000704 <dd>Two declarations of an object with hidden visibility refer to the same
Bill Wendlingf85859d2009-07-20 02:29:24 +0000705 object if they are in the same shared object. Usually, hidden visibility
706 indicates that the symbol will not be placed into the dynamic symbol
707 table, so no other module (executable or shared library) can reference it
708 directly.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000709
710 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000711 <dd>On ELF, protected visibility indicates that the symbol will be placed in
Bill Wendlingf85859d2009-07-20 02:29:24 +0000712 the dynamic symbol table, but that references within the defining module
713 will bind to the local symbol. That is, the symbol cannot be overridden by
714 another module.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000715</dl>
716
717</div>
718
719<!-- ======================================================================= -->
720<div class="doc_subsection">
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000721 <a name="namedtypes">Named Types</a>
722</div>
723
724<div class="doc_text">
725
726<p>LLVM IR allows you to specify name aliases for certain types. This can make
Bill Wendlingf85859d2009-07-20 02:29:24 +0000727 it easier to read the IR and make the IR more condensed (particularly when
728 recursive types are involved). An example of a name specification is:</p>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000729
730<div class="doc_code">
731<pre>
732%mytype = type { %mytype*, i32 }
733</pre>
734</div>
735
Bill Wendlingf85859d2009-07-20 02:29:24 +0000736<p>You may give a name to any <a href="#typesystem">type</a> except
737 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
738 is expected with the syntax "%mytype".</p>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000739
740<p>Note that type names are aliases for the structural type that they indicate,
Bill Wendlingf85859d2009-07-20 02:29:24 +0000741 and that you can therefore specify multiple names for the same type. This
742 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
743 uses structural typing, the name is not part of the type. When printing out
744 LLVM IR, the printer will pick <em>one name</em> to render all types of a
745 particular shape. This means that if you have code where two different
746 source types end up having the same LLVM type, that the dumper will sometimes
747 print the "wrong" or unexpected type. This is an important design point and
748 isn't going to change.</p>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000749
750</div>
751
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000752<!-- ======================================================================= -->
753<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000754 <a name="globalvars">Global Variables</a>
755</div>
756
757<div class="doc_text">
758
759<p>Global variables define regions of memory allocated at compilation time
Bill Wendlingf85859d2009-07-20 02:29:24 +0000760 instead of run-time. Global variables may optionally be initialized, may
761 have an explicit section to be placed in, and may have an optional explicit
762 alignment specified. A variable may be defined as "thread_local", which
763 means that it will not be shared by threads (each thread will have a
764 separated copy of the variable). A variable may be defined as a global
765 "constant," which indicates that the contents of the variable
766 will <b>never</b> be modified (enabling better optimization, allowing the
767 global data to be placed in the read-only section of an executable, etc).
768 Note that variables that need runtime initialization cannot be marked
769 "constant" as there is a store to the variable.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000770
Bill Wendlingf85859d2009-07-20 02:29:24 +0000771<p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
772 constant, even if the final definition of the global is not. This capability
773 can be used to enable slightly better optimization of the program, but
774 requires the language definition to guarantee that optimizations based on the
775 'constantness' are valid for the translation units that do not include the
776 definition.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000777
Bill Wendlingf85859d2009-07-20 02:29:24 +0000778<p>As SSA values, global variables define pointer values that are in scope
779 (i.e. they dominate) all basic blocks in the program. Global variables
780 always define a pointer to their "content" type because they describe a
781 region of memory, and all memory objects in LLVM are accessed through
782 pointers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000783
Bill Wendlingf85859d2009-07-20 02:29:24 +0000784<p>A global variable may be declared to reside in a target-specific numbered
785 address space. For targets that support them, address spaces may affect how
786 optimizations are performed and/or what target instructions are used to
787 access the variable. The default address space is zero. The address space
788 qualifier must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000789
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000790<p>LLVM allows an explicit section to be specified for globals. If the target
Bill Wendlingf85859d2009-07-20 02:29:24 +0000791 supports it, it will emit globals to the section specified.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000792
793<p>An explicit alignment may be specified for a global. If not present, or if
Bill Wendlingf85859d2009-07-20 02:29:24 +0000794 the alignment is set to zero, the alignment of the global is set by the
795 target to whatever it feels convenient. If an explicit alignment is
796 specified, the global is forced to have at least that much alignment. All
797 alignments must be a power of 2.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000798
Bill Wendlingf85859d2009-07-20 02:29:24 +0000799<p>For example, the following defines a global in a numbered address space with
800 an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000801
802<div class="doc_code">
803<pre>
Dan Gohman21ef02c2009-01-11 00:40:00 +0000804@G = addrspace(5) constant float 1.0, section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000805</pre>
806</div>
807
808</div>
809
810
811<!-- ======================================================================= -->
812<div class="doc_subsection">
813 <a name="functionstructure">Functions</a>
814</div>
815
816<div class="doc_text">
817
Bill Wendlingf85859d2009-07-20 02:29:24 +0000818<p>LLVM function definitions consist of the "<tt>define</tt>" keyord, an
819 optional <a href="#linkage">linkage type</a>, an optional
820 <a href="#visibility">visibility style</a>, an optional
821 <a href="#callingconv">calling convention</a>, a return type, an optional
822 <a href="#paramattrs">parameter attribute</a> for the return type, a function
823 name, a (possibly empty) argument list (each with optional
824 <a href="#paramattrs">parameter attributes</a>), optional
825 <a href="#fnattrs">function attributes</a>, an optional section, an optional
826 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
827 curly brace, a list of basic blocks, and a closing curly brace.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000828
Bill Wendlingf85859d2009-07-20 02:29:24 +0000829<p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
830 optional <a href="#linkage">linkage type</a>, an optional
831 <a href="#visibility">visibility style</a>, an optional
832 <a href="#callingconv">calling convention</a>, a return type, an optional
833 <a href="#paramattrs">parameter attribute</a> for the return type, a function
834 name, a possibly empty list of arguments, an optional alignment, and an
835 optional <a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000836
Chris Lattner96451482008-08-05 18:29:16 +0000837<p>A function definition contains a list of basic blocks, forming the CFG
Bill Wendlingf85859d2009-07-20 02:29:24 +0000838 (Control Flow Graph) for the function. Each basic block may optionally start
839 with a label (giving the basic block a symbol table entry), contains a list
840 of instructions, and ends with a <a href="#terminators">terminator</a>
841 instruction (such as a branch or function return).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000842
843<p>The first basic block in a function is special in two ways: it is immediately
Bill Wendlingf85859d2009-07-20 02:29:24 +0000844 executed on entrance to the function, and it is not allowed to have
845 predecessor basic blocks (i.e. there can not be any branches to the entry
846 block of a function). Because the block can have no predecessors, it also
847 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000848
849<p>LLVM allows an explicit section to be specified for functions. If the target
Bill Wendlingf85859d2009-07-20 02:29:24 +0000850 supports it, it will emit functions to the section specified.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000851
852<p>An explicit alignment may be specified for a function. If not present, or if
Bill Wendlingf85859d2009-07-20 02:29:24 +0000853 the alignment is set to zero, the alignment of the function is set by the
854 target to whatever it feels convenient. If an explicit alignment is
855 specified, the function is forced to have at least that much alignment. All
856 alignments must be a power of 2.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000857
Bill Wendling6ec40612009-07-20 02:39:26 +0000858<h5>Syntax:</h5>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000859<div class="doc_code">
Bill Wendlingf85859d2009-07-20 02:29:24 +0000860<pre>
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000861define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
Bill Wendlingf85859d2009-07-20 02:29:24 +0000862 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
863 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
864 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
865 [<a href="#gc">gc</a>] { ... }
866</pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000867</div>
868
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000869</div>
870
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000871<!-- ======================================================================= -->
872<div class="doc_subsection">
873 <a name="aliasstructure">Aliases</a>
874</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000875
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000876<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +0000877
878<p>Aliases act as "second name" for the aliasee value (which can be either
879 function, global variable, another alias or bitcast of global value). Aliases
880 may have an optional <a href="#linkage">linkage type</a>, and an
881 optional <a href="#visibility">visibility style</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000882
Bill Wendling6ec40612009-07-20 02:39:26 +0000883<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000884<div class="doc_code">
885<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000886@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000887</pre>
888</div>
889
890</div>
891
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000892<!-- ======================================================================= -->
893<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000894
Bill Wendlingf85859d2009-07-20 02:29:24 +0000895<div class="doc_text">
896
897<p>The return type and each parameter of a function type may have a set of
898 <i>parameter attributes</i> associated with them. Parameter attributes are
899 used to communicate additional information about the result or parameters of
900 a function. Parameter attributes are considered to be part of the function,
901 not of the function type, so functions with different parameter attributes
902 can have the same function type.</p>
903
904<p>Parameter attributes are simple keywords that follow the type specified. If
905 multiple parameter attributes are needed, they are space separated. For
906 example:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000907
908<div class="doc_code">
909<pre>
Nick Lewycky3022a742009-02-15 23:06:14 +0000910declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattnerf33b8452008-10-04 18:33:34 +0000911declare i32 @atoi(i8 zeroext)
912declare signext i8 @returns_signed_char()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000913</pre>
914</div>
915
Bill Wendlingf85859d2009-07-20 02:29:24 +0000916<p>Note that any attributes for the function result (<tt>nounwind</tt>,
917 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000918
Bill Wendlingf85859d2009-07-20 02:29:24 +0000919<p>Currently, only the following parameter attributes are defined:</p>
Chris Lattner275e6be2008-01-11 06:20:47 +0000920
Bill Wendlingf85859d2009-07-20 02:29:24 +0000921<dl>
922 <dt><tt>zeroext</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000923 <dd>This indicates to the code generator that the parameter or return value
924 should be zero-extended to a 32-bit value by the caller (for a parameter)
925 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000926
Bill Wendlingf85859d2009-07-20 02:29:24 +0000927 <dt><tt>signext</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000928 <dd>This indicates to the code generator that the parameter or return value
929 should be sign-extended to a 32-bit value by the caller (for a parameter)
930 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000931
Bill Wendlingf85859d2009-07-20 02:29:24 +0000932 <dt><tt>inreg</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000933 <dd>This indicates that this parameter or return value should be treated in a
934 special target-dependent fashion during while emitting code for a function
935 call or return (usually, by putting it in a register as opposed to memory,
936 though some targets use it to distinguish between two different kinds of
937 registers). Use of this attribute is target-specific.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000938
Bill Wendlingf85859d2009-07-20 02:29:24 +0000939 <dt><tt><a name="byval">byval</a></tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000940 <dd>This indicates that the pointer parameter should really be passed by value
941 to the function. The attribute implies that a hidden copy of the pointee
942 is made between the caller and the callee, so the callee is unable to
943 modify the value in the callee. This attribute is only valid on LLVM
944 pointer arguments. It is generally used to pass structs and arrays by
945 value, but is also valid on pointers to scalars. The copy is considered
946 to belong to the caller not the callee (for example,
947 <tt><a href="#readonly">readonly</a></tt> functions should not write to
948 <tt>byval</tt> parameters). This is not a valid attribute for return
949 values. The byval attribute also supports specifying an alignment with
950 the align attribute. This has a target-specific effect on the code
951 generator that usually indicates a desired alignment for the synthesized
952 stack slot.</dd>
953
954 <dt><tt>sret</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000955 <dd>This indicates that the pointer parameter specifies the address of a
956 structure that is the return value of the function in the source program.
957 This pointer must be guaranteed by the caller to be valid: loads and
958 stores to the structure may be assumed by the callee to not to trap. This
959 may only be applied to the first parameter. This is not a valid attribute
960 for return values. </dd>
961
962 <dt><tt>noalias</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000963 <dd>This indicates that the pointer does not alias any global or any other
964 parameter. The caller is responsible for ensuring that this is the
965 case. On a function return value, <tt>noalias</tt> additionally indicates
966 that the pointer does not alias any other pointers visible to the
967 caller. For further details, please see the discussion of the NoAlias
968 response in
969 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
970 analysis</a>.</dd>
971
972 <dt><tt>nocapture</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000973 <dd>This indicates that the callee does not make any copies of the pointer
974 that outlive the callee itself. This is not a valid attribute for return
975 values.</dd>
976
977 <dt><tt>nest</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +0000978 <dd>This indicates that the pointer parameter can be excised using the
979 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
980 attribute for return values.</dd>
981</dl>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000982
983</div>
984
985<!-- ======================================================================= -->
986<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000987 <a name="gc">Garbage Collector Names</a>
988</div>
989
990<div class="doc_text">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000991
Bill Wendlingf85859d2009-07-20 02:29:24 +0000992<p>Each function may specify a garbage collector name, which is simply a
993 string:</p>
994
995<div class="doc_code">
996<pre>
997define void @f() gc "name" { ...
998</pre>
999</div>
Gordon Henriksen13fe5e32007-12-10 03:18:06 +00001000
1001<p>The compiler declares the supported values of <i>name</i>. Specifying a
Bill Wendlingf85859d2009-07-20 02:29:24 +00001002 collector which will cause the compiler to alter its output in order to
1003 support the named garbage collection algorithm.</p>
1004
Gordon Henriksen13fe5e32007-12-10 03:18:06 +00001005</div>
1006
1007<!-- ======================================================================= -->
1008<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +00001009 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +00001010</div>
1011
1012<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +00001013
Bill Wendlingf85859d2009-07-20 02:29:24 +00001014<p>Function attributes are set to communicate additional information about a
1015 function. Function attributes are considered to be part of the function, not
1016 of the function type, so functions with different parameter attributes can
1017 have the same function type.</p>
Devang Patel008cd3e2008-09-26 23:51:19 +00001018
Bill Wendlingf85859d2009-07-20 02:29:24 +00001019<p>Function attributes are simple keywords that follow the type specified. If
1020 multiple attributes are needed, they are space separated. For example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +00001021
1022<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +00001023<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +00001024define void @f() noinline { ... }
1025define void @f() alwaysinline { ... }
1026define void @f() alwaysinline optsize { ... }
1027define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +00001028</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +00001029</div>
1030
Bill Wendling74d3eac2008-09-07 10:26:33 +00001031<dl>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001032 <dt><tt>alwaysinline</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001033 <dd>This attribute indicates that the inliner should attempt to inline this
1034 function into callers whenever possible, ignoring any active inlining size
1035 threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +00001036
Bill Wendlingf85859d2009-07-20 02:29:24 +00001037 <dt><tt>noinline</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001038 <dd>This attribute indicates that the inliner should never inline this
1039 function in any situation. This attribute may not be used together with
1040 the <tt>alwaysinline</tt> attribute.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +00001041
Bill Wendlingf85859d2009-07-20 02:29:24 +00001042 <dt><tt>optsize</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001043 <dd>This attribute suggests that optimization passes and code generator passes
1044 make choices that keep the code size of this function low, and otherwise
1045 do optimizations specifically to reduce code size.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +00001046
Bill Wendlingf85859d2009-07-20 02:29:24 +00001047 <dt><tt>noreturn</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001048 <dd>This function attribute indicates that the function never returns
1049 normally. This produces undefined behavior at runtime if the function
1050 ever does dynamically return.</dd>
Bill Wendlingdfaabba2008-11-13 01:02:51 +00001051
Bill Wendlingf85859d2009-07-20 02:29:24 +00001052 <dt><tt>nounwind</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001053 <dd>This function attribute indicates that the function never returns with an
1054 unwind or exceptional control flow. If the function does unwind, its
1055 runtime behavior is undefined.</dd>
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +00001056
Bill Wendlingf85859d2009-07-20 02:29:24 +00001057 <dt><tt>readnone</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001058 <dd>This attribute indicates that the function computes its result (or decides
1059 to unwind an exception) based strictly on its arguments, without
1060 dereferencing any pointer arguments or otherwise accessing any mutable
1061 state (e.g. memory, control registers, etc) visible to caller functions.
1062 It does not write through any pointer arguments
1063 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1064 changes any state visible to callers. This means that it cannot unwind
1065 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1066 could use the <tt>unwind</tt> instruction.</dd>
Devang Patela2f9f412009-06-12 19:45:19 +00001067
Bill Wendlingf85859d2009-07-20 02:29:24 +00001068 <dt><tt><a name="readonly">readonly</a></tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001069 <dd>This attribute indicates that the function does not write through any
1070 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1071 arguments) or otherwise modify any state (e.g. memory, control registers,
1072 etc) visible to caller functions. It may dereference pointer arguments
1073 and read state that may be set in the caller. A readonly function always
1074 returns the same value (or unwinds an exception identically) when called
1075 with the same set of arguments and global state. It cannot unwind an
1076 exception by calling the <tt>C++</tt> exception throwing methods, but may
1077 use the <tt>unwind</tt> instruction.</dd>
Anton Korobeynikovedd7d112009-07-17 18:07:26 +00001078
Bill Wendlingf85859d2009-07-20 02:29:24 +00001079 <dt><tt><a name="ssp">ssp</a></tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001080 <dd>This attribute indicates that the function should emit a stack smashing
1081 protector. It is in the form of a "canary"&mdash;a random value placed on
1082 the stack before the local variables that's checked upon return from the
1083 function to see if it has been overwritten. A heuristic is used to
1084 determine if a function needs stack protectors or not.<br>
1085<br>
1086 If a function that has an <tt>ssp</tt> attribute is inlined into a
1087 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1088 function will have an <tt>ssp</tt> attribute.</dd>
1089
1090 <dt><tt>sspreq</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001091 <dd>This attribute indicates that the function should <em>always</em> emit a
1092 stack smashing protector. This overrides
Bill Wendling6ec40612009-07-20 02:39:26 +00001093 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1094<br>
1095 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1096 function that doesn't have an <tt>sspreq</tt> attribute or which has
1097 an <tt>ssp</tt> attribute, then the resulting function will have
1098 an <tt>sspreq</tt> attribute.</dd>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001099
1100 <dt><tt>noredzone</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001101 <dd>This attribute indicates that the code generator should not use a red
1102 zone, even if the target-specific ABI normally permits it.</dd>
1103
1104 <dt><tt>noimplicitfloat</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001105 <dd>This attributes disables implicit floating point instructions.</dd>
1106
1107 <dt><tt>naked</tt></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001108 <dd>This attribute disables prologue / epilogue emission for the function.
1109 This can have very system-specific consequences.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +00001110</dl>
1111
Devang Pateld468f1c2008-09-04 23:05:13 +00001112</div>
1113
1114<!-- ======================================================================= -->
1115<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001116 <a name="moduleasm">Module-Level Inline Assembly</a>
1117</div>
1118
1119<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001120
1121<p>Modules may contain "module-level inline asm" blocks, which corresponds to
1122 the GCC "file scope inline asm" blocks. These blocks are internally
1123 concatenated by LLVM and treated as a single unit, but may be separated in
1124 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001125
1126<div class="doc_code">
1127<pre>
1128module asm "inline asm code goes here"
1129module asm "more can go here"
1130</pre>
1131</div>
1132
1133<p>The strings can contain any character by escaping non-printable characters.
1134 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
Bill Wendlingf85859d2009-07-20 02:29:24 +00001135 for the number.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001136
Bill Wendlingf85859d2009-07-20 02:29:24 +00001137<p>The inline asm code is simply printed to the machine code .s file when
1138 assembly code is generated.</p>
1139
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001140</div>
1141
1142<!-- ======================================================================= -->
1143<div class="doc_subsection">
1144 <a name="datalayout">Data Layout</a>
1145</div>
1146
1147<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001148
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001149<p>A module may specify a target specific data layout string that specifies how
Bill Wendlingf85859d2009-07-20 02:29:24 +00001150 data is to be laid out in memory. The syntax for the data layout is
1151 simply:</p>
1152
1153<div class="doc_code">
1154<pre>
1155target datalayout = "<i>layout specification</i>"
1156</pre>
1157</div>
1158
1159<p>The <i>layout specification</i> consists of a list of specifications
1160 separated by the minus sign character ('-'). Each specification starts with
1161 a letter and may include other information after the letter to define some
1162 aspect of the data layout. The specifications accepted are as follows:</p>
1163
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001164<dl>
1165 <dt><tt>E</tt></dt>
1166 <dd>Specifies that the target lays out data in big-endian form. That is, the
Bill Wendlingf85859d2009-07-20 02:29:24 +00001167 bits with the most significance have the lowest address location.</dd>
1168
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001169 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +00001170 <dd>Specifies that the target lays out data in little-endian form. That is,
Bill Wendlingf85859d2009-07-20 02:29:24 +00001171 the bits with the least significance have the lowest address
1172 location.</dd>
1173
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001174 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1175 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
Bill Wendlingf85859d2009-07-20 02:29:24 +00001176 <i>preferred</i> alignments. All sizes are in bits. Specifying
1177 the <i>pref</i> alignment is optional. If omitted, the
1178 preceding <tt>:</tt> should be omitted too.</dd>
1179
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001180 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1181 <dd>This specifies the alignment for an integer type of a given bit
Bill Wendlingf85859d2009-07-20 02:29:24 +00001182 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1183
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001184 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1185 <dd>This specifies the alignment for a vector type of a given bit
Bill Wendlingf85859d2009-07-20 02:29:24 +00001186 <i>size</i>.</dd>
1187
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001188 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1189 <dd>This specifies the alignment for a floating point type of a given bit
Bill Wendlingf85859d2009-07-20 02:29:24 +00001190 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1191 (double).</dd>
1192
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001193 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1194 <dd>This specifies the alignment for an aggregate type of a given bit
Bill Wendlingf85859d2009-07-20 02:29:24 +00001195 <i>size</i>.</dd>
1196
Daniel Dunbard88a97b2009-06-08 22:17:53 +00001197 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1198 <dd>This specifies the alignment for a stack object of a given bit
Bill Wendlingf85859d2009-07-20 02:29:24 +00001199 <i>size</i>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001200</dl>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001201
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001202<p>When constructing the data layout for a given target, LLVM starts with a
Bill Wendlingf85859d2009-07-20 02:29:24 +00001203 default set of specifications which are then (possibly) overriden by the
1204 specifications in the <tt>datalayout</tt> keyword. The default specifications
1205 are given in this list:</p>
1206
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001207<ul>
1208 <li><tt>E</tt> - big endian</li>
1209 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1210 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1211 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1212 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1213 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +00001214 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001215 alignment of 64-bits</li>
1216 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1217 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1218 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1219 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1220 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
Daniel Dunbard88a97b2009-06-08 22:17:53 +00001221 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001222</ul>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001223
1224<p>When LLVM is determining the alignment for a given type, it uses the
1225 following rules:</p>
1226
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001227<ol>
1228 <li>If the type sought is an exact match for one of the specifications, that
Bill Wendlingf85859d2009-07-20 02:29:24 +00001229 specification is used.</li>
1230
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001231 <li>If no match is found, and the type sought is an integer type, then the
Bill Wendlingf85859d2009-07-20 02:29:24 +00001232 smallest integer type that is larger than the bitwidth of the sought type
1233 is used. If none of the specifications are larger than the bitwidth then
1234 the the largest integer type is used. For example, given the default
1235 specifications above, the i7 type will use the alignment of i8 (next
1236 largest) while both i65 and i256 will use the alignment of i64 (largest
1237 specified).</li>
1238
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001239 <li>If no match is found, and the type sought is a vector type, then the
Bill Wendlingf85859d2009-07-20 02:29:24 +00001240 largest vector type that is smaller than the sought vector type will be
1241 used as a fall back. This happens because &lt;128 x double&gt; can be
1242 implemented in terms of 64 &lt;2 x double&gt;, for example.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001243</ol>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001244
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001245</div>
1246
Dan Gohman27b47012009-07-27 18:07:55 +00001247<!-- ======================================================================= -->
1248<div class="doc_subsection">
1249 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1250</div>
1251
1252<div class="doc_text">
1253
1254<p>Any memory access must be done though a pointer value associated
Andreas Bolka23bece42009-07-27 20:37:10 +00001255with an address range of the memory access, otherwise the behavior
Dan Gohman27b47012009-07-27 18:07:55 +00001256is undefined. Pointer values are associated with address ranges
1257according to the following rules:</p>
1258
1259<ul>
1260 <li>A pointer value formed from a <tt>getelementptr</tt> instruction is
1261 associated with the addresses associated with the first operand of
1262 the <tt>getelementptr</tt>.</li>
1263 <li>An addresses of a global variable is associated with the address
1264 range of the variable's storage.</li>
1265 <li>The result value of an allocation instruction is associated with
1266 the address range of the allocated storage.</li>
1267 <li>A null pointer in the default address-space is associated with
1268 no addresses.</li>
1269 <li>A pointer value formed by an <tt>inttoptr</tt> is associated with
1270 all address ranges of all pointer values that contribute (directly
1271 or indirectly) to the computation of the pointer's value.</li>
1272 <li>The result value of a <tt>bitcast</tt> is associated with all
1273 addresses associated with the operand of the <tt>bitcast</tt>.</li>
1274 <li>An integer constant other than zero or a pointer value returned
1275 from a function not defined within LLVM may be associated with address
1276 ranges allocated through mechanisms other than those provided by
1277 LLVM. Such ranges shall not overlap with any ranges of address
1278 allocated by mechanisms provided by LLVM.</li>
1279 </ul>
1280
1281<p>LLVM IR does not associate types with memory. The result type of a
1282<tt>load</tt> merely indicates the size and alignment of the memory from
1283which to load, as well as the interpretation of the value. The first
1284operand of a <tt>store</tt> similarly only indicates the size and
1285alignment of the store.</p>
1286
1287<p>Consequently, type-based alias analysis, aka TBAA, aka
1288<tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1289LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1290additional information which specialized optimization passes may use
1291to implement type-based alias analysis.</p>
1292
1293</div>
1294
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001295<!-- *********************************************************************** -->
1296<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1297<!-- *********************************************************************** -->
1298
1299<div class="doc_text">
1300
1301<p>The LLVM type system is one of the most important features of the
Bill Wendlingf85859d2009-07-20 02:29:24 +00001302 intermediate representation. Being typed enables a number of optimizations
1303 to be performed on the intermediate representation directly, without having
1304 to do extra analyses on the side before the transformation. A strong type
1305 system makes it easier to read the generated code and enables novel analyses
1306 and transformations that are not feasible to perform on normal three address
1307 code representations.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001308
1309</div>
1310
1311<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +00001312<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001313Classifications</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001314
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001315<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001316
1317<p>The types fall into a few useful classifications:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001318
1319<table border="1" cellspacing="0" cellpadding="4">
1320 <tbody>
1321 <tr><th>Classification</th><th>Types</th></tr>
1322 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +00001323 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001324 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1325 </tr>
1326 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +00001327 <td><a href="#t_floating">floating point</a></td>
1328 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001329 </tr>
1330 <tr>
1331 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +00001332 <td><a href="#t_integer">integer</a>,
1333 <a href="#t_floating">floating point</a>,
1334 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +00001335 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +00001336 <a href="#t_struct">structure</a>,
1337 <a href="#t_array">array</a>,
Nick Lewycky29aaef82009-05-30 05:06:04 +00001338 <a href="#t_label">label</a>,
1339 <a href="#t_metadata">metadata</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001340 </td>
1341 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +00001342 <tr>
1343 <td><a href="#t_primitive">primitive</a></td>
1344 <td><a href="#t_label">label</a>,
1345 <a href="#t_void">void</a>,
Nick Lewycky29aaef82009-05-30 05:06:04 +00001346 <a href="#t_floating">floating point</a>,
1347 <a href="#t_metadata">metadata</a>.</td>
Chris Lattner488772f2008-01-04 04:32:38 +00001348 </tr>
1349 <tr>
1350 <td><a href="#t_derived">derived</a></td>
1351 <td><a href="#t_integer">integer</a>,
1352 <a href="#t_array">array</a>,
1353 <a href="#t_function">function</a>,
1354 <a href="#t_pointer">pointer</a>,
1355 <a href="#t_struct">structure</a>,
1356 <a href="#t_pstruct">packed structure</a>,
1357 <a href="#t_vector">vector</a>,
1358 <a href="#t_opaque">opaque</a>.
Dan Gohman032ba852008-10-14 16:32:04 +00001359 </td>
Chris Lattner488772f2008-01-04 04:32:38 +00001360 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001361 </tbody>
1362</table>
1363
Bill Wendlingf85859d2009-07-20 02:29:24 +00001364<p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1365 important. Values of these types are the only ones which can be produced by
1366 instructions, passed as arguments, or used as operands to instructions.</p>
1367
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001368</div>
1369
1370<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +00001371<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +00001372
Chris Lattner488772f2008-01-04 04:32:38 +00001373<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001374
Chris Lattner488772f2008-01-04 04:32:38 +00001375<p>The primitive types are the fundamental building blocks of the LLVM
Bill Wendlingf85859d2009-07-20 02:29:24 +00001376 system.</p>
Chris Lattner488772f2008-01-04 04:32:38 +00001377
Chris Lattner86437612008-01-04 04:34:14 +00001378</div>
1379
Chris Lattner488772f2008-01-04 04:32:38 +00001380<!-- _______________________________________________________________________ -->
1381<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1382
1383<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001384
1385<table>
1386 <tbody>
1387 <tr><th>Type</th><th>Description</th></tr>
1388 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1389 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1390 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1391 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1392 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1393 </tbody>
1394</table>
1395
Chris Lattner488772f2008-01-04 04:32:38 +00001396</div>
1397
1398<!-- _______________________________________________________________________ -->
1399<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1400
1401<div class="doc_text">
Bill Wendling6ec40612009-07-20 02:39:26 +00001402
Chris Lattner488772f2008-01-04 04:32:38 +00001403<h5>Overview:</h5>
1404<p>The void type does not represent any value and has no size.</p>
1405
1406<h5>Syntax:</h5>
Chris Lattner488772f2008-01-04 04:32:38 +00001407<pre>
1408 void
1409</pre>
Bill Wendling6ec40612009-07-20 02:39:26 +00001410
Chris Lattner488772f2008-01-04 04:32:38 +00001411</div>
1412
1413<!-- _______________________________________________________________________ -->
1414<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1415
1416<div class="doc_text">
Bill Wendling6ec40612009-07-20 02:39:26 +00001417
Chris Lattner488772f2008-01-04 04:32:38 +00001418<h5>Overview:</h5>
1419<p>The label type represents code labels.</p>
1420
1421<h5>Syntax:</h5>
Chris Lattner488772f2008-01-04 04:32:38 +00001422<pre>
1423 label
1424</pre>
Bill Wendling6ec40612009-07-20 02:39:26 +00001425
Chris Lattner488772f2008-01-04 04:32:38 +00001426</div>
1427
Nick Lewycky29aaef82009-05-30 05:06:04 +00001428<!-- _______________________________________________________________________ -->
1429<div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1430
1431<div class="doc_text">
Bill Wendling6ec40612009-07-20 02:39:26 +00001432
Nick Lewycky29aaef82009-05-30 05:06:04 +00001433<h5>Overview:</h5>
1434<p>The metadata type represents embedded metadata. The only derived type that
Bill Wendlingf85859d2009-07-20 02:29:24 +00001435 may contain metadata is <tt>metadata*</tt> or a function type that returns or
1436 takes metadata typed parameters, but not pointer to metadata types.</p>
Nick Lewycky29aaef82009-05-30 05:06:04 +00001437
1438<h5>Syntax:</h5>
Nick Lewycky29aaef82009-05-30 05:06:04 +00001439<pre>
1440 metadata
1441</pre>
Bill Wendling6ec40612009-07-20 02:39:26 +00001442
Nick Lewycky29aaef82009-05-30 05:06:04 +00001443</div>
1444
Chris Lattner488772f2008-01-04 04:32:38 +00001445
1446<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001447<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1448
1449<div class="doc_text">
1450
Bill Wendlingf85859d2009-07-20 02:29:24 +00001451<p>The real power in LLVM comes from the derived types in the system. This is
1452 what allows a programmer to represent arrays, functions, pointers, and other
1453 useful types. Note that these derived types may be recursive: For example,
1454 it is possible to have a two dimensional array.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001455
1456</div>
1457
1458<!-- _______________________________________________________________________ -->
1459<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1460
1461<div class="doc_text">
1462
1463<h5>Overview:</h5>
1464<p>The integer type is a very simple derived type that simply specifies an
Bill Wendlingf85859d2009-07-20 02:29:24 +00001465 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1466 2^23-1 (about 8 million) can be specified.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001467
1468<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001469<pre>
1470 iN
1471</pre>
1472
1473<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001474 value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001475
1476<h5>Examples:</h5>
1477<table class="layout">
Nick Lewycky39382d62009-05-24 02:46:06 +00001478 <tr class="layout">
1479 <td class="left"><tt>i1</tt></td>
1480 <td class="left">a single-bit integer.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001481 </tr>
Nick Lewycky39382d62009-05-24 02:46:06 +00001482 <tr class="layout">
1483 <td class="left"><tt>i32</tt></td>
1484 <td class="left">a 32-bit integer.</td>
1485 </tr>
1486 <tr class="layout">
1487 <td class="left"><tt>i1942652</tt></td>
1488 <td class="left">a really big integer of over 1 million bits.</td>
1489 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001490</table>
djge93155c2009-01-24 15:58:40 +00001491
Bill Wendlingf85859d2009-07-20 02:29:24 +00001492<p>Note that the code generator does not yet support large integer types to be
1493 used as function return types. The specific limit on how large a return type
1494 the code generator can currently handle is target-dependent; currently it's
1495 often 64 bits for 32-bit targets and 128 bits for 64-bit targets.</p>
djge93155c2009-01-24 15:58:40 +00001496
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001497</div>
1498
1499<!-- _______________________________________________________________________ -->
1500<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1501
1502<div class="doc_text">
1503
1504<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001505<p>The array type is a very simple derived type that arranges elements
Bill Wendlingf85859d2009-07-20 02:29:24 +00001506 sequentially in memory. The array type requires a size (number of elements)
1507 and an underlying data type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001508
1509<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001510<pre>
1511 [&lt;# elements&gt; x &lt;elementtype&gt;]
1512</pre>
1513
Bill Wendlingf85859d2009-07-20 02:29:24 +00001514<p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1515 be any type with a size.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001516
1517<h5>Examples:</h5>
1518<table class="layout">
1519 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +00001520 <td class="left"><tt>[40 x i32]</tt></td>
1521 <td class="left">Array of 40 32-bit integer values.</td>
1522 </tr>
1523 <tr class="layout">
1524 <td class="left"><tt>[41 x i32]</tt></td>
1525 <td class="left">Array of 41 32-bit integer values.</td>
1526 </tr>
1527 <tr class="layout">
1528 <td class="left"><tt>[4 x i8]</tt></td>
1529 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001530 </tr>
1531</table>
1532<p>Here are some examples of multidimensional arrays:</p>
1533<table class="layout">
1534 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +00001535 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1536 <td class="left">3x4 array of 32-bit integer values.</td>
1537 </tr>
1538 <tr class="layout">
1539 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1540 <td class="left">12x10 array of single precision floating point values.</td>
1541 </tr>
1542 <tr class="layout">
1543 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1544 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001545 </tr>
1546</table>
1547
Bill Wendlingf85859d2009-07-20 02:29:24 +00001548<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1549 length array. Normally, accesses past the end of an array are undefined in
1550 LLVM (e.g. it is illegal to access the 5th element of a 3 element array). As
1551 a special case, however, zero length arrays are recognized to be variable
1552 length. This allows implementation of 'pascal style arrays' with the LLVM
1553 type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001554
Bill Wendlingf85859d2009-07-20 02:29:24 +00001555<p>Note that the code generator does not yet support large aggregate types to be
1556 used as function return types. The specific limit on how large an aggregate
1557 return type the code generator can currently handle is target-dependent, and
1558 also dependent on the aggregate element types.</p>
djge93155c2009-01-24 15:58:40 +00001559
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001560</div>
1561
1562<!-- _______________________________________________________________________ -->
1563<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001564
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001565<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +00001566
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001567<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001568<p>The function type can be thought of as a function signature. It consists of
1569 a return type and a list of formal parameter types. The return type of a
1570 function type is a scalar type, a void type, or a struct type. If the return
1571 type is a struct type then all struct elements must be of first class types,
1572 and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +00001573
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001574<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +00001575<pre>
1576 &lt;returntype list&gt; (&lt;parameter list&gt;)
1577</pre>
1578
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001579<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
Bill Wendlingf85859d2009-07-20 02:29:24 +00001580 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1581 which indicates that the function takes a variable number of arguments.
1582 Variable argument functions can access their arguments with
1583 the <a href="#int_varargs">variable argument handling intrinsic</a>
1584 functions. '<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1585 <a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +00001586
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001587<h5>Examples:</h5>
1588<table class="layout">
1589 <tr class="layout">
1590 <td class="left"><tt>i32 (i32)</tt></td>
1591 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1592 </td>
1593 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +00001594 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001595 </tt></td>
1596 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1597 an <tt>i16</tt> that should be sign extended and a
1598 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1599 <tt>float</tt>.
1600 </td>
1601 </tr><tr class="layout">
1602 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1603 <td class="left">A vararg function that takes at least one
1604 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1605 which returns an integer. This is the signature for <tt>printf</tt> in
1606 LLVM.
1607 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +00001608 </tr><tr class="layout">
1609 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Misha Brukmanafc88b02008-11-27 06:41:20 +00001610 <td class="left">A function taking an <tt>i32</tt>, returning two
1611 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +00001612 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001613 </tr>
1614</table>
1615
1616</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001617
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001618<!-- _______________________________________________________________________ -->
1619<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001620
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001621<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001622
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001623<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001624<p>The structure type is used to represent a collection of data members together
1625 in memory. The packing of the field types is defined to match the ABI of the
1626 underlying processor. The elements of a structure may be any type that has a
1627 size.</p>
1628
1629<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1630 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1631 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1632
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001633<h5>Syntax:</h5>
Bill Wendling6ec40612009-07-20 02:39:26 +00001634<pre>
1635 { &lt;type list&gt; }
1636</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001637
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001638<h5>Examples:</h5>
1639<table class="layout">
1640 <tr class="layout">
1641 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1642 <td class="left">A triple of three <tt>i32</tt> values</td>
1643 </tr><tr class="layout">
1644 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1645 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1646 second element is a <a href="#t_pointer">pointer</a> to a
1647 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1648 an <tt>i32</tt>.</td>
1649 </tr>
1650</table>
djge93155c2009-01-24 15:58:40 +00001651
Bill Wendlingf85859d2009-07-20 02:29:24 +00001652<p>Note that the code generator does not yet support large aggregate types to be
1653 used as function return types. The specific limit on how large an aggregate
1654 return type the code generator can currently handle is target-dependent, and
1655 also dependent on the aggregate element types.</p>
djge93155c2009-01-24 15:58:40 +00001656
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001657</div>
1658
1659<!-- _______________________________________________________________________ -->
1660<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1661</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001662
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001663<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001664
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001665<h5>Overview:</h5>
1666<p>The packed structure type is used to represent a collection of data members
Bill Wendlingf85859d2009-07-20 02:29:24 +00001667 together in memory. There is no padding between fields. Further, the
1668 alignment of a packed structure is 1 byte. The elements of a packed
1669 structure may be any type that has a size.</p>
1670
1671<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1672 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1673 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1674
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001675<h5>Syntax:</h5>
Bill Wendling6ec40612009-07-20 02:39:26 +00001676<pre>
1677 &lt; { &lt;type list&gt; } &gt;
1678</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001679
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001680<h5>Examples:</h5>
1681<table class="layout">
1682 <tr class="layout">
1683 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1684 <td class="left">A triple of three <tt>i32</tt> values</td>
1685 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +00001686 <td class="left">
1687<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001688 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1689 second element is a <a href="#t_pointer">pointer</a> to a
1690 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1691 an <tt>i32</tt>.</td>
1692 </tr>
1693</table>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001694
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001695</div>
1696
1697<!-- _______________________________________________________________________ -->
1698<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
Chris Lattner96edbd32009-02-08 19:53:29 +00001699
Bill Wendlingf85859d2009-07-20 02:29:24 +00001700<div class="doc_text">
1701
1702<h5>Overview:</h5>
1703<p>As in many languages, the pointer type represents a pointer or reference to
1704 another object, which must live in memory. Pointer types may have an optional
1705 address space attribute defining the target-specific numbered address space
1706 where the pointed-to object resides. The default address space is zero.</p>
1707
1708<p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1709 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +00001710
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001711<h5>Syntax:</h5>
Bill Wendling6ec40612009-07-20 02:39:26 +00001712<pre>
1713 &lt;type&gt; *
1714</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001715
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001716<h5>Examples:</h5>
1717<table class="layout">
1718 <tr class="layout">
Dan Gohman01852382009-01-04 23:44:43 +00001719 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner7311d222007-12-19 05:04:11 +00001720 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1721 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1722 </tr>
1723 <tr class="layout">
1724 <td class="left"><tt>i32 (i32 *) *</tt></td>
1725 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001726 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +00001727 <tt>i32</tt>.</td>
1728 </tr>
1729 <tr class="layout">
1730 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1731 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1732 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001733 </tr>
1734</table>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001735
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001736</div>
1737
1738<!-- _______________________________________________________________________ -->
1739<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001740
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001741<div class="doc_text">
1742
1743<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001744<p>A vector type is a simple derived type that represents a vector of elements.
1745 Vector types are used when multiple primitive data are operated in parallel
1746 using a single instruction (SIMD). A vector type requires a size (number of
1747 elements) and an underlying primitive data type. Vectors must have a power
1748 of two length (1, 2, 4, 8, 16 ...). Vector types are considered
1749 <a href="#t_firstclass">first class</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001750
1751<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001752<pre>
1753 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1754</pre>
1755
Bill Wendlingf85859d2009-07-20 02:29:24 +00001756<p>The number of elements is a constant integer value; elementtype may be any
1757 integer or floating point type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001758
1759<h5>Examples:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001760<table class="layout">
1761 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +00001762 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1763 <td class="left">Vector of 4 32-bit integer values.</td>
1764 </tr>
1765 <tr class="layout">
1766 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1767 <td class="left">Vector of 8 32-bit floating-point values.</td>
1768 </tr>
1769 <tr class="layout">
1770 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1771 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001772 </tr>
1773</table>
djge93155c2009-01-24 15:58:40 +00001774
Bill Wendlingf85859d2009-07-20 02:29:24 +00001775<p>Note that the code generator does not yet support large vector types to be
1776 used as function return types. The specific limit on how large a vector
1777 return type codegen can currently handle is target-dependent; currently it's
1778 often a few times longer than a hardware vector register.</p>
djge93155c2009-01-24 15:58:40 +00001779
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001780</div>
1781
1782<!-- _______________________________________________________________________ -->
1783<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1784<div class="doc_text">
1785
1786<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001787<p>Opaque types are used to represent unknown types in the system. This
Bill Wendlingf85859d2009-07-20 02:29:24 +00001788 corresponds (for example) to the C notion of a forward declared structure
1789 type. In LLVM, opaque types can eventually be resolved to any type (not just
1790 a structure type).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001791
1792<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001793<pre>
1794 opaque
1795</pre>
1796
1797<h5>Examples:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001798<table class="layout">
1799 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +00001800 <td class="left"><tt>opaque</tt></td>
1801 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001802 </tr>
1803</table>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001804
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001805</div>
1806
Chris Lattner515195a2009-02-02 07:32:36 +00001807<!-- ======================================================================= -->
1808<div class="doc_subsection">
1809 <a name="t_uprefs">Type Up-references</a>
1810</div>
1811
1812<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001813
Chris Lattner515195a2009-02-02 07:32:36 +00001814<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001815<p>An "up reference" allows you to refer to a lexically enclosing type without
1816 requiring it to have a name. For instance, a structure declaration may
1817 contain a pointer to any of the types it is lexically a member of. Example
1818 of up references (with their equivalent as named type declarations)
1819 include:</p>
Chris Lattner515195a2009-02-02 07:32:36 +00001820
1821<pre>
Chris Lattner5ad632d2009-02-09 10:00:56 +00001822 { \2 * } %x = type { %x* }
Chris Lattner515195a2009-02-02 07:32:36 +00001823 { \2 }* %y = type { %y }*
1824 \1* %z = type %z*
1825</pre>
1826
Bill Wendlingf85859d2009-07-20 02:29:24 +00001827<p>An up reference is needed by the asmprinter for printing out cyclic types
1828 when there is no declared name for a type in the cycle. Because the
1829 asmprinter does not want to print out an infinite type string, it needs a
1830 syntax to handle recursive types that have no names (all names are optional
1831 in llvm IR).</p>
Chris Lattner515195a2009-02-02 07:32:36 +00001832
1833<h5>Syntax:</h5>
1834<pre>
1835 \&lt;level&gt;
1836</pre>
1837
Bill Wendlingf85859d2009-07-20 02:29:24 +00001838<p>The level is the count of the lexical type that is being referred to.</p>
Chris Lattner515195a2009-02-02 07:32:36 +00001839
1840<h5>Examples:</h5>
Chris Lattner515195a2009-02-02 07:32:36 +00001841<table class="layout">
1842 <tr class="layout">
1843 <td class="left"><tt>\1*</tt></td>
1844 <td class="left">Self-referential pointer.</td>
1845 </tr>
1846 <tr class="layout">
1847 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1848 <td class="left">Recursive structure where the upref refers to the out-most
1849 structure.</td>
1850 </tr>
1851</table>
Chris Lattner515195a2009-02-02 07:32:36 +00001852
Bill Wendlingf85859d2009-07-20 02:29:24 +00001853</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001854
1855<!-- *********************************************************************** -->
1856<div class="doc_section"> <a name="constants">Constants</a> </div>
1857<!-- *********************************************************************** -->
1858
1859<div class="doc_text">
1860
1861<p>LLVM has several different basic types of constants. This section describes
Bill Wendlingf85859d2009-07-20 02:29:24 +00001862 them all and their syntax.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001863
1864</div>
1865
1866<!-- ======================================================================= -->
1867<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1868
1869<div class="doc_text">
1870
1871<dl>
1872 <dt><b>Boolean constants</b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001873 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
Bill Wendlingf85859d2009-07-20 02:29:24 +00001874 constants of the <tt><a href="#t_primitive">i1</a></tt> type.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001875
1876 <dt><b>Integer constants</b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00001877 <dd>Standard integers (such as '4') are constants of
1878 the <a href="#t_integer">integer</a> type. Negative numbers may be used
1879 with integer types.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001880
1881 <dt><b>Floating point constants</b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001882 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
Bill Wendlingf85859d2009-07-20 02:29:24 +00001883 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1884 notation (see below). The assembler requires the exact decimal value of a
1885 floating-point constant. For example, the assembler accepts 1.25 but
1886 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1887 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001888
1889 <dt><b>Null pointer constants</b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001890 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
Bill Wendlingf85859d2009-07-20 02:29:24 +00001891 and must be of <a href="#t_pointer">pointer type</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001892</dl>
1893
Bill Wendlingf85859d2009-07-20 02:29:24 +00001894<p>The one non-intuitive notation for constants is the hexadecimal form of
1895 floating point constants. For example, the form '<tt>double
1896 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
1897 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
1898 constants are required (and the only time that they are generated by the
1899 disassembler) is when a floating point constant must be emitted but it cannot
1900 be represented as a decimal floating point number in a reasonable number of
1901 digits. For example, NaN's, infinities, and other special values are
1902 represented in their IEEE hexadecimal format so that assembly and disassembly
1903 do not cause any bits to change in the constants.</p>
1904
Dale Johannesenf82a52f2009-02-11 22:14:51 +00001905<p>When using the hexadecimal form, constants of types float and double are
Bill Wendlingf85859d2009-07-20 02:29:24 +00001906 represented using the 16-digit form shown above (which matches the IEEE754
1907 representation for double); float values must, however, be exactly
1908 representable as IEE754 single precision. Hexadecimal format is always used
1909 for long double, and there are three forms of long double. The 80-bit format
1910 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
1911 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1912 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
1913 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
1914 currently supported target uses this format. Long doubles will only work if
1915 they match the long double format on your target. All hexadecimal formats
1916 are big-endian (sign bit at the left).</p>
1917
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001918</div>
1919
1920<!-- ======================================================================= -->
Chris Lattner97063852009-02-28 18:32:25 +00001921<div class="doc_subsection">
Bill Wendling1a2630a2009-07-20 02:32:41 +00001922<a name="aggregateconstants"></a> <!-- old anchor -->
1923<a name="complexconstants">Complex Constants</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001924</div>
1925
1926<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00001927
Chris Lattner97063852009-02-28 18:32:25 +00001928<p>Complex constants are a (potentially recursive) combination of simple
Bill Wendlingf85859d2009-07-20 02:29:24 +00001929 constants and smaller complex constants.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001930
1931<dl>
1932 <dt><b>Structure constants</b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001933 <dd>Structure constants are represented with notation similar to structure
Bill Wendlingf85859d2009-07-20 02:29:24 +00001934 type definitions (a comma separated list of elements, surrounded by braces
1935 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1936 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
1937 Structure constants must have <a href="#t_struct">structure type</a>, and
1938 the number and types of elements must match those specified by the
1939 type.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001940
1941 <dt><b>Array constants</b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001942 <dd>Array constants are represented with notation similar to array type
Bill Wendlingf85859d2009-07-20 02:29:24 +00001943 definitions (a comma separated list of elements, surrounded by square
1944 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
1945 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
1946 the number and types of elements must match those specified by the
1947 type.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001948
1949 <dt><b>Vector constants</b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001950 <dd>Vector constants are represented with notation similar to vector type
Bill Wendlingf85859d2009-07-20 02:29:24 +00001951 definitions (a comma separated list of elements, surrounded by
1952 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32
1953 42, i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must
1954 have <a href="#t_vector">vector type</a>, and the number and types of
1955 elements must match those specified by the type.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001956
1957 <dt><b>Zero initialization</b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001958 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
Bill Wendlingf85859d2009-07-20 02:29:24 +00001959 value to zero of <em>any</em> type, including scalar and aggregate types.
1960 This is often used to avoid having to print large zero initializers
1961 (e.g. for large arrays) and is always exactly equivalent to using explicit
1962 zero initializers.</dd>
Nick Lewycky4dcf8102009-04-04 07:22:01 +00001963
1964 <dt><b>Metadata node</b></dt>
Nick Lewyckyf122c7e2009-05-30 16:08:30 +00001965 <dd>A metadata node is a structure-like constant with
Bill Wendlingf85859d2009-07-20 02:29:24 +00001966 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
1967 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
1968 be interpreted as part of the instruction stream, metadata is a place to
1969 attach additional information such as debug info.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001970</dl>
1971
1972</div>
1973
1974<!-- ======================================================================= -->
1975<div class="doc_subsection">
1976 <a name="globalconstants">Global Variable and Function Addresses</a>
1977</div>
1978
1979<div class="doc_text">
1980
Bill Wendlingf85859d2009-07-20 02:29:24 +00001981<p>The addresses of <a href="#globalvars">global variables</a>
1982 and <a href="#functionstructure">functions</a> are always implicitly valid
1983 (link-time) constants. These constants are explicitly referenced when
1984 the <a href="#identifiers">identifier for the global</a> is used and always
1985 have <a href="#t_pointer">pointer</a> type. For example, the following is a
1986 legal LLVM file:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00001987
1988<div class="doc_code">
1989<pre>
1990@X = global i32 17
1991@Y = global i32 42
1992@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1993</pre>
1994</div>
1995
1996</div>
1997
1998<!-- ======================================================================= -->
1999<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2000<div class="doc_text">
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002001
Bill Wendlingf85859d2009-07-20 02:29:24 +00002002<p>The string '<tt>undef</tt>' is recognized as a type-less constant that has no
2003 specific value. Undefined values may be of any type and be used anywhere a
2004 constant is permitted.</p>
2005
2006<p>Undefined values indicate to the compiler that the program is well defined no
2007 matter what value is used, giving the compiler more freedom to optimize.</p>
2008
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002009</div>
2010
2011<!-- ======================================================================= -->
2012<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2013</div>
2014
2015<div class="doc_text">
2016
2017<p>Constant expressions are used to allow expressions involving other constants
Bill Wendlingf85859d2009-07-20 02:29:24 +00002018 to be used as constants. Constant expressions may be of
2019 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2020 operation that does not have side effects (e.g. load and call are not
2021 supported). The following is the syntax for constant expressions:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002022
2023<dl>
2024 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002025 <dd>Truncate a constant to another type. The bit size of CST must be larger
2026 than the bit size of TYPE. Both types must be integers.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002027
2028 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002029 <dd>Zero extend a constant to another type. The bit size of CST must be
2030 smaller or equal to the bit size of TYPE. Both types must be
2031 integers.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002032
2033 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002034 <dd>Sign extend a constant to another type. The bit size of CST must be
2035 smaller or equal to the bit size of TYPE. Both types must be
2036 integers.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002037
2038 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002039 <dd>Truncate a floating point constant to another floating point type. The
2040 size of CST must be larger than the size of TYPE. Both types must be
2041 floating point.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002042
2043 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002044 <dd>Floating point extend a constant to another type. The size of CST must be
2045 smaller or equal to the size of TYPE. Both types must be floating
2046 point.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002047
Reid Spencere6adee82007-07-31 14:40:14 +00002048 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002049 <dd>Convert a floating point constant to the corresponding unsigned integer
Bill Wendlingf85859d2009-07-20 02:29:24 +00002050 constant. TYPE must be a scalar or vector integer type. CST must be of
2051 scalar or vector floating point type. Both CST and TYPE must be scalars,
2052 or vectors of the same number of elements. If the value won't fit in the
2053 integer type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002054
2055 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
2056 <dd>Convert a floating point constant to the corresponding signed integer
Bill Wendlingf85859d2009-07-20 02:29:24 +00002057 constant. TYPE must be a scalar or vector integer type. CST must be of
2058 scalar or vector floating point type. Both CST and TYPE must be scalars,
2059 or vectors of the same number of elements. If the value won't fit in the
2060 integer type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002061
2062 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
2063 <dd>Convert an unsigned integer constant to the corresponding floating point
Bill Wendlingf85859d2009-07-20 02:29:24 +00002064 constant. TYPE must be a scalar or vector floating point type. CST must be
2065 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2066 vectors of the same number of elements. If the value won't fit in the
2067 floating point type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002068
2069 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
2070 <dd>Convert a signed integer constant to the corresponding floating point
Bill Wendlingf85859d2009-07-20 02:29:24 +00002071 constant. TYPE must be a scalar or vector floating point type. CST must be
2072 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2073 vectors of the same number of elements. If the value won't fit in the
2074 floating point type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002075
2076 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
2077 <dd>Convert a pointer typed constant to the corresponding integer constant
Bill Wendlingf85859d2009-07-20 02:29:24 +00002078 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2079 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2080 make it fit in <tt>TYPE</tt>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002081
2082 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002083 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2084 type. CST must be of integer type. The CST value is zero extended,
2085 truncated, or unchanged to make it fit in a pointer size. This one is
2086 <i>really</i> dangerous!</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002087
2088 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
Chris Lattner557bc5d2009-02-28 18:27:03 +00002089 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2090 are the same as those for the <a href="#i_bitcast">bitcast
2091 instruction</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002092
2093 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002094 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
Bill Wendlingf85859d2009-07-20 02:29:24 +00002095 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2096 instruction, the index list may have zero or more indexes, which are
2097 required to make sense for the type of "CSTPTR".</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002098
2099 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002100 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002101
2102 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2103 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2104
2105 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2106 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2107
2108 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002109 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2110 constants.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002111
2112 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002113 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2114 constants.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002115
2116 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002117 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2118 constants.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002119
2120 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002121 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2122 be any of the <a href="#binaryops">binary</a>
2123 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2124 on operands are the same as those for the corresponding instruction
2125 (e.g. no bitwise operations on floating point values are allowed).</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002126</dl>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002127
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002128</div>
2129
Nick Lewycky4dcf8102009-04-04 07:22:01 +00002130<!-- ======================================================================= -->
2131<div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2132</div>
2133
2134<div class="doc_text">
2135
Bill Wendlingf85859d2009-07-20 02:29:24 +00002136<p>Embedded metadata provides a way to attach arbitrary data to the instruction
2137 stream without affecting the behaviour of the program. There are two
2138 metadata primitives, strings and nodes. All metadata has the
2139 <tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2140 point ('<tt>!</tt>').</p>
Nick Lewycky4dcf8102009-04-04 07:22:01 +00002141
2142<p>A metadata string is a string surrounded by double quotes. It can contain
Bill Wendlingf85859d2009-07-20 02:29:24 +00002143 any character by escaping non-printable characters with "\xx" where "xx" is
2144 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
Nick Lewycky4dcf8102009-04-04 07:22:01 +00002145
2146<p>Metadata nodes are represented with notation similar to structure constants
Bill Wendlingf85859d2009-07-20 02:29:24 +00002147 (a comma separated list of elements, surrounded by braces and preceeded by an
2148 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2149 10}</tt>".</p>
Nick Lewycky4dcf8102009-04-04 07:22:01 +00002150
Bill Wendlingf85859d2009-07-20 02:29:24 +00002151<p>A metadata node will attempt to track changes to the values it holds. In the
2152 event that a value is deleted, it will be replaced with a typeless
2153 "<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
Nick Lewycky117f4382009-05-10 20:57:05 +00002154
Nick Lewycky4dcf8102009-04-04 07:22:01 +00002155<p>Optimizations may rely on metadata to provide additional information about
Bill Wendlingf85859d2009-07-20 02:29:24 +00002156 the program that isn't available in the instructions, or that isn't easily
2157 computable. Similarly, the code generator may expect a certain metadata
2158 format to be used to express debugging information.</p>
2159
Nick Lewycky4dcf8102009-04-04 07:22:01 +00002160</div>
2161
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002162<!-- *********************************************************************** -->
2163<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2164<!-- *********************************************************************** -->
2165
2166<!-- ======================================================================= -->
2167<div class="doc_subsection">
2168<a name="inlineasm">Inline Assembler Expressions</a>
2169</div>
2170
2171<div class="doc_text">
2172
Bill Wendlingf85859d2009-07-20 02:29:24 +00002173<p>LLVM supports inline assembler expressions (as opposed
2174 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2175 a special value. This value represents the inline assembler as a string
2176 (containing the instructions to emit), a list of operand constraints (stored
2177 as a string), and a flag that indicates whether or not the inline asm
2178 expression has side effects. An example inline assembler expression is:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002179
2180<div class="doc_code">
2181<pre>
2182i32 (i32) asm "bswap $0", "=r,r"
2183</pre>
2184</div>
2185
Bill Wendlingf85859d2009-07-20 02:29:24 +00002186<p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2187 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2188 have:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002189
2190<div class="doc_code">
2191<pre>
2192%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2193</pre>
2194</div>
2195
Bill Wendlingf85859d2009-07-20 02:29:24 +00002196<p>Inline asms with side effects not visible in the constraint list must be
2197 marked as having side effects. This is done through the use of the
2198 '<tt>sideeffect</tt>' keyword, like so:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002199
2200<div class="doc_code">
2201<pre>
2202call void asm sideeffect "eieio", ""()
2203</pre>
2204</div>
2205
2206<p>TODO: The format of the asm and constraints string still need to be
Bill Wendlingf85859d2009-07-20 02:29:24 +00002207 documented here. Constraints on what can be done (e.g. duplication, moving,
2208 etc need to be documented). This is probably best done by reference to
2209 another document that covers inline asm from a holistic perspective.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002210
2211</div>
2212
Chris Lattner75c24e02009-07-20 05:55:19 +00002213
2214<!-- *********************************************************************** -->
2215<div class="doc_section">
2216 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2217</div>
2218<!-- *********************************************************************** -->
2219
2220<p>LLVM has a number of "magic" global variables that contain data that affect
2221code generation or other IR semantics. These are documented here. All globals
Chris Lattner1e0e0d12009-07-20 06:14:25 +00002222of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2223section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2224by LLVM.</p>
Chris Lattner75c24e02009-07-20 05:55:19 +00002225
2226<!-- ======================================================================= -->
2227<div class="doc_subsection">
2228<a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2229</div>
2230
2231<div class="doc_text">
2232
2233<p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2234href="#linkage_appending">appending linkage</a>. This array contains a list of
2235pointers to global variables and functions which may optionally have a pointer
2236cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2237
2238<pre>
2239 @X = global i8 4
2240 @Y = global i32 123
2241
2242 @llvm.used = appending global [2 x i8*] [
2243 i8* @X,
2244 i8* bitcast (i32* @Y to i8*)
2245 ], section "llvm.metadata"
2246</pre>
2247
2248<p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2249compiler, assembler, and linker are required to treat the symbol as if there is
2250a reference to the global that it cannot see. For example, if a variable has
2251internal linkage and no references other than that from the <tt>@llvm.used</tt>
2252list, it cannot be deleted. This is commonly used to represent references from
2253inline asms and other things the compiler cannot "see", and corresponds to
2254"attribute((used))" in GNU C.</p>
2255
2256<p>On some targets, the code generator must emit a directive to the assembler or
2257object file to prevent the assembler and linker from molesting the symbol.</p>
2258
2259</div>
2260
2261<!-- ======================================================================= -->
2262<div class="doc_subsection">
Chris Lattner1e0e0d12009-07-20 06:14:25 +00002263<a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2264</div>
2265
2266<div class="doc_text">
2267
2268<p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2269<tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2270touching the symbol. On targets that support it, this allows an intelligent
2271linker to optimize references to the symbol without being impeded as it would be
2272by <tt>@llvm.used</tt>.</p>
2273
2274<p>This is a rare construct that should only be used in rare circumstances, and
2275should not be exposed to source languages.</p>
2276
2277</div>
2278
2279<!-- ======================================================================= -->
2280<div class="doc_subsection">
Chris Lattner75c24e02009-07-20 05:55:19 +00002281<a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2282</div>
2283
2284<div class="doc_text">
2285
2286<p>TODO: Describe this.</p>
2287
2288</div>
2289
2290<!-- ======================================================================= -->
2291<div class="doc_subsection">
2292<a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2293</div>
2294
2295<div class="doc_text">
2296
2297<p>TODO: Describe this.</p>
2298
2299</div>
2300
2301
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002302<!-- *********************************************************************** -->
2303<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2304<!-- *********************************************************************** -->
2305
2306<div class="doc_text">
2307
Bill Wendlingf85859d2009-07-20 02:29:24 +00002308<p>The LLVM instruction set consists of several different classifications of
2309 instructions: <a href="#terminators">terminator
2310 instructions</a>, <a href="#binaryops">binary instructions</a>,
2311 <a href="#bitwiseops">bitwise binary instructions</a>,
2312 <a href="#memoryops">memory instructions</a>, and
2313 <a href="#otherops">other instructions</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002314
2315</div>
2316
2317<!-- ======================================================================= -->
2318<div class="doc_subsection"> <a name="terminators">Terminator
2319Instructions</a> </div>
2320
2321<div class="doc_text">
2322
Bill Wendlingf85859d2009-07-20 02:29:24 +00002323<p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2324 in a program ends with a "Terminator" instruction, which indicates which
2325 block should be executed after the current block is finished. These
2326 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2327 control flow, not values (the one exception being the
2328 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2329
2330<p>There are six different terminator instructions: the
2331 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2332 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2333 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2334 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2335 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2336 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002337
2338</div>
2339
2340<!-- _______________________________________________________________________ -->
2341<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2342Instruction</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002343
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002344<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00002345
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002346<h5>Syntax:</h5>
Dan Gohman3e700032008-10-04 19:00:07 +00002347<pre>
2348 ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002349 ret void <i>; Return from void function</i>
2350</pre>
Chris Lattner43030e72008-04-23 04:59:35 +00002351
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002352<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002353<p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2354 a value) from a function back to the caller.</p>
2355
2356<p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2357 value and then causes control flow, and one that just causes control flow to
2358 occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +00002359
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002360<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002361<p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2362 return value. The type of the return value must be a
2363 '<a href="#t_firstclass">first class</a>' type.</p>
Dan Gohman3e700032008-10-04 19:00:07 +00002364
Bill Wendlingf85859d2009-07-20 02:29:24 +00002365<p>A function is not <a href="#wellformed">well formed</a> if it it has a
2366 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2367 value or a return value with a type that does not match its type, or if it
2368 has a void return type and contains a '<tt>ret</tt>' instruction with a
2369 return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +00002370
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002371<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002372<p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2373 the calling function's context. If the caller is a
2374 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2375 instruction after the call. If the caller was an
2376 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2377 the beginning of the "normal" destination block. If the instruction returns
2378 a value, that value shall set the call or invoke instruction's return
2379 value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +00002380
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002381<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +00002382<pre>
2383 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002384 ret void <i>; Return from a void function</i>
Bill Wendlingd163e2d2009-02-28 22:12:54 +00002385 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002386</pre>
Dan Gohman60967192009-01-12 23:12:39 +00002387
djge93155c2009-01-24 15:58:40 +00002388<p>Note that the code generator does not yet fully support large
2389 return values. The specific sizes that are currently supported are
2390 dependent on the target. For integers, on 32-bit targets the limit
2391 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2392 For aggregate types, the current limits are dependent on the element
2393 types; for example targets are often limited to 2 total integer
2394 elements and 2 total floating-point elements.</p>
Dan Gohman60967192009-01-12 23:12:39 +00002395
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002396</div>
2397<!-- _______________________________________________________________________ -->
2398<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002399
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002400<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00002401
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002402<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002403<pre>
2404 br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002405</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002406
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002407<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002408<p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2409 different basic block in the current function. There are two forms of this
2410 instruction, corresponding to a conditional branch and an unconditional
2411 branch.</p>
2412
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002413<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002414<p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2415 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2416 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2417 target.</p>
2418
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002419<h5>Semantics:</h5>
2420<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
Bill Wendlingf85859d2009-07-20 02:29:24 +00002421 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2422 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2423 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2424
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002425<h5>Example:</h5>
Bill Wendling6ec40612009-07-20 02:39:26 +00002426<pre>
2427Test:
2428 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2429 br i1 %cond, label %IfEqual, label %IfUnequal
2430IfEqual:
2431 <a href="#i_ret">ret</a> i32 1
2432IfUnequal:
2433 <a href="#i_ret">ret</a> i32 0
2434</pre>
2435
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002436</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002437
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002438<!-- _______________________________________________________________________ -->
2439<div class="doc_subsubsection">
2440 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2441</div>
2442
2443<div class="doc_text">
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002444
Bill Wendlingf85859d2009-07-20 02:29:24 +00002445<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002446<pre>
2447 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
2448</pre>
2449
2450<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002451<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
Bill Wendlingf85859d2009-07-20 02:29:24 +00002452 several different places. It is a generalization of the '<tt>br</tt>'
2453 instruction, allowing a branch to occur to one of many possible
2454 destinations.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002455
2456<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002457<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
Bill Wendlingf85859d2009-07-20 02:29:24 +00002458 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2459 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2460 The table is not allowed to contain duplicate constant entries.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002461
2462<h5>Semantics:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002463<p>The <tt>switch</tt> instruction specifies a table of values and
Bill Wendlingf85859d2009-07-20 02:29:24 +00002464 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2465 is searched for the given value. If the value is found, control flow is
2466 transfered to the corresponding destination; otherwise, control flow is
2467 transfered to the default destination.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002468
2469<h5>Implementation:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002470<p>Depending on properties of the target machine and the particular
Bill Wendlingf85859d2009-07-20 02:29:24 +00002471 <tt>switch</tt> instruction, this instruction may be code generated in
2472 different ways. For example, it could be generated as a series of chained
2473 conditional branches or with a lookup table.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002474
2475<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002476<pre>
2477 <i>; Emulate a conditional br instruction</i>
2478 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman01852382009-01-04 23:44:43 +00002479 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002480
2481 <i>; Emulate an unconditional br instruction</i>
2482 switch i32 0, label %dest [ ]
2483
2484 <i>; Implement a jump table:</i>
Dan Gohman01852382009-01-04 23:44:43 +00002485 switch i32 %val, label %otherwise [ i32 0, label %onzero
2486 i32 1, label %onone
2487 i32 2, label %ontwo ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002488</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002489
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002490</div>
2491
2492<!-- _______________________________________________________________________ -->
2493<div class="doc_subsubsection">
2494 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2495</div>
2496
2497<div class="doc_text">
2498
2499<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002500<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +00002501 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] &lt;ptr to function ty&gt; &lt;function ptr val&gt;(&lt;function args&gt;) [<a href="#fnattrs">fn attrs</a>]
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002502 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2503</pre>
2504
2505<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002506<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
Bill Wendlingf85859d2009-07-20 02:29:24 +00002507 function, with the possibility of control flow transfer to either the
2508 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
2509 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
2510 control flow will return to the "normal" label. If the callee (or any
2511 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
2512 instruction, control is interrupted and continued at the dynamically nearest
2513 "exception" label.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002514
2515<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002516<p>This instruction requires several arguments:</p>
2517
2518<ol>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002519 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
2520 convention</a> the call should use. If none is specified, the call
2521 defaults to using C calling conventions.</li>
Devang Patelac2fc272008-10-06 18:50:38 +00002522
2523 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
Bill Wendlingf85859d2009-07-20 02:29:24 +00002524 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
2525 '<tt>inreg</tt>' attributes are valid here.</li>
Devang Patelac2fc272008-10-06 18:50:38 +00002526
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002527 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
Bill Wendlingf85859d2009-07-20 02:29:24 +00002528 function value being invoked. In most cases, this is a direct function
2529 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
2530 off an arbitrary pointer to function value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002531
2532 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
Bill Wendlingf85859d2009-07-20 02:29:24 +00002533 function to be invoked. </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002534
2535 <li>'<tt>function args</tt>': argument list whose types match the function
Bill Wendlingf85859d2009-07-20 02:29:24 +00002536 signature argument types. If the function signature indicates the
2537 function accepts a variable number of arguments, the extra arguments can
2538 be specified.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002539
2540 <li>'<tt>normal label</tt>': the label reached when the called function
Bill Wendlingf85859d2009-07-20 02:29:24 +00002541 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002542
2543 <li>'<tt>exception label</tt>': the label reached when a callee returns with
Bill Wendlingf85859d2009-07-20 02:29:24 +00002544 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002545
Devang Pateld0bfcc72008-10-07 17:48:33 +00002546 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Bill Wendlingf85859d2009-07-20 02:29:24 +00002547 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2548 '<tt>readnone</tt>' attributes are valid here.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002549</ol>
2550
2551<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002552<p>This instruction is designed to operate as a standard
2553 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
2554 primary difference is that it establishes an association with a label, which
2555 is used by the runtime library to unwind the stack.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002556
2557<p>This instruction is used in languages with destructors to ensure that proper
Bill Wendlingf85859d2009-07-20 02:29:24 +00002558 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2559 exception. Additionally, this is important for implementation of
2560 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002561
Bill Wendlingf85859d2009-07-20 02:29:24 +00002562<p>For the purposes of the SSA form, the definition of the value returned by the
2563 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
2564 block to the "normal" label. If the callee unwinds then no return value is
2565 available.</p>
Dan Gohman140ba5d2009-05-22 21:47:08 +00002566
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002567<h5>Example:</h5>
2568<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +00002569 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002570 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +00002571 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002572 unwind label %TestCleanup <i>; {i32}:retval set</i>
2573</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002574
Bill Wendlingf85859d2009-07-20 02:29:24 +00002575</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002576
2577<!-- _______________________________________________________________________ -->
2578
2579<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2580Instruction</a> </div>
2581
2582<div class="doc_text">
2583
2584<h5>Syntax:</h5>
2585<pre>
2586 unwind
2587</pre>
2588
2589<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002590<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
Bill Wendlingf85859d2009-07-20 02:29:24 +00002591 at the first callee in the dynamic call stack which used
2592 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
2593 This is primarily used to implement exception handling.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002594
2595<h5>Semantics:</h5>
Chris Lattner8b094fc2008-04-19 21:01:16 +00002596<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Bill Wendlingf85859d2009-07-20 02:29:24 +00002597 immediately halt. The dynamic call stack is then searched for the
2598 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
2599 Once found, execution continues at the "exceptional" destination block
2600 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
2601 instruction in the dynamic call chain, undefined behavior results.</p>
2602
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002603</div>
2604
2605<!-- _______________________________________________________________________ -->
2606
2607<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2608Instruction</a> </div>
2609
2610<div class="doc_text">
2611
2612<h5>Syntax:</h5>
2613<pre>
2614 unreachable
2615</pre>
2616
2617<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002618<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
Bill Wendlingf85859d2009-07-20 02:29:24 +00002619 instruction is used to inform the optimizer that a particular portion of the
2620 code is not reachable. This can be used to indicate that the code after a
2621 no-return function cannot be reached, and other facts.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002622
2623<h5>Semantics:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002624<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002625
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002626</div>
2627
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002628<!-- ======================================================================= -->
2629<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002630
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002631<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00002632
2633<p>Binary operators are used to do most of the computation in a program. They
2634 require two operands of the same type, execute an operation on them, and
2635 produce a single value. The operands might represent multiple data, as is
2636 the case with the <a href="#t_vector">vector</a> data type. The result value
2637 has the same type as its operands.</p>
2638
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002639<p>There are several different binary operators:</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002640
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002641</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002642
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002643<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +00002644<div class="doc_subsubsection">
2645 <a name="i_add">'<tt>add</tt>' Instruction</a>
2646</div>
2647
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002648<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +00002649
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002650<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00002651<pre>
Dan Gohmaned5fcb22009-07-20 22:41:19 +00002652 &lt;result&gt; = add &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohman46e96012009-07-22 22:44:56 +00002653 &lt;result&gt; = nuw add &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2654 &lt;result&gt; = nsw add &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2655 &lt;result&gt; = nuw nsw add &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002656</pre>
Chris Lattner6704c212008-05-20 20:48:21 +00002657
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002658<h5>Overview:</h5>
2659<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002660
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002661<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002662<p>The two arguments to the '<tt>add</tt>' instruction must
2663 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2664 integer values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002665
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002666<h5>Semantics:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002667<p>The value produced is the integer sum of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002668
Bill Wendlingf85859d2009-07-20 02:29:24 +00002669<p>If the sum has unsigned overflow, the result returned is the mathematical
2670 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002671
Bill Wendlingf85859d2009-07-20 02:29:24 +00002672<p>Because LLVM integers use a two's complement representation, this instruction
2673 is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002674
Dan Gohman46e96012009-07-22 22:44:56 +00002675<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
2676 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
2677 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
2678 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
Dan Gohmaned5fcb22009-07-20 22:41:19 +00002679
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002680<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00002681<pre>
2682 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002683</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002684
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002685</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002686
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002687<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +00002688<div class="doc_subsubsection">
Dan Gohman7ce405e2009-06-04 22:49:04 +00002689 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
2690</div>
2691
2692<div class="doc_text">
2693
2694<h5>Syntax:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002695<pre>
2696 &lt;result&gt; = fadd &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2697</pre>
2698
2699<h5>Overview:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002700<p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
2701
2702<h5>Arguments:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002703<p>The two arguments to the '<tt>fadd</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00002704 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2705 floating point values. Both arguments must have identical types.</p>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002706
2707<h5>Semantics:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002708<p>The value produced is the floating point sum of the two operands.</p>
2709
2710<h5>Example:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002711<pre>
2712 &lt;result&gt; = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
2713</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002714
Dan Gohman7ce405e2009-06-04 22:49:04 +00002715</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002716
Dan Gohman7ce405e2009-06-04 22:49:04 +00002717<!-- _______________________________________________________________________ -->
2718<div class="doc_subsubsection">
Chris Lattner6704c212008-05-20 20:48:21 +00002719 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2720</div>
2721
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002722<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +00002723
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002724<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00002725<pre>
Dan Gohman46e96012009-07-22 22:44:56 +00002726 &lt;result&gt; = sub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2727 &lt;result&gt; = nuw sub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2728 &lt;result&gt; = nsw sub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2729 &lt;result&gt; = nuw nsw sub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002730</pre>
Chris Lattner6704c212008-05-20 20:48:21 +00002731
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002732<h5>Overview:</h5>
2733<p>The '<tt>sub</tt>' instruction returns the difference of its two
Bill Wendlingf85859d2009-07-20 02:29:24 +00002734 operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002735
2736<p>Note that the '<tt>sub</tt>' instruction is used to represent the
Bill Wendlingf85859d2009-07-20 02:29:24 +00002737 '<tt>neg</tt>' instruction present in most other intermediate
2738 representations.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002739
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002740<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002741<p>The two arguments to the '<tt>sub</tt>' instruction must
2742 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2743 integer values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002744
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002745<h5>Semantics:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002746<p>The value produced is the integer difference of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002747
Dan Gohman7ce405e2009-06-04 22:49:04 +00002748<p>If the difference has unsigned overflow, the result returned is the
Bill Wendlingf85859d2009-07-20 02:29:24 +00002749 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
2750 result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002751
Bill Wendlingf85859d2009-07-20 02:29:24 +00002752<p>Because LLVM integers use a two's complement representation, this instruction
2753 is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002754
Dan Gohman46e96012009-07-22 22:44:56 +00002755<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
2756 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
2757 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
2758 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
Dan Gohmaned5fcb22009-07-20 22:41:19 +00002759
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002760<h5>Example:</h5>
2761<pre>
2762 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2763 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2764</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002765
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002766</div>
Chris Lattner6704c212008-05-20 20:48:21 +00002767
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002768<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +00002769<div class="doc_subsubsection">
Dan Gohman7ce405e2009-06-04 22:49:04 +00002770 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
2771</div>
2772
2773<div class="doc_text">
2774
2775<h5>Syntax:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002776<pre>
2777 &lt;result&gt; = fsub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2778</pre>
2779
2780<h5>Overview:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002781<p>The '<tt>fsub</tt>' instruction returns the difference of its two
Bill Wendlingf85859d2009-07-20 02:29:24 +00002782 operands.</p>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002783
2784<p>Note that the '<tt>fsub</tt>' instruction is used to represent the
Bill Wendlingf85859d2009-07-20 02:29:24 +00002785 '<tt>fneg</tt>' instruction present in most other intermediate
2786 representations.</p>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002787
2788<h5>Arguments:</h5>
Bill Wendling1a2630a2009-07-20 02:32:41 +00002789<p>The two arguments to the '<tt>fsub</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00002790 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2791 floating point values. Both arguments must have identical types.</p>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002792
2793<h5>Semantics:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002794<p>The value produced is the floating point difference of the two operands.</p>
2795
2796<h5>Example:</h5>
2797<pre>
2798 &lt;result&gt; = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
2799 &lt;result&gt; = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
2800</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002801
Dan Gohman7ce405e2009-06-04 22:49:04 +00002802</div>
2803
2804<!-- _______________________________________________________________________ -->
2805<div class="doc_subsubsection">
Chris Lattner6704c212008-05-20 20:48:21 +00002806 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2807</div>
2808
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002809<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +00002810
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002811<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002812<pre>
Dan Gohman46e96012009-07-22 22:44:56 +00002813 &lt;result&gt; = mul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2814 &lt;result&gt; = nuw mul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2815 &lt;result&gt; = nsw mul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2816 &lt;result&gt; = nuw nsw mul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002817</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002818
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002819<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002820<p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002821
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002822<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002823<p>The two arguments to the '<tt>mul</tt>' instruction must
2824 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2825 integer values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002826
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002827<h5>Semantics:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002828<p>The value produced is the integer product of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002829
Bill Wendlingf85859d2009-07-20 02:29:24 +00002830<p>If the result of the multiplication has unsigned overflow, the result
2831 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
2832 width of the result.</p>
2833
2834<p>Because LLVM integers use a two's complement representation, and the result
2835 is the same width as the operands, this instruction returns the correct
2836 result for both signed and unsigned integers. If a full product
2837 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
2838 be sign-extended or zero-extended as appropriate to the width of the full
2839 product.</p>
2840
Dan Gohman46e96012009-07-22 22:44:56 +00002841<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
2842 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
2843 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
2844 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
Dan Gohmaned5fcb22009-07-20 22:41:19 +00002845
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002846<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002847<pre>
2848 &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002849</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002850
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002851</div>
Chris Lattner6704c212008-05-20 20:48:21 +00002852
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002853<!-- _______________________________________________________________________ -->
Dan Gohman7ce405e2009-06-04 22:49:04 +00002854<div class="doc_subsubsection">
2855 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
2856</div>
2857
2858<div class="doc_text">
2859
2860<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002861<pre>
2862 &lt;result&gt; = fmul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002863</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002864
Dan Gohman7ce405e2009-06-04 22:49:04 +00002865<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002866<p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002867
2868<h5>Arguments:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002869<p>The two arguments to the '<tt>fmul</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00002870 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2871 floating point values. Both arguments must have identical types.</p>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002872
2873<h5>Semantics:</h5>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002874<p>The value produced is the floating point product of the two operands.</p>
2875
2876<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002877<pre>
2878 &lt;result&gt; = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
Dan Gohman7ce405e2009-06-04 22:49:04 +00002879</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002880
Dan Gohman7ce405e2009-06-04 22:49:04 +00002881</div>
2882
2883<!-- _______________________________________________________________________ -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002884<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2885</a></div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002886
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002887<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00002888
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002889<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002890<pre>
2891 &lt;result&gt; = udiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002892</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002893
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002894<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002895<p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002896
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002897<h5>Arguments:</h5>
2898<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00002899 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2900 values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002901
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002902<h5>Semantics:</h5>
Chris Lattner9aba1e22008-01-28 00:36:27 +00002903<p>The value produced is the unsigned integer quotient of the two operands.</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002904
Chris Lattner9aba1e22008-01-28 00:36:27 +00002905<p>Note that unsigned integer division and signed integer division are distinct
Bill Wendlingf85859d2009-07-20 02:29:24 +00002906 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2907
Chris Lattner9aba1e22008-01-28 00:36:27 +00002908<p>Division by zero leads to undefined behavior.</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002909
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002910<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002911<pre>
2912 &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002913</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002914
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002915</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002916
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002917<!-- _______________________________________________________________________ -->
2918<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2919</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002920
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002921<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00002922
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002923<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00002924<pre>
Dan Gohmaned5fcb22009-07-20 22:41:19 +00002925 &lt;result&gt; = sdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
2926 &lt;result&gt; = exact sdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002927</pre>
Chris Lattner6704c212008-05-20 20:48:21 +00002928
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002929<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002930<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002931
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002932<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00002933<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00002934 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2935 values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002936
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002937<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002938<p>The value produced is the signed integer quotient of the two operands rounded
2939 towards zero.</p>
2940
Chris Lattner9aba1e22008-01-28 00:36:27 +00002941<p>Note that signed integer division and unsigned integer division are distinct
Bill Wendlingf85859d2009-07-20 02:29:24 +00002942 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2943
Chris Lattner9aba1e22008-01-28 00:36:27 +00002944<p>Division by zero leads to undefined behavior. Overflow also leads to
Bill Wendlingf85859d2009-07-20 02:29:24 +00002945 undefined behavior; this is a rare case, but can occur, for example, by doing
2946 a 32-bit division of -2147483648 by -1.</p>
2947
Dan Gohman67fa48e2009-07-22 00:04:19 +00002948<p>If the <tt>exact</tt> keyword is present, the result value of the
2949 <tt>sdiv</tt> is undefined if the result would be rounded or if overflow
2950 would occur.</p>
Dan Gohmaned5fcb22009-07-20 22:41:19 +00002951
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002952<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002953<pre>
2954 &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002955</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002956
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002957</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002958
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002959<!-- _______________________________________________________________________ -->
2960<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2961Instruction</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002962
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002963<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00002964
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002965<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00002966<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +00002967 &lt;result&gt; = fdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002968</pre>
Chris Lattner6704c212008-05-20 20:48:21 +00002969
Bill Wendlingf85859d2009-07-20 02:29:24 +00002970<h5>Overview:</h5>
2971<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002972
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002973<h5>Arguments:</h5>
2974<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00002975 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2976 floating point values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002977
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002978<h5>Semantics:</h5>
2979<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00002980
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002981<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00002982<pre>
2983 &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002984</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002985
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002986</div>
Chris Lattner6704c212008-05-20 20:48:21 +00002987
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002988<!-- _______________________________________________________________________ -->
2989<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2990</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002991
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002992<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00002993
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002994<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002995<pre>
2996 &lt;result&gt; = urem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002997</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00002998
Dan Gohmanf17a25c2007-07-18 16:29:46 +00002999<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003000<p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3001 division of its two arguments.</p>
3002
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003003<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00003004<p>The two arguments to the '<tt>urem</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00003005 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3006 values. Both arguments must have identical types.</p>
3007
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003008<h5>Semantics:</h5>
3009<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Bill Wendlingf85859d2009-07-20 02:29:24 +00003010 This instruction always performs an unsigned division to get the
3011 remainder.</p>
3012
Chris Lattner9aba1e22008-01-28 00:36:27 +00003013<p>Note that unsigned integer remainder and signed integer remainder are
Bill Wendlingf85859d2009-07-20 02:29:24 +00003014 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3015
Chris Lattner9aba1e22008-01-28 00:36:27 +00003016<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003017
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003018<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003019<pre>
3020 &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003021</pre>
3022
3023</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003024
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003025<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +00003026<div class="doc_subsubsection">
3027 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3028</div>
3029
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003030<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +00003031
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003032<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00003033<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +00003034 &lt;result&gt; = srem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003035</pre>
Chris Lattner6704c212008-05-20 20:48:21 +00003036
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003037<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003038<p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3039 division of its two operands. This instruction can also take
3040 <a href="#t_vector">vector</a> versions of the values in which case the
3041 elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +00003042
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003043<h5>Arguments:</h5>
3044<p>The two arguments to the '<tt>srem</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00003045 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3046 values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00003047
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003048<h5>Semantics:</h5>
3049<p>This instruction returns the <i>remainder</i> of a division (where the result
Bill Wendlingf85859d2009-07-20 02:29:24 +00003050 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3051 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3052 a value. For more information about the difference,
3053 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3054 Math Forum</a>. For a table of how this is implemented in various languages,
3055 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3056 Wikipedia: modulo operation</a>.</p>
3057
Chris Lattner9aba1e22008-01-28 00:36:27 +00003058<p>Note that signed integer remainder and unsigned integer remainder are
Bill Wendlingf85859d2009-07-20 02:29:24 +00003059 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3060
Chris Lattner9aba1e22008-01-28 00:36:27 +00003061<p>Taking the remainder of a division by zero leads to undefined behavior.
Bill Wendlingf85859d2009-07-20 02:29:24 +00003062 Overflow also leads to undefined behavior; this is a rare case, but can
3063 occur, for example, by taking the remainder of a 32-bit division of
3064 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3065 lets srem be implemented using instructions that return both the result of
3066 the division and the remainder.)</p>
3067
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003068<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003069<pre>
3070 &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003071</pre>
3072
3073</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003074
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003075<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +00003076<div class="doc_subsubsection">
3077 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3078
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003079<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +00003080
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003081<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003082<pre>
3083 &lt;result&gt; = frem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003084</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003085
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003086<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003087<p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3088 its two operands.</p>
3089
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003090<h5>Arguments:</h5>
3091<p>The two arguments to the '<tt>frem</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00003092 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3093 floating point values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00003094
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003095<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003096<p>This instruction returns the <i>remainder</i> of a division. The remainder
3097 has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00003098
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003099<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00003100<pre>
3101 &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003102</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003103
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003104</div>
3105
3106<!-- ======================================================================= -->
3107<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3108Operations</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003109
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003110<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00003111
3112<p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3113 program. They are generally very efficient instructions and can commonly be
3114 strength reduced from other instructions. They require two operands of the
3115 same type, execute an operation on them, and produce a single value. The
3116 resulting value is the same type as its operands.</p>
3117
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003118</div>
3119
3120<!-- _______________________________________________________________________ -->
3121<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3122Instruction</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003123
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003124<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00003125
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003126<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003127<pre>
3128 &lt;result&gt; = shl &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003129</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +00003130
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003131<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003132<p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3133 a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +00003134
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003135<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003136<p>Both arguments to the '<tt>shl</tt>' instruction must be the
3137 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3138 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +00003139
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003140<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003141<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3142 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3143 is (statically or dynamically) negative or equal to or larger than the number
3144 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3145 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3146 shift amount in <tt>op2</tt>.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +00003147
Bill Wendlingf85859d2009-07-20 02:29:24 +00003148<h5>Example:</h5>
3149<pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003150 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
3151 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
3152 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +00003153 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +00003154 &lt;result&gt; = shl &lt;2 x i32&gt; &lt; i32 1, i32 1&gt;, &lt; i32 1, i32 2&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 2, i32 4&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003155</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003156
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003157</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003158
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003159<!-- _______________________________________________________________________ -->
3160<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3161Instruction</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003162
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003163<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00003164
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003165<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003166<pre>
3167 &lt;result&gt; = lshr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003168</pre>
3169
3170<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003171<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3172 operand shifted to the right a specified number of bits with zero fill.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003173
3174<h5>Arguments:</h5>
3175<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Bill Wendlingf85859d2009-07-20 02:29:24 +00003176 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3177 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003178
3179<h5>Semantics:</h5>
3180<p>This instruction always performs a logical shift right operation. The most
Bill Wendlingf85859d2009-07-20 02:29:24 +00003181 significant bits of the result will be filled with zero bits after the shift.
3182 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3183 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3184 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3185 shift amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003186
3187<h5>Example:</h5>
3188<pre>
3189 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3190 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3191 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3192 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +00003193 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +00003194 &lt;result&gt; = lshr &lt;2 x i32&gt; &lt; i32 -2, i32 4&gt;, &lt; i32 1, i32 2&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0x7FFFFFFF, i32 1&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003195</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003196
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003197</div>
3198
3199<!-- _______________________________________________________________________ -->
3200<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3201Instruction</a> </div>
3202<div class="doc_text">
3203
3204<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003205<pre>
3206 &lt;result&gt; = ashr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003207</pre>
3208
3209<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003210<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3211 operand shifted to the right a specified number of bits with sign
3212 extension.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003213
3214<h5>Arguments:</h5>
3215<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Bill Wendlingf85859d2009-07-20 02:29:24 +00003216 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3217 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003218
3219<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003220<p>This instruction always performs an arithmetic shift right operation, The
3221 most significant bits of the result will be filled with the sign bit
3222 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3223 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3224 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3225 the corresponding shift amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003226
3227<h5>Example:</h5>
3228<pre>
3229 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3230 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3231 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3232 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +00003233 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +00003234 &lt;result&gt; = ashr &lt;2 x i32&gt; &lt; i32 -2, i32 4&gt;, &lt; i32 1, i32 3&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 -1, i32 0&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003235</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003236
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003237</div>
3238
3239<!-- _______________________________________________________________________ -->
3240<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3241Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +00003242
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003243<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +00003244
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003245<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00003246<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +00003247 &lt;result&gt; = and &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003248</pre>
Chris Lattner6704c212008-05-20 20:48:21 +00003249
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003250<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003251<p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3252 operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00003253
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003254<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00003255<p>The two arguments to the '<tt>and</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00003256 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3257 values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00003258
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003259<h5>Semantics:</h5>
3260<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003261
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003262<table border="1" cellspacing="0" cellpadding="4">
3263 <tbody>
3264 <tr>
3265 <td>In0</td>
3266 <td>In1</td>
3267 <td>Out</td>
3268 </tr>
3269 <tr>
3270 <td>0</td>
3271 <td>0</td>
3272 <td>0</td>
3273 </tr>
3274 <tr>
3275 <td>0</td>
3276 <td>1</td>
3277 <td>0</td>
3278 </tr>
3279 <tr>
3280 <td>1</td>
3281 <td>0</td>
3282 <td>0</td>
3283 </tr>
3284 <tr>
3285 <td>1</td>
3286 <td>1</td>
3287 <td>1</td>
3288 </tr>
3289 </tbody>
3290</table>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003291
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003292<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00003293<pre>
3294 &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003295 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3296 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3297</pre>
3298</div>
3299<!-- _______________________________________________________________________ -->
3300<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +00003301
Bill Wendlingf85859d2009-07-20 02:29:24 +00003302<div class="doc_text">
3303
3304<h5>Syntax:</h5>
3305<pre>
3306 &lt;result&gt; = or &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3307</pre>
3308
3309<h5>Overview:</h5>
3310<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3311 two operands.</p>
3312
3313<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00003314<p>The two arguments to the '<tt>or</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00003315 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3316 values. Both arguments must have identical types.</p>
3317
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003318<h5>Semantics:</h5>
3319<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003320
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003321<table border="1" cellspacing="0" cellpadding="4">
3322 <tbody>
3323 <tr>
3324 <td>In0</td>
3325 <td>In1</td>
3326 <td>Out</td>
3327 </tr>
3328 <tr>
3329 <td>0</td>
3330 <td>0</td>
3331 <td>0</td>
3332 </tr>
3333 <tr>
3334 <td>0</td>
3335 <td>1</td>
3336 <td>1</td>
3337 </tr>
3338 <tr>
3339 <td>1</td>
3340 <td>0</td>
3341 <td>1</td>
3342 </tr>
3343 <tr>
3344 <td>1</td>
3345 <td>1</td>
3346 <td>1</td>
3347 </tr>
3348 </tbody>
3349</table>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003350
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003351<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003352<pre>
3353 &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003354 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3355 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3356</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003357
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003358</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003359
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003360<!-- _______________________________________________________________________ -->
3361<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3362Instruction</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003363
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003364<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00003365
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003366<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003367<pre>
3368 &lt;result&gt; = xor &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003369</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003370
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003371<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003372<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3373 its two operands. The <tt>xor</tt> is used to implement the "one's
3374 complement" operation, which is the "~" operator in C.</p>
3375
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003376<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00003377<p>The two arguments to the '<tt>xor</tt>' instruction must be
Bill Wendlingf85859d2009-07-20 02:29:24 +00003378 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3379 values. Both arguments must have identical types.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00003380
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003381<h5>Semantics:</h5>
3382<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003383
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003384<table border="1" cellspacing="0" cellpadding="4">
3385 <tbody>
3386 <tr>
3387 <td>In0</td>
3388 <td>In1</td>
3389 <td>Out</td>
3390 </tr>
3391 <tr>
3392 <td>0</td>
3393 <td>0</td>
3394 <td>0</td>
3395 </tr>
3396 <tr>
3397 <td>0</td>
3398 <td>1</td>
3399 <td>1</td>
3400 </tr>
3401 <tr>
3402 <td>1</td>
3403 <td>0</td>
3404 <td>1</td>
3405 </tr>
3406 <tr>
3407 <td>1</td>
3408 <td>1</td>
3409 <td>0</td>
3410 </tr>
3411 </tbody>
3412</table>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003413
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003414<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003415<pre>
3416 &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003417 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3418 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3419 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3420</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003421
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003422</div>
3423
3424<!-- ======================================================================= -->
3425<div class="doc_subsection">
3426 <a name="vectorops">Vector Operations</a>
3427</div>
3428
3429<div class="doc_text">
3430
3431<p>LLVM supports several instructions to represent vector operations in a
Bill Wendlingf85859d2009-07-20 02:29:24 +00003432 target-independent manner. These instructions cover the element-access and
3433 vector-specific operations needed to process vectors effectively. While LLVM
3434 does directly support these vector operations, many sophisticated algorithms
3435 will want to use target-specific intrinsics to take full advantage of a
3436 specific target.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003437
3438</div>
3439
3440<!-- _______________________________________________________________________ -->
3441<div class="doc_subsubsection">
3442 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3443</div>
3444
3445<div class="doc_text">
3446
3447<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003448<pre>
3449 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
3450</pre>
3451
3452<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003453<p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3454 from a vector at a specified index.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003455
3456
3457<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003458<p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3459 of <a href="#t_vector">vector</a> type. The second operand is an index
3460 indicating the position from which to extract the element. The index may be
3461 a variable.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003462
3463<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003464<p>The result is a scalar of the same type as the element type of
3465 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3466 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3467 results are undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003468
3469<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003470<pre>
3471 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
3472</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003473
Bill Wendlingf85859d2009-07-20 02:29:24 +00003474</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003475
3476<!-- _______________________________________________________________________ -->
3477<div class="doc_subsubsection">
3478 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3479</div>
3480
3481<div class="doc_text">
3482
3483<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003484<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +00003485 &lt;result&gt; = insertelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, &lt;ty&gt; &lt;elt&gt;, i32 &lt;idx&gt; <i>; yields &lt;n x &lt;ty&gt;&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003486</pre>
3487
3488<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003489<p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
3490 vector at a specified index.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003491
3492<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003493<p>The first operand of an '<tt>insertelement</tt>' instruction is a value
3494 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
3495 whose type must equal the element type of the first operand. The third
3496 operand is an index indicating the position at which to insert the value.
3497 The index may be a variable.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003498
3499<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003500<p>The result is a vector of the same type as <tt>val</tt>. Its element values
3501 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
3502 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3503 results are undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003504
3505<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003506<pre>
3507 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
3508</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003509
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003510</div>
3511
3512<!-- _______________________________________________________________________ -->
3513<div class="doc_subsubsection">
3514 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3515</div>
3516
3517<div class="doc_text">
3518
3519<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003520<pre>
Mon P Wangbff5d9c2008-11-10 04:46:22 +00003521 &lt;result&gt; = shufflevector &lt;n x &lt;ty&gt;&gt; &lt;v1&gt;, &lt;n x &lt;ty&gt;&gt; &lt;v2&gt;, &lt;m x i32&gt; &lt;mask&gt; <i>; yields &lt;m x &lt;ty&gt;&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003522</pre>
3523
3524<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003525<p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3526 from two input vectors, returning a vector with the same element type as the
3527 input and length that is the same as the shuffle mask.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003528
3529<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003530<p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3531 with types that match each other. The third argument is a shuffle mask whose
3532 element type is always 'i32'. The result of the instruction is a vector
3533 whose length is the same as the shuffle mask and whose element type is the
3534 same as the element type of the first two operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003535
Bill Wendlingf85859d2009-07-20 02:29:24 +00003536<p>The shuffle mask operand is required to be a constant vector with either
3537 constant integer or undef values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003538
3539<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003540<p>The elements of the two input vectors are numbered from left to right across
3541 both of the vectors. The shuffle mask operand specifies, for each element of
3542 the result vector, which element of the two input vectors the result element
3543 gets. The element selector may be undef (meaning "don't care") and the
3544 second operand may be undef if performing a shuffle from only one vector.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003545
3546<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003547<pre>
3548 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3549 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
3550 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
3551 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i> - Identity shuffle.
Mon P Wangbff5d9c2008-11-10 04:46:22 +00003552 %result = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
3553 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i>
3554 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3555 &lt;8 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 &gt; <i>; yields &lt;8 x i32&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003556</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003557
Bill Wendlingf85859d2009-07-20 02:29:24 +00003558</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003559
3560<!-- ======================================================================= -->
3561<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +00003562 <a name="aggregateops">Aggregate Operations</a>
3563</div>
3564
3565<div class="doc_text">
3566
Bill Wendlingf85859d2009-07-20 02:29:24 +00003567<p>LLVM supports several instructions for working with aggregate values.</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003568
3569</div>
3570
3571<!-- _______________________________________________________________________ -->
3572<div class="doc_subsubsection">
3573 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3574</div>
3575
3576<div class="doc_text">
3577
3578<h5>Syntax:</h5>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003579<pre>
3580 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
3581</pre>
3582
3583<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003584<p>The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3585 or array element from an aggregate value.</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003586
3587<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003588<p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
3589 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3590 operands are constant indices to specify which value to extract in a similar
3591 manner as indices in a
3592 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003593
3594<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003595<p>The result is the value at the position in the aggregate specified by the
3596 index operands.</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003597
3598<h5>Example:</h5>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003599<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +00003600 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003601</pre>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003602
Bill Wendlingf85859d2009-07-20 02:29:24 +00003603</div>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003604
3605<!-- _______________________________________________________________________ -->
3606<div class="doc_subsubsection">
3607 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3608</div>
3609
3610<div class="doc_text">
3611
3612<h5>Syntax:</h5>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003613<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +00003614 &lt;result&gt; = insertvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;ty&gt; &lt;val&gt;, &lt;idx&gt; <i>; yields &lt;n x &lt;ty&gt;&gt;</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003615</pre>
3616
3617<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003618<p>The '<tt>insertvalue</tt>' instruction inserts a value into a struct field or
3619 array element in an aggregate.</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003620
3621
3622<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003623<p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
3624 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3625 second operand is a first-class value to insert. The following operands are
3626 constant indices indicating the position at which to insert the value in a
3627 similar manner as indices in a
3628 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
3629 value to insert must have the same type as the value identified by the
3630 indices.</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003631
3632<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003633<p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
3634 that of <tt>val</tt> except that the value at the position specified by the
3635 indices is that of <tt>elt</tt>.</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003636
3637<h5>Example:</h5>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003638<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +00003639 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +00003640</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003641
Dan Gohman74d6faf2008-05-12 23:51:09 +00003642</div>
3643
3644
3645<!-- ======================================================================= -->
3646<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003647 <a name="memoryops">Memory Access and Addressing Operations</a>
3648</div>
3649
3650<div class="doc_text">
3651
Bill Wendlingf85859d2009-07-20 02:29:24 +00003652<p>A key design point of an SSA-based representation is how it represents
3653 memory. In LLVM, no memory locations are in SSA form, which makes things
3654 very simple. This section describes how to read, write, allocate, and free
3655 memory in LLVM.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003656
3657</div>
3658
3659<!-- _______________________________________________________________________ -->
3660<div class="doc_subsubsection">
3661 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3662</div>
3663
3664<div class="doc_text">
3665
3666<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003667<pre>
3668 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3669</pre>
3670
3671<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003672<p>The '<tt>malloc</tt>' instruction allocates memory from the system heap and
3673 returns a pointer to it. The object is always allocated in the generic
3674 address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003675
3676<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003677<p>The '<tt>malloc</tt>' instruction allocates
Bill Wendlingf85859d2009-07-20 02:29:24 +00003678 <tt>sizeof(&lt;type&gt;)*NumElements</tt> bytes of memory from the operating
3679 system and returns a pointer of the appropriate type to the program. If
3680 "NumElements" is specified, it is the number of elements allocated, otherwise
3681 "NumElements" is defaulted to be one. If a constant alignment is specified,
3682 the value result of the allocation is guaranteed to be aligned to at least
3683 that boundary. If not specified, or if zero, the target can choose to align
3684 the allocation on any convenient boundary compatible with the type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003685
3686<p>'<tt>type</tt>' must be a sized type.</p>
3687
3688<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003689<p>Memory is allocated using the system "<tt>malloc</tt>" function, and a
3690 pointer is returned. The result of a zero byte allocation is undefined. The
3691 result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003692
3693<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003694<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +00003695 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003696
3697 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3698 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3699 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3700 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3701 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3702</pre>
Dan Gohman60967192009-01-12 23:12:39 +00003703
Bill Wendlingf85859d2009-07-20 02:29:24 +00003704<p>Note that the code generator does not yet respect the alignment value.</p>
Dan Gohman60967192009-01-12 23:12:39 +00003705
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003706</div>
3707
3708<!-- _______________________________________________________________________ -->
3709<div class="doc_subsubsection">
3710 <a name="i_free">'<tt>free</tt>' Instruction</a>
3711</div>
3712
3713<div class="doc_text">
3714
3715<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003716<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +00003717 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003718</pre>
3719
3720<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003721<p>The '<tt>free</tt>' instruction returns memory back to the unused memory heap
3722 to be reallocated in the future.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003723
3724<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003725<p>'<tt>value</tt>' shall be a pointer value that points to a value that was
3726 allocated with the '<tt><a href="#i_malloc">malloc</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003727
3728<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003729<p>Access to the memory pointed to by the pointer is no longer defined after
3730 this instruction executes. If the pointer is null, the operation is a
3731 noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003732
3733<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003734<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +00003735 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003736 free [4 x i8]* %array
3737</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003738
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003739</div>
3740
3741<!-- _______________________________________________________________________ -->
3742<div class="doc_subsubsection">
3743 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3744</div>
3745
3746<div class="doc_text">
3747
3748<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003749<pre>
3750 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3751</pre>
3752
3753<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003754<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
Bill Wendlingf85859d2009-07-20 02:29:24 +00003755 currently executing function, to be automatically released when this function
3756 returns to its caller. The object is always allocated in the generic address
3757 space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003758
3759<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003760<p>The '<tt>alloca</tt>' instruction
3761 allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt> bytes of memory on the
3762 runtime stack, returning a pointer of the appropriate type to the program.
3763 If "NumElements" is specified, it is the number of elements allocated,
3764 otherwise "NumElements" is defaulted to be one. If a constant alignment is
3765 specified, the value result of the allocation is guaranteed to be aligned to
3766 at least that boundary. If not specified, or if zero, the target can choose
3767 to align the allocation on any convenient boundary compatible with the
3768 type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003769
3770<p>'<tt>type</tt>' may be any sized type.</p>
3771
3772<h5>Semantics:</h5>
Bill Wendling2a454572009-05-08 20:49:29 +00003773<p>Memory is allocated; a pointer is returned. The operation is undefined if
Bill Wendlingf85859d2009-07-20 02:29:24 +00003774 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3775 memory is automatically released when the function returns. The
3776 '<tt>alloca</tt>' instruction is commonly used to represent automatic
3777 variables that must have an address available. When the function returns
3778 (either with the <tt><a href="#i_ret">ret</a></tt>
3779 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
3780 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003781
3782<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003783<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +00003784 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3785 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3786 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3787 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003788</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003789
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003790</div>
3791
3792<!-- _______________________________________________________________________ -->
3793<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3794Instruction</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003795
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003796<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00003797
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003798<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003799<pre>
3800 &lt;result&gt; = load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;]
3801 &lt;result&gt; = volatile load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;]
3802</pre>
3803
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003804<h5>Overview:</h5>
3805<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003806
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003807<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003808<p>The argument to the '<tt>load</tt>' instruction specifies the memory address
3809 from which to load. The pointer must point to
3810 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3811 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
3812 number or order of execution of this <tt>load</tt> with other
3813 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3814 instructions. </p>
3815
3816<p>The optional constant "align" argument specifies the alignment of the
3817 operation (that is, the alignment of the memory address). A value of 0 or an
3818 omitted "align" argument means that the operation has the preferential
3819 alignment for the target. It is the responsibility of the code emitter to
3820 ensure that the alignment information is correct. Overestimating the
3821 alignment results in an undefined behavior. Underestimating the alignment may
3822 produce less efficient code. An alignment of 1 is always safe.</p>
3823
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003824<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003825<p>The location of memory pointed to is loaded. If the value being loaded is of
3826 scalar type then the number of bytes read does not exceed the minimum number
3827 of bytes needed to hold all bits of the type. For example, loading an
3828 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
3829 <tt>i20</tt> with a size that is not an integral number of bytes, the result
3830 is undefined if the value was not originally written using a store of the
3831 same type.</p>
3832
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003833<h5>Examples:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003834<pre>
3835 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3836 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003837 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3838</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003839
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003840</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003841
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003842<!-- _______________________________________________________________________ -->
3843<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3844Instruction</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003845
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003846<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00003847
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003848<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003849<pre>
3850 store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003851 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3852</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003853
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003854<h5>Overview:</h5>
3855<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003856
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003857<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003858<p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
3859 and an address at which to store it. The type of the
3860 '<tt>&lt;pointer&gt;</tt>' operand must be a pointer to
3861 the <a href="#t_firstclass">first class</a> type of the
3862 '<tt>&lt;value&gt;</tt>' operand. If the <tt>store</tt> is marked
3863 as <tt>volatile</tt>, then the optimizer is not allowed to modify the number
3864 or order of execution of this <tt>store</tt> with other
3865 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3866 instructions.</p>
3867
3868<p>The optional constant "align" argument specifies the alignment of the
3869 operation (that is, the alignment of the memory address). A value of 0 or an
3870 omitted "align" argument means that the operation has the preferential
3871 alignment for the target. It is the responsibility of the code emitter to
3872 ensure that the alignment information is correct. Overestimating the
3873 alignment results in an undefined behavior. Underestimating the alignment may
3874 produce less efficient code. An alignment of 1 is always safe.</p>
3875
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003876<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003877<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>' at the
3878 location specified by the '<tt>&lt;pointer&gt;</tt>' operand. If
3879 '<tt>&lt;value&gt;</tt>' is of scalar type then the number of bytes written
3880 does not exceed the minimum number of bytes needed to hold all bits of the
3881 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
3882 writing a value of a type like <tt>i20</tt> with a size that is not an
3883 integral number of bytes, it is unspecified what happens to the extra bits
3884 that do not belong to the type, but they will typically be overwritten.</p>
3885
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003886<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003887<pre>
3888 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +00003889 store i32 3, i32* %ptr <i>; yields {void}</i>
3890 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003891</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003892
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003893</div>
3894
3895<!-- _______________________________________________________________________ -->
3896<div class="doc_subsubsection">
3897 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3898</div>
3899
3900<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00003901
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003902<h5>Syntax:</h5>
3903<pre>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +00003904 &lt;result&gt; = getelementptr &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003905</pre>
3906
3907<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00003908<p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
3909 subelement of an aggregate data structure. It performs address calculation
3910 only and does not access memory.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003911
3912<h5>Arguments:</h5>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +00003913<p>The first argument is always a pointer, and forms the basis of the
Bill Wendlingf85859d2009-07-20 02:29:24 +00003914 calculation. The remaining arguments are indices, that indicate which of the
3915 elements of the aggregate object are indexed. The interpretation of each
3916 index is dependent on the type being indexed into. The first index always
3917 indexes the pointer value given as the first argument, the second index
3918 indexes a value of the type pointed to (not necessarily the value directly
3919 pointed to, since the first index can be non-zero), etc. The first type
3920 indexed into must be a pointer value, subsequent types can be arrays, vectors
3921 and structs. Note that subsequent types being indexed into can never be
3922 pointers, since that would require loading the pointer before continuing
3923 calculation.</p>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +00003924
3925<p>The type of each index argument depends on the type it is indexing into.
Bill Wendlingf85859d2009-07-20 02:29:24 +00003926 When indexing into a (packed) structure, only <tt>i32</tt> integer
3927 <b>constants</b> are allowed. When indexing into an array, pointer or
3928 vector, integers of any width are allowed (also non-constants).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003929
Bill Wendlingf85859d2009-07-20 02:29:24 +00003930<p>For example, let's consider a C code fragment and how it gets compiled to
3931 LLVM:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003932
3933<div class="doc_code">
3934<pre>
3935struct RT {
3936 char A;
3937 int B[10][20];
3938 char C;
3939};
3940struct ST {
3941 int X;
3942 double Y;
3943 struct RT Z;
3944};
3945
3946int *foo(struct ST *s) {
3947 return &amp;s[1].Z.B[5][13];
3948}
3949</pre>
3950</div>
3951
3952<p>The LLVM code generated by the GCC frontend is:</p>
3953
3954<div class="doc_code">
3955<pre>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +00003956%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3957%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003958
Dan Gohman47360842009-07-25 02:23:48 +00003959define i32* @foo(%ST* %s) {
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003960entry:
3961 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3962 ret i32* %reg
3963}
3964</pre>
3965</div>
3966
3967<h5>Semantics:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003968<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
Bill Wendlingf85859d2009-07-20 02:29:24 +00003969 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3970 }</tt>' type, a structure. The second index indexes into the third element
3971 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3972 i8 }</tt>' type, another structure. The third index indexes into the second
3973 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3974 array. The two dimensions of the array are subscripted into, yielding an
3975 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
3976 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003977
Bill Wendlingf85859d2009-07-20 02:29:24 +00003978<p>Note that it is perfectly legal to index partially through a structure,
3979 returning a pointer to an inner element. Because of this, the LLVM code for
3980 the given testcase is equivalent to:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003981
3982<pre>
Dan Gohman47360842009-07-25 02:23:48 +00003983 define i32* @foo(%ST* %s) {
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003984 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3985 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3986 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3987 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3988 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3989 ret i32* %t5
3990 }
3991</pre>
3992
Bill Wendlingf85859d2009-07-20 02:29:24 +00003993<p>The getelementptr instruction is often confusing. For some more insight into
3994 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003995
3996<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00003997<pre>
3998 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +00003999 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4000 <i>; yields i8*:vptr</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +00004001 %vptr = getelementptr {i32, &lt;2 x i8&gt;}* %svptr, i64 0, i32 1, i32 1
Matthijs Kooijman32a080e2008-10-13 13:44:15 +00004002 <i>; yields i8*:eptr</i>
4003 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Sanjiv Gupta4f9a0dc2009-04-25 07:27:44 +00004004 <i>; yields i32*:iptr</i>
Sanjiv Gupta1e46c582009-04-24 16:38:13 +00004005 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004006</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004007
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004008</div>
4009
4010<!-- ======================================================================= -->
4011<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4012</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004013
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004014<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00004015
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004016<p>The instructions in this category are the conversion instructions (casting)
Bill Wendlingf85859d2009-07-20 02:29:24 +00004017 which all take a single operand and a type. They perform various bit
4018 conversions on the operand.</p>
4019
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004020</div>
4021
4022<!-- _______________________________________________________________________ -->
4023<div class="doc_subsubsection">
4024 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4025</div>
4026<div class="doc_text">
4027
4028<h5>Syntax:</h5>
4029<pre>
4030 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4031</pre>
4032
4033<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004034<p>The '<tt>trunc</tt>' instruction truncates its operand to the
4035 type <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004036
4037<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004038<p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4039 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4040 size and type of the result, which must be
4041 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4042 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4043 allowed.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004044
4045<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004046<p>The '<tt>trunc</tt>' instruction truncates the high order bits
4047 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4048 source size must be larger than the destination size, <tt>trunc</tt> cannot
4049 be a <i>no-op cast</i>. It will always truncate bits.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004050
4051<h5>Example:</h5>
4052<pre>
4053 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4054 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4055 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
4056</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004057
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004058</div>
4059
4060<!-- _______________________________________________________________________ -->
4061<div class="doc_subsubsection">
4062 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4063</div>
4064<div class="doc_text">
4065
4066<h5>Syntax:</h5>
4067<pre>
4068 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4069</pre>
4070
4071<h5>Overview:</h5>
4072<p>The '<tt>zext</tt>' instruction zero extends its operand to type
Bill Wendlingf85859d2009-07-20 02:29:24 +00004073 <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004074
4075
4076<h5>Arguments:</h5>
4077<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
Bill Wendlingf85859d2009-07-20 02:29:24 +00004078 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4079 also be of <a href="#t_integer">integer</a> type. The bit size of the
4080 <tt>value</tt> must be smaller than the bit size of the destination type,
4081 <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004082
4083<h5>Semantics:</h5>
4084<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
Bill Wendlingf85859d2009-07-20 02:29:24 +00004085 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004086
4087<p>When zero extending from i1, the result will always be either 0 or 1.</p>
4088
4089<h5>Example:</h5>
4090<pre>
4091 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4092 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4093</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004094
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004095</div>
4096
4097<!-- _______________________________________________________________________ -->
4098<div class="doc_subsubsection">
4099 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4100</div>
4101<div class="doc_text">
4102
4103<h5>Syntax:</h5>
4104<pre>
4105 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4106</pre>
4107
4108<h5>Overview:</h5>
4109<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4110
4111<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004112<p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4113 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4114 also be of <a href="#t_integer">integer</a> type. The bit size of the
4115 <tt>value</tt> must be smaller than the bit size of the destination type,
4116 <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004117
4118<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004119<p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4120 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4121 of the type <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004122
4123<p>When sign extending from i1, the extension always results in -1 or 0.</p>
4124
4125<h5>Example:</h5>
4126<pre>
4127 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4128 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4129</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004130
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004131</div>
4132
4133<!-- _______________________________________________________________________ -->
4134<div class="doc_subsubsection">
4135 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4136</div>
4137
4138<div class="doc_text">
4139
4140<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004141<pre>
4142 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4143</pre>
4144
4145<h5>Overview:</h5>
4146<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
Bill Wendlingf85859d2009-07-20 02:29:24 +00004147 <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004148
4149<h5>Arguments:</h5>
4150<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
Bill Wendlingf85859d2009-07-20 02:29:24 +00004151 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4152 to cast it to. The size of <tt>value</tt> must be larger than the size of
4153 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4154 <i>no-op cast</i>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004155
4156<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004157<p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4158 <a href="#t_floating">floating point</a> type to a smaller
4159 <a href="#t_floating">floating point</a> type. If the value cannot fit
4160 within the destination type, <tt>ty2</tt>, then the results are
4161 undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004162
4163<h5>Example:</h5>
4164<pre>
4165 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4166 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4167</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004168
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004169</div>
4170
4171<!-- _______________________________________________________________________ -->
4172<div class="doc_subsubsection">
4173 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4174</div>
4175<div class="doc_text">
4176
4177<h5>Syntax:</h5>
4178<pre>
4179 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4180</pre>
4181
4182<h5>Overview:</h5>
4183<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
Bill Wendlingf85859d2009-07-20 02:29:24 +00004184 floating point value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004185
4186<h5>Arguments:</h5>
4187<p>The '<tt>fpext</tt>' instruction takes a
Bill Wendlingf85859d2009-07-20 02:29:24 +00004188 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4189 a <a href="#t_floating">floating point</a> type to cast it to. The source
4190 type must be smaller than the destination type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004191
4192<h5>Semantics:</h5>
4193<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
Bill Wendlingf85859d2009-07-20 02:29:24 +00004194 <a href="#t_floating">floating point</a> type to a larger
4195 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4196 used to make a <i>no-op cast</i> because it always changes bits. Use
4197 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004198
4199<h5>Example:</h5>
4200<pre>
4201 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4202 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4203</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004204
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004205</div>
4206
4207<!-- _______________________________________________________________________ -->
4208<div class="doc_subsubsection">
4209 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4210</div>
4211<div class="doc_text">
4212
4213<h5>Syntax:</h5>
4214<pre>
Reid Spencere6adee82007-07-31 14:40:14 +00004215 &lt;result&gt; = fptoui &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004216</pre>
4217
4218<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +00004219<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Bill Wendlingf85859d2009-07-20 02:29:24 +00004220 unsigned integer equivalent of type <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004221
4222<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004223<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4224 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4225 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4226 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4227 vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004228
4229<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004230<p>The '<tt>fptoui</tt>' instruction converts its
4231 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4232 towards zero) unsigned integer value. If the value cannot fit
4233 in <tt>ty2</tt>, the results are undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004234
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004235<h5>Example:</h5>
4236<pre>
Reid Spencere6adee82007-07-31 14:40:14 +00004237 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +00004238 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +00004239 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004240</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004241
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004242</div>
4243
4244<!-- _______________________________________________________________________ -->
4245<div class="doc_subsubsection">
4246 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4247</div>
4248<div class="doc_text">
4249
4250<h5>Syntax:</h5>
4251<pre>
4252 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4253</pre>
4254
4255<h5>Overview:</h5>
4256<p>The '<tt>fptosi</tt>' instruction converts
Bill Wendlingf85859d2009-07-20 02:29:24 +00004257 <a href="#t_floating">floating point</a> <tt>value</tt> to
4258 type <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004259
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004260<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004261<p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4262 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4263 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4264 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4265 vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004266
4267<h5>Semantics:</h5>
4268<p>The '<tt>fptosi</tt>' instruction converts its
Bill Wendlingf85859d2009-07-20 02:29:24 +00004269 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4270 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4271 the results are undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004272
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004273<h5>Example:</h5>
4274<pre>
4275 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +00004276 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004277 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4278</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004279
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004280</div>
4281
4282<!-- _______________________________________________________________________ -->
4283<div class="doc_subsubsection">
4284 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4285</div>
4286<div class="doc_text">
4287
4288<h5>Syntax:</h5>
4289<pre>
4290 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4291</pre>
4292
4293<h5>Overview:</h5>
4294<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
Bill Wendlingf85859d2009-07-20 02:29:24 +00004295 integer and converts that value to the <tt>ty2</tt> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004296
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004297<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +00004298<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlingf85859d2009-07-20 02:29:24 +00004299 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4300 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4301 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4302 floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004303
4304<h5>Semantics:</h5>
4305<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
Bill Wendlingf85859d2009-07-20 02:29:24 +00004306 integer quantity and converts it to the corresponding floating point
4307 value. If the value cannot fit in the floating point value, the results are
4308 undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004309
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004310<h5>Example:</h5>
4311<pre>
4312 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +00004313 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004314</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004315
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004316</div>
4317
4318<!-- _______________________________________________________________________ -->
4319<div class="doc_subsubsection">
4320 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4321</div>
4322<div class="doc_text">
4323
4324<h5>Syntax:</h5>
4325<pre>
4326 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4327</pre>
4328
4329<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004330<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4331 and converts that value to the <tt>ty2</tt> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004332
4333<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +00004334<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlingf85859d2009-07-20 02:29:24 +00004335 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4336 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4337 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4338 floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004339
4340<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004341<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4342 quantity and converts it to the corresponding floating point value. If the
4343 value cannot fit in the floating point value, the results are undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004344
4345<h5>Example:</h5>
4346<pre>
4347 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +00004348 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004349</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004350
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004351</div>
4352
4353<!-- _______________________________________________________________________ -->
4354<div class="doc_subsubsection">
4355 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4356</div>
4357<div class="doc_text">
4358
4359<h5>Syntax:</h5>
4360<pre>
4361 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4362</pre>
4363
4364<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004365<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4366 the integer type <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004367
4368<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004369<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4370 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4371 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004372
4373<h5>Semantics:</h5>
4374<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
Bill Wendlingf85859d2009-07-20 02:29:24 +00004375 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4376 truncating or zero extending that value to the size of the integer type. If
4377 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4378 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4379 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4380 change.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004381
4382<h5>Example:</h5>
4383<pre>
4384 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4385 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4386</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004387
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004388</div>
4389
4390<!-- _______________________________________________________________________ -->
4391<div class="doc_subsubsection">
4392 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4393</div>
4394<div class="doc_text">
4395
4396<h5>Syntax:</h5>
4397<pre>
4398 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4399</pre>
4400
4401<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004402<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4403 pointer type, <tt>ty2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004404
4405<h5>Arguments:</h5>
4406<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004407 value to cast, and a type to cast it to, which must be a
4408 <a href="#t_pointer">pointer</a> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004409
4410<h5>Semantics:</h5>
4411<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingf85859d2009-07-20 02:29:24 +00004412 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4413 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4414 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4415 than the size of a pointer then a zero extension is done. If they are the
4416 same size, nothing is done (<i>no-op cast</i>).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004417
4418<h5>Example:</h5>
4419<pre>
4420 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4421 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4422 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4423</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004424
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004425</div>
4426
4427<!-- _______________________________________________________________________ -->
4428<div class="doc_subsubsection">
4429 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4430</div>
4431<div class="doc_text">
4432
4433<h5>Syntax:</h5>
4434<pre>
4435 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4436</pre>
4437
4438<h5>Overview:</h5>
4439<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingf85859d2009-07-20 02:29:24 +00004440 <tt>ty2</tt> without changing any bits.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004441
4442<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004443<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4444 non-aggregate first class value, and a type to cast it to, which must also be
4445 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4446 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4447 identical. If the source type is a pointer, the destination type must also be
4448 a pointer. This instruction supports bitwise conversion of vectors to
4449 integers and to vectors of other types (as long as they have the same
4450 size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004451
4452<h5>Semantics:</h5>
4453<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingf85859d2009-07-20 02:29:24 +00004454 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4455 this conversion. The conversion is done as if the <tt>value</tt> had been
4456 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4457 be converted to other pointer types with this instruction. To convert
4458 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4459 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004460
4461<h5>Example:</h5>
4462<pre>
4463 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4464 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +00004465 %Z = bitcast &lt;2 x int&gt; %V to i64; <i>; yields i64: %V</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004466</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004467
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004468</div>
4469
4470<!-- ======================================================================= -->
4471<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004472
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004473<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00004474
4475<p>The instructions in this category are the "miscellaneous" instructions, which
4476 defy better classification.</p>
4477
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004478</div>
4479
4480<!-- _______________________________________________________________________ -->
4481<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4482</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004483
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004484<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00004485
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004486<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004487<pre>
4488 &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {i1} or {&lt;N x i1&gt;}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004489</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004490
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004491<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004492<p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4493 boolean values based on comparison of its two integer, integer vector, or
4494 pointer operands.</p>
4495
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004496<h5>Arguments:</h5>
4497<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
Bill Wendlingf85859d2009-07-20 02:29:24 +00004498 the condition code indicating the kind of comparison to perform. It is not a
4499 value, just a keyword. The possible condition code are:</p>
4500
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004501<ol>
4502 <li><tt>eq</tt>: equal</li>
4503 <li><tt>ne</tt>: not equal </li>
4504 <li><tt>ugt</tt>: unsigned greater than</li>
4505 <li><tt>uge</tt>: unsigned greater or equal</li>
4506 <li><tt>ult</tt>: unsigned less than</li>
4507 <li><tt>ule</tt>: unsigned less or equal</li>
4508 <li><tt>sgt</tt>: signed greater than</li>
4509 <li><tt>sge</tt>: signed greater or equal</li>
4510 <li><tt>slt</tt>: signed less than</li>
4511 <li><tt>sle</tt>: signed less or equal</li>
4512</ol>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004513
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004514<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Bill Wendlingf85859d2009-07-20 02:29:24 +00004515 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
4516 typed. They must also be identical types.</p>
4517
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004518<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004519<p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
4520 condition code given as <tt>cond</tt>. The comparison performed always yields
4521 either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt>
4522 result, as follows:</p>
4523
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004524<ol>
4525 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
Bill Wendlingf85859d2009-07-20 02:29:24 +00004526 <tt>false</tt> otherwise. No sign interpretation is necessary or
4527 performed.</li>
4528
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004529 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Bill Wendlingf85859d2009-07-20 02:29:24 +00004530 <tt>false</tt> otherwise. No sign interpretation is necessary or
4531 performed.</li>
4532
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004533 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Bill Wendlingf85859d2009-07-20 02:29:24 +00004534 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4535
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004536 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Bill Wendlingf85859d2009-07-20 02:29:24 +00004537 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4538 to <tt>op2</tt>.</li>
4539
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004540 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Bill Wendlingf85859d2009-07-20 02:29:24 +00004541 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4542
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004543 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Bill Wendlingf85859d2009-07-20 02:29:24 +00004544 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4545
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004546 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Bill Wendlingf85859d2009-07-20 02:29:24 +00004547 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4548
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004549 <li><tt>sge</tt>: interprets the operands as signed values and yields
Bill Wendlingf85859d2009-07-20 02:29:24 +00004550 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4551 to <tt>op2</tt>.</li>
4552
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004553 <li><tt>slt</tt>: interprets the operands as signed values and yields
Bill Wendlingf85859d2009-07-20 02:29:24 +00004554 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4555
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004556 <li><tt>sle</tt>: interprets the operands as signed values and yields
Bill Wendlingf85859d2009-07-20 02:29:24 +00004557 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004558</ol>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004559
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004560<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
Bill Wendlingf85859d2009-07-20 02:29:24 +00004561 values are compared as if they were integers.</p>
4562
4563<p>If the operands are integer vectors, then they are compared element by
4564 element. The result is an <tt>i1</tt> vector with the same number of elements
4565 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004566
4567<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004568<pre>
4569 &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004570 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
4571 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
4572 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4573 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4574 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
4575</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +00004576
4577<p>Note that the code generator does not yet support vector types with
4578 the <tt>icmp</tt> instruction.</p>
4579
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004580</div>
4581
4582<!-- _______________________________________________________________________ -->
4583<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4584</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004585
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004586<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00004587
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004588<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004589<pre>
4590 &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {i1} or {&lt;N x i1&gt;}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004591</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004592
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004593<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004594<p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
4595 values based on comparison of its operands.</p>
4596
4597<p>If the operands are floating point scalars, then the result type is a boolean
4598(<a href="#t_primitive"><tt>i1</tt></a>).</p>
4599
4600<p>If the operands are floating point vectors, then the result type is a vector
4601 of boolean with the same number of elements as the operands being
4602 compared.</p>
4603
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004604<h5>Arguments:</h5>
4605<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
Bill Wendlingf85859d2009-07-20 02:29:24 +00004606 the condition code indicating the kind of comparison to perform. It is not a
4607 value, just a keyword. The possible condition code are:</p>
4608
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004609<ol>
4610 <li><tt>false</tt>: no comparison, always returns false</li>
4611 <li><tt>oeq</tt>: ordered and equal</li>
4612 <li><tt>ogt</tt>: ordered and greater than </li>
4613 <li><tt>oge</tt>: ordered and greater than or equal</li>
4614 <li><tt>olt</tt>: ordered and less than </li>
4615 <li><tt>ole</tt>: ordered and less than or equal</li>
4616 <li><tt>one</tt>: ordered and not equal</li>
4617 <li><tt>ord</tt>: ordered (no nans)</li>
4618 <li><tt>ueq</tt>: unordered or equal</li>
4619 <li><tt>ugt</tt>: unordered or greater than </li>
4620 <li><tt>uge</tt>: unordered or greater than or equal</li>
4621 <li><tt>ult</tt>: unordered or less than </li>
4622 <li><tt>ule</tt>: unordered or less than or equal</li>
4623 <li><tt>une</tt>: unordered or not equal</li>
4624 <li><tt>uno</tt>: unordered (either nans)</li>
4625 <li><tt>true</tt>: no comparison, always returns true</li>
4626</ol>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004627
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004628<p><i>Ordered</i> means that neither operand is a QNAN while
Bill Wendlingf85859d2009-07-20 02:29:24 +00004629 <i>unordered</i> means that either operand may be a QNAN.</p>
4630
4631<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
4632 a <a href="#t_floating">floating point</a> type or
4633 a <a href="#t_vector">vector</a> of floating point type. They must have
4634 identical types.</p>
4635
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004636<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +00004637<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004638 according to the condition code given as <tt>cond</tt>. If the operands are
4639 vectors, then the vectors are compared element by element. Each comparison
4640 performed always yields an <a href="#t_primitive">i1</a> result, as
4641 follows:</p>
4642
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004643<ol>
4644 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004645
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004646 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingf85859d2009-07-20 02:29:24 +00004647 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4648
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004649 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingf85859d2009-07-20 02:29:24 +00004650 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4651
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004652 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingf85859d2009-07-20 02:29:24 +00004653 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4654
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004655 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingf85859d2009-07-20 02:29:24 +00004656 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4657
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004658 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingf85859d2009-07-20 02:29:24 +00004659 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4660
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004661 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingf85859d2009-07-20 02:29:24 +00004662 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4663
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004664 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004665
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004666 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingf85859d2009-07-20 02:29:24 +00004667 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4668
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004669 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingf85859d2009-07-20 02:29:24 +00004670 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4671
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004672 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingf85859d2009-07-20 02:29:24 +00004673 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4674
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004675 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingf85859d2009-07-20 02:29:24 +00004676 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4677
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004678 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingf85859d2009-07-20 02:29:24 +00004679 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4680
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004681 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingf85859d2009-07-20 02:29:24 +00004682 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4683
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004684 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004685
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004686 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4687</ol>
4688
4689<h5>Example:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004690<pre>
4691 &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +00004692 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4693 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4694 &lt;result&gt; = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004695</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +00004696
4697<p>Note that the code generator does not yet support vector types with
4698 the <tt>fcmp</tt> instruction.</p>
4699
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004700</div>
4701
4702<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +00004703<div class="doc_subsubsection">
Chris Lattner6704c212008-05-20 20:48:21 +00004704 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4705</div>
4706
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004707<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +00004708
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004709<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004710<pre>
4711 &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...
4712</pre>
Chris Lattner6704c212008-05-20 20:48:21 +00004713
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004714<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004715<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in the
4716 SSA graph representing the function.</p>
4717
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004718<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004719<p>The type of the incoming values is specified with the first type field. After
4720 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
4721 one pair for each predecessor basic block of the current block. Only values
4722 of <a href="#t_firstclass">first class</a> type may be used as the value
4723 arguments to the PHI node. Only labels may be used as the label
4724 arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00004725
Bill Wendlingf85859d2009-07-20 02:29:24 +00004726<p>There must be no non-phi instructions between the start of a basic block and
4727 the PHI instructions: i.e. PHI instructions must be first in a basic
4728 block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00004729
Bill Wendlingf85859d2009-07-20 02:29:24 +00004730<p>For the purposes of the SSA form, the use of each incoming value is deemed to
4731 occur on the edge from the corresponding predecessor block to the current
4732 block (but after any definition of an '<tt>invoke</tt>' instruction's return
4733 value on the same edge).</p>
Jay Foad8e2fd2c2009-06-03 10:20:10 +00004734
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004735<h5>Semantics:</h5>
4736<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
Bill Wendlingf85859d2009-07-20 02:29:24 +00004737 specified by the pair corresponding to the predecessor basic block that
4738 executed just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +00004739
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004740<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +00004741<pre>
4742Loop: ; Infinite loop that counts from 0 on up...
4743 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4744 %nextindvar = add i32 %indvar, 1
4745 br label %Loop
4746</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004747
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004748</div>
4749
4750<!-- _______________________________________________________________________ -->
4751<div class="doc_subsubsection">
4752 <a name="i_select">'<tt>select</tt>' Instruction</a>
4753</div>
4754
4755<div class="doc_text">
4756
4757<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004758<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +00004759 &lt;result&gt; = select <i>selty</i> &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
4760
Dan Gohman2672f3e2008-10-14 16:51:45 +00004761 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004762</pre>
4763
4764<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004765<p>The '<tt>select</tt>' instruction is used to choose one value based on a
4766 condition, without branching.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004767
4768
4769<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004770<p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
4771 values indicating the condition, and two values of the
4772 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
4773 vectors and the condition is a scalar, then entire vectors are selected, not
4774 individual elements.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004775
4776<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004777<p>If the condition is an i1 and it evaluates to 1, the instruction returns the
4778 first value argument; otherwise, it returns the second value argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004779
Bill Wendlingf85859d2009-07-20 02:29:24 +00004780<p>If the condition is a vector of i1, then the value arguments must be vectors
4781 of the same size, and the selection is done element by element.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004782
4783<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004784<pre>
4785 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4786</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +00004787
4788<p>Note that the code generator does not yet support conditions
4789 with vector type.</p>
4790
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004791</div>
4792
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004793<!-- _______________________________________________________________________ -->
4794<div class="doc_subsubsection">
4795 <a name="i_call">'<tt>call</tt>' Instruction</a>
4796</div>
4797
4798<div class="doc_text">
4799
4800<h5>Syntax:</h5>
4801<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +00004802 &lt;result&gt; = [tail] call [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] &lt;ty&gt; [&lt;fnty&gt;*] &lt;fnptrval&gt;(&lt;function args&gt;) [<a href="#fnattrs">fn attrs</a>]
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004803</pre>
4804
4805<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004806<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4807
4808<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004809<p>This instruction requires several arguments:</p>
4810
4811<ol>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004812 <li>The optional "tail" marker indicates whether the callee function accesses
4813 any allocas or varargs in the caller. If the "tail" marker is present,
4814 the function call is eligible for tail call optimization. Note that calls
4815 may be marked "tail" even if they do not occur before
4816 a <a href="#i_ret"><tt>ret</tt></a> instruction.</li>
Devang Patelac2fc272008-10-06 18:50:38 +00004817
Bill Wendlingf85859d2009-07-20 02:29:24 +00004818 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
4819 convention</a> the call should use. If none is specified, the call
4820 defaults to using C calling conventions.</li>
Devang Patelac2fc272008-10-06 18:50:38 +00004821
Bill Wendlingf85859d2009-07-20 02:29:24 +00004822 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4823 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
4824 '<tt>inreg</tt>' attributes are valid here.</li>
4825
4826 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
4827 type of the return value. Functions that return no value are marked
4828 <tt><a href="#t_void">void</a></tt>.</li>
4829
4830 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
4831 being invoked. The argument types must match the types implied by this
4832 signature. This type can be omitted if the function is not varargs and if
4833 the function type does not return a pointer to a function.</li>
4834
4835 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4836 be invoked. In most cases, this is a direct function invocation, but
4837 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4838 to function value.</li>
4839
4840 <li>'<tt>function args</tt>': argument list whose types match the function
4841 signature argument types. All arguments must be of
4842 <a href="#t_firstclass">first class</a> type. If the function signature
4843 indicates the function accepts a variable number of arguments, the extra
4844 arguments can be specified.</li>
4845
4846 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
4847 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4848 '<tt>readnone</tt>' attributes are valid here.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004849</ol>
4850
4851<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004852<p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
4853 a specified function, with its incoming arguments bound to the specified
4854 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
4855 function, control flow continues with the instruction after the function
4856 call, and the return value of the function is bound to the result
4857 argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004858
4859<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004860<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +00004861 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +00004862 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4863 %X = tail call i32 @foo() <i>; yields i32</i>
4864 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4865 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +00004866
4867 %struct.A = type { i32, i8 }
Devang Patelac2fc272008-10-06 18:50:38 +00004868 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohman3e700032008-10-04 19:00:07 +00004869 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4870 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattnerac454b32008-10-08 06:26:11 +00004871 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijman2c4e05a2008-10-07 10:03:45 +00004872 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004873</pre>
4874
4875</div>
4876
4877<!-- _______________________________________________________________________ -->
4878<div class="doc_subsubsection">
4879 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4880</div>
4881
4882<div class="doc_text">
4883
4884<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004885<pre>
4886 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4887</pre>
4888
4889<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004890<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
Bill Wendlingf85859d2009-07-20 02:29:24 +00004891 the "variable argument" area of a function call. It is used to implement the
4892 <tt>va_arg</tt> macro in C.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004893
4894<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004895<p>This instruction takes a <tt>va_list*</tt> value and the type of the
4896 argument. It returns a value of the specified argument type and increments
4897 the <tt>va_list</tt> to point to the next argument. The actual type
4898 of <tt>va_list</tt> is target specific.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004899
4900<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004901<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
4902 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
4903 to the next argument. For more information, see the variable argument
4904 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004905
4906<p>It is legal for this instruction to be called in a function which does not
Bill Wendlingf85859d2009-07-20 02:29:24 +00004907 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4908 function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004909
Bill Wendlingf85859d2009-07-20 02:29:24 +00004910<p><tt>va_arg</tt> is an LLVM instruction instead of
4911 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
4912 argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004913
4914<h5>Example:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004915<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4916
Bill Wendlingf85859d2009-07-20 02:29:24 +00004917<p>Note that the code generator does not yet fully support va_arg on many
4918 targets. Also, it does not currently support va_arg with aggregate types on
4919 any target.</p>
Dan Gohman60967192009-01-12 23:12:39 +00004920
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004921</div>
4922
4923<!-- *********************************************************************** -->
4924<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4925<!-- *********************************************************************** -->
4926
4927<div class="doc_text">
4928
4929<p>LLVM supports the notion of an "intrinsic function". These functions have
Bill Wendlingf85859d2009-07-20 02:29:24 +00004930 well known names and semantics and are required to follow certain
4931 restrictions. Overall, these intrinsics represent an extension mechanism for
4932 the LLVM language that does not require changing all of the transformations
4933 in LLVM when adding to the language (or the bitcode reader/writer, the
4934 parser, etc...).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004935
4936<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
Bill Wendlingf85859d2009-07-20 02:29:24 +00004937 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4938 begin with this prefix. Intrinsic functions must always be external
4939 functions: you cannot define the body of intrinsic functions. Intrinsic
4940 functions may only be used in call or invoke instructions: it is illegal to
4941 take the address of an intrinsic function. Additionally, because intrinsic
4942 functions are part of the LLVM language, it is required if any are added that
4943 they be documented here.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004944
Bill Wendlingf85859d2009-07-20 02:29:24 +00004945<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
4946 family of functions that perform the same operation but on different data
4947 types. Because LLVM can represent over 8 million different integer types,
4948 overloading is used commonly to allow an intrinsic function to operate on any
4949 integer type. One or more of the argument types or the result type can be
4950 overloaded to accept any integer type. Argument types may also be defined as
4951 exactly matching a previous argument's type or the result type. This allows
4952 an intrinsic function which accepts multiple arguments, but needs all of them
4953 to be of the same type, to only be overloaded with respect to a single
4954 argument or the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004955
Bill Wendlingf85859d2009-07-20 02:29:24 +00004956<p>Overloaded intrinsics will have the names of its overloaded argument types
4957 encoded into its function name, each preceded by a period. Only those types
4958 which are overloaded result in a name suffix. Arguments whose type is matched
4959 against another type do not. For example, the <tt>llvm.ctpop</tt> function
4960 can take an integer of any width and returns an integer of exactly the same
4961 integer width. This leads to a family of functions such as
4962 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
4963 %val)</tt>. Only one type, the return type, is overloaded, and only one type
4964 suffix is required. Because the argument's type is matched against the return
4965 type, it does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004966
4967<p>To learn how to add an intrinsic function, please see the
Bill Wendlingf85859d2009-07-20 02:29:24 +00004968 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004969
4970</div>
4971
4972<!-- ======================================================================= -->
4973<div class="doc_subsection">
4974 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4975</div>
4976
4977<div class="doc_text">
4978
Bill Wendlingf85859d2009-07-20 02:29:24 +00004979<p>Variable argument support is defined in LLVM with
4980 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4981 intrinsic functions. These functions are related to the similarly named
4982 macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004983
Bill Wendlingf85859d2009-07-20 02:29:24 +00004984<p>All of these functions operate on arguments that use a target-specific value
4985 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
4986 not define what this type is, so all transformations should be prepared to
4987 handle these functions regardless of the type used.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004988
4989<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
Bill Wendlingf85859d2009-07-20 02:29:24 +00004990 instruction and the variable argument handling intrinsic functions are
4991 used.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00004992
4993<div class="doc_code">
4994<pre>
4995define i32 @test(i32 %X, ...) {
4996 ; Initialize variable argument processing
4997 %ap = alloca i8*
4998 %ap2 = bitcast i8** %ap to i8*
4999 call void @llvm.va_start(i8* %ap2)
5000
5001 ; Read a single integer argument
5002 %tmp = va_arg i8** %ap, i32
5003
5004 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5005 %aq = alloca i8*
5006 %aq2 = bitcast i8** %aq to i8*
5007 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5008 call void @llvm.va_end(i8* %aq2)
5009
5010 ; Stop processing of arguments.
5011 call void @llvm.va_end(i8* %ap2)
5012 ret i32 %tmp
5013}
5014
5015declare void @llvm.va_start(i8*)
5016declare void @llvm.va_copy(i8*, i8*)
5017declare void @llvm.va_end(i8*)
5018</pre>
5019</div>
5020
5021</div>
5022
5023<!-- _______________________________________________________________________ -->
5024<div class="doc_subsubsection">
5025 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5026</div>
5027
5028
5029<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00005030
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005031<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005032<pre>
5033 declare void %llvm.va_start(i8* &lt;arglist&gt;)
5034</pre>
5035
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005036<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005037<p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*&lt;arglist&gt;</tt>
5038 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005039
5040<h5>Arguments:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +00005041<p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005042
5043<h5>Semantics:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +00005044<p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005045 macro available in C. In a target-dependent way, it initializes
5046 the <tt>va_list</tt> element to which the argument points, so that the next
5047 call to <tt>va_arg</tt> will produce the first variable argument passed to
5048 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5049 need to know the last argument of the function as the compiler can figure
5050 that out.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005051
5052</div>
5053
5054<!-- _______________________________________________________________________ -->
5055<div class="doc_subsubsection">
5056 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5057</div>
5058
5059<div class="doc_text">
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005060
Bill Wendlingf85859d2009-07-20 02:29:24 +00005061<h5>Syntax:</h5>
5062<pre>
5063 declare void @llvm.va_end(i8* &lt;arglist&gt;)
5064</pre>
5065
5066<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005067<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
Bill Wendlingf85859d2009-07-20 02:29:24 +00005068 which has been initialized previously
5069 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5070 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005071
5072<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005073<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5074
5075<h5>Semantics:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005076<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005077 macro available in C. In a target-dependent way, it destroys
5078 the <tt>va_list</tt> element to which the argument points. Calls
5079 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5080 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5081 with calls to <tt>llvm.va_end</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005082
5083</div>
5084
5085<!-- _______________________________________________________________________ -->
5086<div class="doc_subsubsection">
5087 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5088</div>
5089
5090<div class="doc_text">
5091
5092<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005093<pre>
5094 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
5095</pre>
5096
5097<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005098<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
Bill Wendlingf85859d2009-07-20 02:29:24 +00005099 from the source argument list to the destination argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005100
5101<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005102<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
Bill Wendlingf85859d2009-07-20 02:29:24 +00005103 The second argument is a pointer to a <tt>va_list</tt> element to copy
5104 from.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005105
5106<h5>Semantics:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005107<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005108 macro available in C. In a target-dependent way, it copies the
5109 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5110 element. This intrinsic is necessary because
5111 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5112 arbitrarily complex and require, for example, memory allocation.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005113
5114</div>
5115
5116<!-- ======================================================================= -->
5117<div class="doc_subsection">
5118 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5119</div>
5120
5121<div class="doc_text">
5122
Bill Wendlingf85859d2009-07-20 02:29:24 +00005123<p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +00005124Collection</a> (GC) requires the implementation and generation of these
Bill Wendlingf85859d2009-07-20 02:29:24 +00005125intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5126roots on the stack</a>, as well as garbage collector implementations that
5127require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5128barriers. Front-ends for type-safe garbage collected languages should generate
5129these intrinsics to make use of the LLVM garbage collectors. For more details,
5130see <a href="GarbageCollection.html">Accurate Garbage Collection with
5131LLVM</a>.</p>
Christopher Lambcfe00962007-12-17 01:00:21 +00005132
Bill Wendlingf85859d2009-07-20 02:29:24 +00005133<p>The garbage collection intrinsics only operate on objects in the generic
5134 address space (address space zero).</p>
Christopher Lambcfe00962007-12-17 01:00:21 +00005135
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005136</div>
5137
5138<!-- _______________________________________________________________________ -->
5139<div class="doc_subsubsection">
5140 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5141</div>
5142
5143<div class="doc_text">
5144
5145<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005146<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +00005147 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005148</pre>
5149
5150<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005151<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
Bill Wendlingf85859d2009-07-20 02:29:24 +00005152 the code generator, and allows some metadata to be associated with it.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005153
5154<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005155<p>The first argument specifies the address of a stack object that contains the
Bill Wendlingf85859d2009-07-20 02:29:24 +00005156 root pointer. The second pointer (which must be either a constant or a
5157 global value address) contains the meta-data to be associated with the
5158 root.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005159
5160<h5>Semantics:</h5>
Chris Lattnera7d94ba2008-04-24 05:59:56 +00005161<p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
Bill Wendlingf85859d2009-07-20 02:29:24 +00005162 location. At compile-time, the code generator generates information to allow
5163 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5164 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5165 algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005166
5167</div>
5168
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005169<!-- _______________________________________________________________________ -->
5170<div class="doc_subsubsection">
5171 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5172</div>
5173
5174<div class="doc_text">
5175
5176<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005177<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +00005178 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005179</pre>
5180
5181<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005182<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
Bill Wendlingf85859d2009-07-20 02:29:24 +00005183 locations, allowing garbage collector implementations that require read
5184 barriers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005185
5186<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005187<p>The second argument is the address to read from, which should be an address
Bill Wendlingf85859d2009-07-20 02:29:24 +00005188 allocated from the garbage collector. The first object is a pointer to the
5189 start of the referenced object, if needed by the language runtime (otherwise
5190 null).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005191
5192<h5>Semantics:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005193<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
Bill Wendlingf85859d2009-07-20 02:29:24 +00005194 instruction, but may be replaced with substantially more complex code by the
5195 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5196 may only be used in a function which <a href="#gc">specifies a GC
5197 algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005198
5199</div>
5200
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005201<!-- _______________________________________________________________________ -->
5202<div class="doc_subsubsection">
5203 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5204</div>
5205
5206<div class="doc_text">
5207
5208<h5>Syntax:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005209<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +00005210 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005211</pre>
5212
5213<h5>Overview:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005214<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
Bill Wendlingf85859d2009-07-20 02:29:24 +00005215 locations, allowing garbage collector implementations that require write
5216 barriers (such as generational or reference counting collectors).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005217
5218<h5>Arguments:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005219<p>The first argument is the reference to store, the second is the start of the
Bill Wendlingf85859d2009-07-20 02:29:24 +00005220 object to store it to, and the third is the address of the field of Obj to
5221 store to. If the runtime does not require a pointer to the object, Obj may
5222 be null.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005223
5224<h5>Semantics:</h5>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005225<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
Bill Wendlingf85859d2009-07-20 02:29:24 +00005226 instruction, but may be replaced with substantially more complex code by the
5227 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5228 may only be used in a function which <a href="#gc">specifies a GC
5229 algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005230
5231</div>
5232
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005233<!-- ======================================================================= -->
5234<div class="doc_subsection">
5235 <a name="int_codegen">Code Generator Intrinsics</a>
5236</div>
5237
5238<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00005239
5240<p>These intrinsics are provided by LLVM to expose special features that may
5241 only be implemented with code generator support.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005242
5243</div>
5244
5245<!-- _______________________________________________________________________ -->
5246<div class="doc_subsubsection">
5247 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5248</div>
5249
5250<div class="doc_text">
5251
5252<h5>Syntax:</h5>
5253<pre>
5254 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
5255</pre>
5256
5257<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005258<p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5259 target-specific value indicating the return address of the current function
5260 or one of its callers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005261
5262<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005263<p>The argument to this intrinsic indicates which function to return the address
5264 for. Zero indicates the calling function, one indicates its caller, etc.
5265 The argument is <b>required</b> to be a constant integer value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005266
5267<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005268<p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5269 indicating the return address of the specified call frame, or zero if it
5270 cannot be identified. The value returned by this intrinsic is likely to be
5271 incorrect or 0 for arguments other than zero, so it should only be used for
5272 debugging purposes.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005273
Bill Wendlingf85859d2009-07-20 02:29:24 +00005274<p>Note that calling this intrinsic does not prevent function inlining or other
5275 aggressive transformations, so the value returned may not be that of the
5276 obvious source-language caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005277
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005278</div>
5279
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005280<!-- _______________________________________________________________________ -->
5281<div class="doc_subsubsection">
5282 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5283</div>
5284
5285<div class="doc_text">
5286
5287<h5>Syntax:</h5>
5288<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +00005289 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005290</pre>
5291
5292<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005293<p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5294 target-specific frame pointer value for the specified stack frame.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005295
5296<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005297<p>The argument to this intrinsic indicates which function to return the frame
5298 pointer for. Zero indicates the calling function, one indicates its caller,
5299 etc. The argument is <b>required</b> to be a constant integer value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005300
5301<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005302<p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5303 indicating the frame address of the specified call frame, or zero if it
5304 cannot be identified. The value returned by this intrinsic is likely to be
5305 incorrect or 0 for arguments other than zero, so it should only be used for
5306 debugging purposes.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005307
Bill Wendlingf85859d2009-07-20 02:29:24 +00005308<p>Note that calling this intrinsic does not prevent function inlining or other
5309 aggressive transformations, so the value returned may not be that of the
5310 obvious source-language caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005311
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005312</div>
5313
5314<!-- _______________________________________________________________________ -->
5315<div class="doc_subsubsection">
5316 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5317</div>
5318
5319<div class="doc_text">
5320
5321<h5>Syntax:</h5>
5322<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +00005323 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005324</pre>
5325
5326<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005327<p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5328 of the function stack, for use
5329 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5330 useful for implementing language features like scoped automatic variable
5331 sized arrays in C99.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005332
5333<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005334<p>This intrinsic returns a opaque pointer value that can be passed
5335 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5336 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5337 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5338 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5339 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5340 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005341
5342</div>
5343
5344<!-- _______________________________________________________________________ -->
5345<div class="doc_subsubsection">
5346 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5347</div>
5348
5349<div class="doc_text">
5350
5351<h5>Syntax:</h5>
5352<pre>
5353 declare void @llvm.stackrestore(i8 * %ptr)
5354</pre>
5355
5356<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005357<p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5358 the function stack to the state it was in when the
5359 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5360 executed. This is useful for implementing language features like scoped
5361 automatic variable sized arrays in C99.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005362
5363<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005364<p>See the description
5365 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005366
5367</div>
5368
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005369<!-- _______________________________________________________________________ -->
5370<div class="doc_subsubsection">
5371 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5372</div>
5373
5374<div class="doc_text">
5375
5376<h5>Syntax:</h5>
5377<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +00005378 declare void @llvm.prefetch(i8* &lt;address&gt;, i32 &lt;rw&gt;, i32 &lt;locality&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005379</pre>
5380
5381<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005382<p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5383 insert a prefetch instruction if supported; otherwise, it is a noop.
5384 Prefetches have no effect on the behavior of the program but can change its
5385 performance characteristics.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005386
5387<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005388<p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5389 specifier determining if the fetch should be for a read (0) or write (1),
5390 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5391 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5392 and <tt>locality</tt> arguments must be constant integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005393
5394<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005395<p>This intrinsic does not modify the behavior of the program. In particular,
5396 prefetches cannot trap and do not produce a value. On targets that support
5397 this intrinsic, the prefetch can provide hints to the processor cache for
5398 better performance.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005399
5400</div>
5401
5402<!-- _______________________________________________________________________ -->
5403<div class="doc_subsubsection">
5404 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5405</div>
5406
5407<div class="doc_text">
5408
5409<h5>Syntax:</h5>
5410<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +00005411 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005412</pre>
5413
5414<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005415<p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5416 Counter (PC) in a region of code to simulators and other tools. The method
5417 is target specific, but it is expected that the marker will use exported
5418 symbols to transmit the PC of the marker. The marker makes no guarantees
5419 that it will remain with any specific instruction after optimizations. It is
5420 possible that the presence of a marker will inhibit optimizations. The
5421 intended use is to be inserted after optimizations to allow correlations of
5422 simulation runs.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005423
5424<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005425<p><tt>id</tt> is a numerical id identifying the marker.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005426
5427<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005428<p>This intrinsic does not modify the behavior of the program. Backends that do
5429 not support this intrinisic may ignore it.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005430
5431</div>
5432
5433<!-- _______________________________________________________________________ -->
5434<div class="doc_subsubsection">
5435 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5436</div>
5437
5438<div class="doc_text">
5439
5440<h5>Syntax:</h5>
5441<pre>
5442 declare i64 @llvm.readcyclecounter( )
5443</pre>
5444
5445<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005446<p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5447 counter register (or similar low latency, high accuracy clocks) on those
5448 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5449 should map to RPCC. As the backing counters overflow quickly (on the order
5450 of 9 seconds on alpha), this should only be used for small timings.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005451
5452<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005453<p>When directly supported, reading the cycle counter should not modify any
5454 memory. Implementations are allowed to either return a application specific
5455 value or a system wide value. On backends without support, this is lowered
5456 to a constant 0.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005457
5458</div>
5459
5460<!-- ======================================================================= -->
5461<div class="doc_subsection">
5462 <a name="int_libc">Standard C Library Intrinsics</a>
5463</div>
5464
5465<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00005466
5467<p>LLVM provides intrinsics for a few important standard C library functions.
5468 These intrinsics allow source-language front-ends to pass information about
5469 the alignment of the pointer arguments to the code generator, providing
5470 opportunity for more efficient code generation.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005471
5472</div>
5473
5474<!-- _______________________________________________________________________ -->
5475<div class="doc_subsubsection">
5476 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5477</div>
5478
5479<div class="doc_text">
5480
5481<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005482<p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
5483 integer bit width. Not all targets support all bit widths however.</p>
5484
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005485<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +00005486 declare void @llvm.memcpy.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Bill Wendlingf85859d2009-07-20 02:29:24 +00005487 i8 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattner82c2e432008-11-21 16:42:48 +00005488 declare void @llvm.memcpy.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5489 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005490 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5491 i32 &lt;len&gt;, i32 &lt;align&gt;)
5492 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5493 i64 &lt;len&gt;, i32 &lt;align&gt;)
5494</pre>
5495
5496<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005497<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5498 source location to the destination location.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005499
Bill Wendlingf85859d2009-07-20 02:29:24 +00005500<p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5501 intrinsics do not return a value, and takes an extra alignment argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005502
5503<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005504<p>The first argument is a pointer to the destination, the second is a pointer
5505 to the source. The third argument is an integer argument specifying the
5506 number of bytes to copy, and the fourth argument is the alignment of the
5507 source and destination locations.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005508
Bill Wendlingf85859d2009-07-20 02:29:24 +00005509<p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5510 then the caller guarantees that both the source and destination pointers are
5511 aligned to that boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005512
5513<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005514<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5515 source location to the destination location, which are not allowed to
5516 overlap. It copies "len" bytes of memory over. If the argument is known to
5517 be aligned to some boundary, this can be specified as the fourth argument,
5518 otherwise it should be set to 0 or 1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005519
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005520</div>
5521
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005522<!-- _______________________________________________________________________ -->
5523<div class="doc_subsubsection">
5524 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5525</div>
5526
5527<div class="doc_text">
5528
5529<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +00005530<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
Bill Wendlingf85859d2009-07-20 02:29:24 +00005531 width. Not all targets support all bit widths however.</p>
5532
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005533<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +00005534 declare void @llvm.memmove.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Bill Wendlingf85859d2009-07-20 02:29:24 +00005535 i8 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattner82c2e432008-11-21 16:42:48 +00005536 declare void @llvm.memmove.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5537 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005538 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5539 i32 &lt;len&gt;, i32 &lt;align&gt;)
5540 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5541 i64 &lt;len&gt;, i32 &lt;align&gt;)
5542</pre>
5543
5544<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005545<p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
5546 source location to the destination location. It is similar to the
5547 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
5548 overlap.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005549
Bill Wendlingf85859d2009-07-20 02:29:24 +00005550<p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5551 intrinsics do not return a value, and takes an extra alignment argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005552
5553<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005554<p>The first argument is a pointer to the destination, the second is a pointer
5555 to the source. The third argument is an integer argument specifying the
5556 number of bytes to copy, and the fourth argument is the alignment of the
5557 source and destination locations.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005558
Bill Wendlingf85859d2009-07-20 02:29:24 +00005559<p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5560 then the caller guarantees that the source and destination pointers are
5561 aligned to that boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005562
5563<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005564<p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
5565 source location to the destination location, which may overlap. It copies
5566 "len" bytes of memory over. If the argument is known to be aligned to some
5567 boundary, this can be specified as the fourth argument, otherwise it should
5568 be set to 0 or 1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005569
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005570</div>
5571
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005572<!-- _______________________________________________________________________ -->
5573<div class="doc_subsubsection">
5574 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5575</div>
5576
5577<div class="doc_text">
5578
5579<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +00005580<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
Bill Wendlingf85859d2009-07-20 02:29:24 +00005581 width. Not all targets support all bit widths however.</p>
5582
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005583<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +00005584 declare void @llvm.memset.i8(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
Bill Wendlingf85859d2009-07-20 02:29:24 +00005585 i8 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattner82c2e432008-11-21 16:42:48 +00005586 declare void @llvm.memset.i16(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5587 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005588 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5589 i32 &lt;len&gt;, i32 &lt;align&gt;)
5590 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5591 i64 &lt;len&gt;, i32 &lt;align&gt;)
5592</pre>
5593
5594<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005595<p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
5596 particular byte value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005597
Bill Wendlingf85859d2009-07-20 02:29:24 +00005598<p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
5599 intrinsic does not return a value, and takes an extra alignment argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005600
5601<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005602<p>The first argument is a pointer to the destination to fill, the second is the
5603 byte value to fill it with, the third argument is an integer argument
5604 specifying the number of bytes to fill, and the fourth argument is the known
5605 alignment of destination location.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005606
Bill Wendlingf85859d2009-07-20 02:29:24 +00005607<p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5608 then the caller guarantees that the destination pointer is aligned to that
5609 boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005610
5611<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005612<p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
5613 at the destination location. If the argument is known to be aligned to some
5614 boundary, this can be specified as the fourth argument, otherwise it should
5615 be set to 0 or 1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005616
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005617</div>
5618
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005619<!-- _______________________________________________________________________ -->
5620<div class="doc_subsubsection">
5621 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5622</div>
5623
5624<div class="doc_text">
5625
5626<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005627<p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5628 floating point or vector of floating point type. Not all targets support all
5629 types however.</p>
5630
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005631<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +00005632 declare float @llvm.sqrt.f32(float %Val)
5633 declare double @llvm.sqrt.f64(double %Val)
5634 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5635 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5636 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005637</pre>
5638
5639<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005640<p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5641 returning the same value as the libm '<tt>sqrt</tt>' functions would.
5642 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
5643 behavior for negative numbers other than -0.0 (which allows for better
5644 optimization, because there is no need to worry about errno being
5645 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005646
5647<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005648<p>The argument and return value are floating point numbers of the same
5649 type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005650
5651<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005652<p>This function returns the sqrt of the specified operand if it is a
5653 nonnegative floating point number.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005654
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005655</div>
5656
5657<!-- _______________________________________________________________________ -->
5658<div class="doc_subsubsection">
5659 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5660</div>
5661
5662<div class="doc_text">
5663
5664<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005665<p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5666 floating point or vector of floating point type. Not all targets support all
5667 types however.</p>
5668
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005669<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +00005670 declare float @llvm.powi.f32(float %Val, i32 %power)
5671 declare double @llvm.powi.f64(double %Val, i32 %power)
5672 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5673 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5674 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005675</pre>
5676
5677<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005678<p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5679 specified (positive or negative) power. The order of evaluation of
5680 multiplications is not defined. When a vector of floating point type is
5681 used, the second argument remains a scalar integer value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005682
5683<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005684<p>The second argument is an integer power, and the first is a value to raise to
5685 that power.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005686
5687<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005688<p>This function returns the first value raised to the second power with an
5689 unspecified sequence of rounding operations.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005690
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005691</div>
5692
Dan Gohman361079c2007-10-15 20:30:11 +00005693<!-- _______________________________________________________________________ -->
5694<div class="doc_subsubsection">
5695 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5696</div>
5697
5698<div class="doc_text">
5699
5700<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005701<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5702 floating point or vector of floating point type. Not all targets support all
5703 types however.</p>
5704
Dan Gohman361079c2007-10-15 20:30:11 +00005705<pre>
5706 declare float @llvm.sin.f32(float %Val)
5707 declare double @llvm.sin.f64(double %Val)
5708 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5709 declare fp128 @llvm.sin.f128(fp128 %Val)
5710 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5711</pre>
5712
5713<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005714<p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005715
5716<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005717<p>The argument and return value are floating point numbers of the same
5718 type.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005719
5720<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005721<p>This function returns the sine of the specified operand, returning the same
5722 values as the libm <tt>sin</tt> functions would, and handles error conditions
5723 in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005724
Dan Gohman361079c2007-10-15 20:30:11 +00005725</div>
5726
5727<!-- _______________________________________________________________________ -->
5728<div class="doc_subsubsection">
5729 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5730</div>
5731
5732<div class="doc_text">
5733
5734<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005735<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5736 floating point or vector of floating point type. Not all targets support all
5737 types however.</p>
5738
Dan Gohman361079c2007-10-15 20:30:11 +00005739<pre>
5740 declare float @llvm.cos.f32(float %Val)
5741 declare double @llvm.cos.f64(double %Val)
5742 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5743 declare fp128 @llvm.cos.f128(fp128 %Val)
5744 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5745</pre>
5746
5747<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005748<p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005749
5750<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005751<p>The argument and return value are floating point numbers of the same
5752 type.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005753
5754<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005755<p>This function returns the cosine of the specified operand, returning the same
5756 values as the libm <tt>cos</tt> functions would, and handles error conditions
5757 in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005758
Dan Gohman361079c2007-10-15 20:30:11 +00005759</div>
5760
5761<!-- _______________________________________________________________________ -->
5762<div class="doc_subsubsection">
5763 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5764</div>
5765
5766<div class="doc_text">
5767
5768<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005769<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5770 floating point or vector of floating point type. Not all targets support all
5771 types however.</p>
5772
Dan Gohman361079c2007-10-15 20:30:11 +00005773<pre>
5774 declare float @llvm.pow.f32(float %Val, float %Power)
5775 declare double @llvm.pow.f64(double %Val, double %Power)
5776 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5777 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5778 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5779</pre>
5780
5781<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005782<p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5783 specified (positive or negative) power.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005784
5785<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005786<p>The second argument is a floating point power, and the first is a value to
5787 raise to that power.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005788
5789<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005790<p>This function returns the first value raised to the second power, returning
5791 the same values as the libm <tt>pow</tt> functions would, and handles error
5792 conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +00005793
Dan Gohman361079c2007-10-15 20:30:11 +00005794</div>
5795
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005796<!-- ======================================================================= -->
5797<div class="doc_subsection">
5798 <a name="int_manip">Bit Manipulation Intrinsics</a>
5799</div>
5800
5801<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00005802
5803<p>LLVM provides intrinsics for a few important bit manipulation operations.
5804 These allow efficient code generation for some algorithms.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005805
5806</div>
5807
5808<!-- _______________________________________________________________________ -->
5809<div class="doc_subsubsection">
5810 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5811</div>
5812
5813<div class="doc_text">
5814
5815<h5>Syntax:</h5>
5816<p>This is an overloaded intrinsic function. You can use bswap on any integer
Bill Wendlingf85859d2009-07-20 02:29:24 +00005817 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5818
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005819<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +00005820 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5821 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5822 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005823</pre>
5824
5825<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005826<p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5827 values with an even number of bytes (positive multiple of 16 bits). These
5828 are useful for performing operations on data that is not in the target's
5829 native byte order.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005830
5831<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005832<p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5833 and low byte of the input i16 swapped. Similarly,
5834 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
5835 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
5836 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
5837 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
5838 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
5839 more, respectively).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005840
5841</div>
5842
5843<!-- _______________________________________________________________________ -->
5844<div class="doc_subsubsection">
5845 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5846</div>
5847
5848<div class="doc_text">
5849
5850<h5>Syntax:</h5>
5851<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
Bill Wendlingf85859d2009-07-20 02:29:24 +00005852 width. Not all targets support all bit widths however.</p>
5853
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005854<pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005855 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +00005856 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005857 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +00005858 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5859 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005860</pre>
5861
5862<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005863<p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
5864 in a value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005865
5866<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005867<p>The only argument is the value to be counted. The argument may be of any
5868 integer type. The return type must match the argument type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005869
5870<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005871<p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005872
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005873</div>
5874
5875<!-- _______________________________________________________________________ -->
5876<div class="doc_subsubsection">
5877 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5878</div>
5879
5880<div class="doc_text">
5881
5882<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005883<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5884 integer bit width. Not all targets support all bit widths however.</p>
5885
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005886<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +00005887 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5888 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005889 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +00005890 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5891 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005892</pre>
5893
5894<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005895<p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5896 leading zeros in a variable.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005897
5898<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005899<p>The only argument is the value to be counted. The argument may be of any
5900 integer type. The return type must match the argument type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005901
5902<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005903<p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
5904 zeros in a variable. If the src == 0 then the result is the size in bits of
5905 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005906
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005907</div>
5908
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005909<!-- _______________________________________________________________________ -->
5910<div class="doc_subsubsection">
5911 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5912</div>
5913
5914<div class="doc_text">
5915
5916<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005917<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5918 integer bit width. Not all targets support all bit widths however.</p>
5919
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005920<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +00005921 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5922 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005923 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +00005924 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5925 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005926</pre>
5927
5928<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005929<p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5930 trailing zeros.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005931
5932<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005933<p>The only argument is the value to be counted. The argument may be of any
5934 integer type. The return type must match the argument type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005935
5936<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005937<p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
5938 zeros in a variable. If the src == 0 then the result is the size in bits of
5939 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005940
Dan Gohmanf17a25c2007-07-18 16:29:46 +00005941</div>
5942
Bill Wendling3e1258b2009-02-08 04:04:40 +00005943<!-- ======================================================================= -->
5944<div class="doc_subsection">
5945 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
5946</div>
5947
5948<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00005949
5950<p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
Bill Wendling3e1258b2009-02-08 04:04:40 +00005951
5952</div>
5953
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005954<!-- _______________________________________________________________________ -->
5955<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +00005956 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005957</div>
5958
5959<div class="doc_text">
5960
5961<h5>Syntax:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005962<p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00005963 on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005964
5965<pre>
5966 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
5967 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
5968 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
5969</pre>
5970
5971<h5>Overview:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005972<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00005973 a signed addition of the two arguments, and indicate whether an overflow
5974 occurred during the signed summation.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005975
5976<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005977<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingf85859d2009-07-20 02:29:24 +00005978 be of integer types of any bit width, but they must have the same bit
5979 width. The second element of the result structure must be of
5980 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
5981 undergo signed addition.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005982
5983<h5>Semantics:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005984<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00005985 a signed addition of the two variables. They return a structure &mdash; the
5986 first element of which is the signed summation, and the second element of
5987 which is a bit specifying if the signed summation resulted in an
5988 overflow.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00005989
5990<h5>Examples:</h5>
5991<pre>
5992 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
5993 %sum = extractvalue {i32, i1} %res, 0
5994 %obit = extractvalue {i32, i1} %res, 1
5995 br i1 %obit, label %overflow, label %normal
5996</pre>
5997
5998</div>
5999
6000<!-- _______________________________________________________________________ -->
6001<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +00006002 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006003</div>
6004
6005<div class="doc_text">
6006
6007<h5>Syntax:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006008<p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006009 on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006010
6011<pre>
6012 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6013 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6014 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6015</pre>
6016
6017<h5>Overview:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006018<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006019 an unsigned addition of the two arguments, and indicate whether a carry
6020 occurred during the unsigned summation.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006021
6022<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006023<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingf85859d2009-07-20 02:29:24 +00006024 be of integer types of any bit width, but they must have the same bit
6025 width. The second element of the result structure must be of
6026 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6027 undergo unsigned addition.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006028
6029<h5>Semantics:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006030<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006031 an unsigned addition of the two arguments. They return a structure &mdash;
6032 the first element of which is the sum, and the second element of which is a
6033 bit specifying if the unsigned summation resulted in a carry.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006034
6035<h5>Examples:</h5>
6036<pre>
6037 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6038 %sum = extractvalue {i32, i1} %res, 0
6039 %obit = extractvalue {i32, i1} %res, 1
6040 br i1 %obit, label %carry, label %normal
6041</pre>
6042
6043</div>
6044
6045<!-- _______________________________________________________________________ -->
6046<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +00006047 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006048</div>
6049
6050<div class="doc_text">
6051
6052<h5>Syntax:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006053<p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006054 on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006055
6056<pre>
6057 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6058 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6059 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6060</pre>
6061
6062<h5>Overview:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006063<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006064 a signed subtraction of the two arguments, and indicate whether an overflow
6065 occurred during the signed subtraction.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006066
6067<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006068<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingf85859d2009-07-20 02:29:24 +00006069 be of integer types of any bit width, but they must have the same bit
6070 width. The second element of the result structure must be of
6071 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6072 undergo signed subtraction.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006073
6074<h5>Semantics:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006075<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006076 a signed subtraction of the two arguments. They return a structure &mdash;
6077 the first element of which is the subtraction, and the second element of
6078 which is a bit specifying if the signed subtraction resulted in an
6079 overflow.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006080
6081<h5>Examples:</h5>
6082<pre>
6083 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6084 %sum = extractvalue {i32, i1} %res, 0
6085 %obit = extractvalue {i32, i1} %res, 1
6086 br i1 %obit, label %overflow, label %normal
6087</pre>
6088
6089</div>
6090
6091<!-- _______________________________________________________________________ -->
6092<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +00006093 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006094</div>
6095
6096<div class="doc_text">
6097
6098<h5>Syntax:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006099<p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006100 on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006101
6102<pre>
6103 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6104 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6105 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6106</pre>
6107
6108<h5>Overview:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006109<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006110 an unsigned subtraction of the two arguments, and indicate whether an
6111 overflow occurred during the unsigned subtraction.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006112
6113<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006114<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingf85859d2009-07-20 02:29:24 +00006115 be of integer types of any bit width, but they must have the same bit
6116 width. The second element of the result structure must be of
6117 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6118 undergo unsigned subtraction.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006119
6120<h5>Semantics:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006121<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006122 an unsigned subtraction of the two arguments. They return a structure &mdash;
6123 the first element of which is the subtraction, and the second element of
6124 which is a bit specifying if the unsigned subtraction resulted in an
6125 overflow.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006126
6127<h5>Examples:</h5>
6128<pre>
6129 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6130 %sum = extractvalue {i32, i1} %res, 0
6131 %obit = extractvalue {i32, i1} %res, 1
6132 br i1 %obit, label %overflow, label %normal
6133</pre>
6134
6135</div>
6136
6137<!-- _______________________________________________________________________ -->
6138<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +00006139 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006140</div>
6141
6142<div class="doc_text">
6143
6144<h5>Syntax:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006145<p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006146 on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006147
6148<pre>
6149 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6150 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6151 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6152</pre>
6153
6154<h5>Overview:</h5>
6155
6156<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006157 a signed multiplication of the two arguments, and indicate whether an
6158 overflow occurred during the signed multiplication.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006159
6160<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006161<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingf85859d2009-07-20 02:29:24 +00006162 be of integer types of any bit width, but they must have the same bit
6163 width. The second element of the result structure must be of
6164 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6165 undergo signed multiplication.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006166
6167<h5>Semantics:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006168<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006169 a signed multiplication of the two arguments. They return a structure &mdash;
6170 the first element of which is the multiplication, and the second element of
6171 which is a bit specifying if the signed multiplication resulted in an
6172 overflow.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +00006173
6174<h5>Examples:</h5>
6175<pre>
6176 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6177 %sum = extractvalue {i32, i1} %res, 0
6178 %obit = extractvalue {i32, i1} %res, 1
6179 br i1 %obit, label %overflow, label %normal
6180</pre>
6181
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006182</div>
6183
Bill Wendlingbda98b62009-02-08 23:00:09 +00006184<!-- _______________________________________________________________________ -->
6185<div class="doc_subsubsection">
6186 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6187</div>
6188
6189<div class="doc_text">
6190
6191<h5>Syntax:</h5>
Bill Wendlingbda98b62009-02-08 23:00:09 +00006192<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006193 on any integer bit width.</p>
Bill Wendlingbda98b62009-02-08 23:00:09 +00006194
6195<pre>
6196 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6197 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6198 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6199</pre>
6200
6201<h5>Overview:</h5>
Bill Wendlingbda98b62009-02-08 23:00:09 +00006202<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006203 a unsigned multiplication of the two arguments, and indicate whether an
6204 overflow occurred during the unsigned multiplication.</p>
Bill Wendlingbda98b62009-02-08 23:00:09 +00006205
6206<h5>Arguments:</h5>
Bill Wendlingbda98b62009-02-08 23:00:09 +00006207<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingf85859d2009-07-20 02:29:24 +00006208 be of integer types of any bit width, but they must have the same bit
6209 width. The second element of the result structure must be of
6210 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6211 undergo unsigned multiplication.</p>
Bill Wendlingbda98b62009-02-08 23:00:09 +00006212
6213<h5>Semantics:</h5>
Bill Wendlingbda98b62009-02-08 23:00:09 +00006214<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingf85859d2009-07-20 02:29:24 +00006215 an unsigned multiplication of the two arguments. They return a structure
6216 &mdash; the first element of which is the multiplication, and the second
6217 element of which is a bit specifying if the unsigned multiplication resulted
6218 in an overflow.</p>
Bill Wendlingbda98b62009-02-08 23:00:09 +00006219
6220<h5>Examples:</h5>
6221<pre>
6222 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6223 %sum = extractvalue {i32, i1} %res, 0
6224 %obit = extractvalue {i32, i1} %res, 1
6225 br i1 %obit, label %overflow, label %normal
6226</pre>
6227
6228</div>
6229
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006230<!-- ======================================================================= -->
6231<div class="doc_subsection">
6232 <a name="int_debugger">Debugger Intrinsics</a>
6233</div>
6234
6235<div class="doc_text">
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006236
Bill Wendlingf85859d2009-07-20 02:29:24 +00006237<p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6238 prefix), are described in
6239 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6240 Level Debugging</a> document.</p>
6241
6242</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006243
6244<!-- ======================================================================= -->
6245<div class="doc_subsection">
6246 <a name="int_eh">Exception Handling Intrinsics</a>
6247</div>
6248
6249<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006250
6251<p>The LLVM exception handling intrinsics (which all start with
6252 <tt>llvm.eh.</tt> prefix), are described in
6253 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6254 Handling</a> document.</p>
6255
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006256</div>
6257
6258<!-- ======================================================================= -->
6259<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +00006260 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +00006261</div>
6262
6263<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006264
6265<p>This intrinsic makes it possible to excise one parameter, marked with
6266 the <tt>nest</tt> attribute, from a function. The result is a callable
6267 function pointer lacking the nest parameter - the caller does not need to
6268 provide a value for it. Instead, the value to use is stored in advance in a
6269 "trampoline", a block of memory usually allocated on the stack, which also
6270 contains code to splice the nest value into the argument list. This is used
6271 to implement the GCC nested function address extension.</p>
6272
6273<p>For example, if the function is
6274 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6275 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6276 follows:</p>
6277
6278<div class="doc_code">
Duncan Sands38947cd2007-07-27 12:58:54 +00006279<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +00006280 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6281 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6282 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6283 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +00006284</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006285</div>
6286
6287<p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6288 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6289
Duncan Sands38947cd2007-07-27 12:58:54 +00006290</div>
6291
6292<!-- _______________________________________________________________________ -->
6293<div class="doc_subsubsection">
6294 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6295</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006296
Duncan Sands38947cd2007-07-27 12:58:54 +00006297<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006298
Duncan Sands38947cd2007-07-27 12:58:54 +00006299<h5>Syntax:</h5>
6300<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006301 declare i8* @llvm.init.trampoline(i8* &lt;tramp&gt;, i8* &lt;func&gt;, i8* &lt;nval&gt;)
Duncan Sands38947cd2007-07-27 12:58:54 +00006302</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006303
Duncan Sands38947cd2007-07-27 12:58:54 +00006304<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006305<p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6306 function pointer suitable for executing it.</p>
6307
Duncan Sands38947cd2007-07-27 12:58:54 +00006308<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006309<p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6310 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6311 sufficiently aligned block of memory; this memory is written to by the
6312 intrinsic. Note that the size and the alignment are target-specific - LLVM
6313 currently provides no portable way of determining them, so a front-end that
6314 generates this intrinsic needs to have some target-specific knowledge.
6315 The <tt>func</tt> argument must hold a function bitcast to
6316 an <tt>i8*</tt>.</p>
6317
Duncan Sands38947cd2007-07-27 12:58:54 +00006318<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006319<p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6320 dependent code, turning it into a function. A pointer to this function is
6321 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6322 function pointer type</a> before being called. The new function's signature
6323 is the same as that of <tt>func</tt> with any arguments marked with
6324 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6325 is allowed, and it must be of pointer type. Calling the new function is
6326 equivalent to calling <tt>func</tt> with the same argument list, but
6327 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6328 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6329 by <tt>tramp</tt> is modified, then the effect of any later call to the
6330 returned function pointer is undefined.</p>
6331
Duncan Sands38947cd2007-07-27 12:58:54 +00006332</div>
6333
6334<!-- ======================================================================= -->
6335<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +00006336 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6337</div>
6338
6339<div class="doc_text">
Andrew Lenharth785610d2008-02-16 01:24:58 +00006340
Bill Wendlingf85859d2009-07-20 02:29:24 +00006341<p>These intrinsic functions expand the "universal IR" of LLVM to represent
6342 hardware constructs for atomic operations and memory synchronization. This
6343 provides an interface to the hardware, not an interface to the programmer. It
6344 is aimed at a low enough level to allow any programming models or APIs
6345 (Application Programming Interfaces) which need atomic behaviors to map
6346 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6347 hardware provides a "universal IR" for source languages, it also provides a
6348 starting point for developing a "universal" atomic operation and
6349 synchronization IR.</p>
6350
6351<p>These do <em>not</em> form an API such as high-level threading libraries,
6352 software transaction memory systems, atomic primitives, and intrinsic
6353 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6354 application libraries. The hardware interface provided by LLVM should allow
6355 a clean implementation of all of these APIs and parallel programming models.
6356 No one model or paradigm should be selected above others unless the hardware
6357 itself ubiquitously does so.</p>
6358
Andrew Lenharth785610d2008-02-16 01:24:58 +00006359</div>
6360
6361<!-- _______________________________________________________________________ -->
6362<div class="doc_subsubsection">
6363 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6364</div>
6365<div class="doc_text">
6366<h5>Syntax:</h5>
6367<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006368 declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;, i1 &lt;device&gt; )
Andrew Lenharth785610d2008-02-16 01:24:58 +00006369</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006370
Andrew Lenharth785610d2008-02-16 01:24:58 +00006371<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006372<p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6373 specific pairs of memory access types.</p>
6374
Andrew Lenharth785610d2008-02-16 01:24:58 +00006375<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006376<p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6377 The first four arguments enables a specific barrier as listed below. The
6378 fith argument specifies that the barrier applies to io or device or uncached
6379 memory.</p>
Andrew Lenharth785610d2008-02-16 01:24:58 +00006380
Bill Wendlingf85859d2009-07-20 02:29:24 +00006381<ul>
6382 <li><tt>ll</tt>: load-load barrier</li>
6383 <li><tt>ls</tt>: load-store barrier</li>
6384 <li><tt>sl</tt>: store-load barrier</li>
6385 <li><tt>ss</tt>: store-store barrier</li>
6386 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6387</ul>
6388
Andrew Lenharth785610d2008-02-16 01:24:58 +00006389<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006390<p>This intrinsic causes the system to enforce some ordering constraints upon
6391 the loads and stores of the program. This barrier does not
6392 indicate <em>when</em> any events will occur, it only enforces
6393 an <em>order</em> in which they occur. For any of the specified pairs of load
6394 and store operations (f.ex. load-load, or store-load), all of the first
6395 operations preceding the barrier will complete before any of the second
6396 operations succeeding the barrier begin. Specifically the semantics for each
6397 pairing is as follows:</p>
Andrew Lenharth785610d2008-02-16 01:24:58 +00006398
Bill Wendlingf85859d2009-07-20 02:29:24 +00006399<ul>
6400 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6401 after the barrier begins.</li>
6402 <li><tt>ls</tt>: All loads before the barrier must complete before any
6403 store after the barrier begins.</li>
6404 <li><tt>ss</tt>: All stores before the barrier must complete before any
6405 store after the barrier begins.</li>
6406 <li><tt>sl</tt>: All stores before the barrier must complete before any
6407 load after the barrier begins.</li>
6408</ul>
6409
6410<p>These semantics are applied with a logical "and" behavior when more than one
6411 is enabled in a single memory barrier intrinsic.</p>
6412
6413<p>Backends may implement stronger barriers than those requested when they do
6414 not support as fine grained a barrier as requested. Some architectures do
6415 not need all types of barriers and on such architectures, these become
6416 noops.</p>
6417
Andrew Lenharth785610d2008-02-16 01:24:58 +00006418<h5>Example:</h5>
6419<pre>
6420%ptr = malloc i32
6421 store i32 4, %ptr
6422
6423%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6424 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6425 <i>; guarantee the above finishes</i>
6426 store i32 8, %ptr <i>; before this begins</i>
6427</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006428
Andrew Lenharth785610d2008-02-16 01:24:58 +00006429</div>
6430
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006431<!-- _______________________________________________________________________ -->
6432<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006433 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006434</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006435
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006436<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006437
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006438<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006439<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6440 any integer bit width and for different address spaces. Not all targets
6441 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006442
6443<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006444 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
6445 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
6446 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
6447 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;cmp&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006448</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006449
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006450<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006451<p>This loads a value in memory and compares it to a given value. If they are
6452 equal, it stores a new value into the memory.</p>
6453
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006454<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006455<p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
6456 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6457 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6458 this integer type. While any bit width integer may be used, targets may only
6459 lower representations they support in hardware.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006460
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006461<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006462<p>This entire intrinsic must be executed atomically. It first loads the value
6463 in memory pointed to by <tt>ptr</tt> and compares it with the
6464 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
6465 memory. The loaded value is yielded in all cases. This provides the
6466 equivalent of an atomic compare-and-swap operation within the SSA
6467 framework.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006468
Bill Wendlingf85859d2009-07-20 02:29:24 +00006469<h5>Examples:</h5>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006470<pre>
6471%ptr = malloc i32
6472 store i32 4, %ptr
6473
6474%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +00006475%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006476 <i>; yields {i32}:result1 = 4</i>
6477%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6478%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6479
6480%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +00006481%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006482 <i>; yields {i32}:result2 = 8</i>
6483%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6484
6485%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6486</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006487
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006488</div>
6489
6490<!-- _______________________________________________________________________ -->
6491<div class="doc_subsubsection">
6492 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6493</div>
6494<div class="doc_text">
6495<h5>Syntax:</h5>
6496
Bill Wendlingf85859d2009-07-20 02:29:24 +00006497<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6498 integer bit width. Not all targets support all bit widths however.</p>
6499
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006500<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006501 declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
6502 declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
6503 declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
6504 declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006505</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006506
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006507<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006508<p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6509 the value from memory. It then stores the value in <tt>val</tt> in the memory
6510 at <tt>ptr</tt>.</p>
6511
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006512<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006513<p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
6514 the <tt>val</tt> argument and the result must be integers of the same bit
6515 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6516 integer type. The targets may only lower integer representations they
6517 support.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006518
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006519<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006520<p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6521 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6522 equivalent of an atomic swap operation within the SSA framework.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006523
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006524<h5>Examples:</h5>
6525<pre>
6526%ptr = malloc i32
6527 store i32 4, %ptr
6528
6529%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +00006530%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006531 <i>; yields {i32}:result1 = 4</i>
6532%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6533%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6534
6535%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +00006536%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006537 <i>; yields {i32}:result2 = 8</i>
6538
6539%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6540%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6541</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006542
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006543</div>
6544
6545<!-- _______________________________________________________________________ -->
6546<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006547 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006548
6549</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006550
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006551<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006552
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006553<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006554<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
6555 any integer bit width. Not all targets support all bit widths however.</p>
6556
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006557<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006558 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6559 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6560 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6561 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006562</pre>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006563
Bill Wendlingf85859d2009-07-20 02:29:24 +00006564<h5>Overview:</h5>
6565<p>This intrinsic adds <tt>delta</tt> to the value stored in memory
6566 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6567
6568<h5>Arguments:</h5>
6569<p>The intrinsic takes two arguments, the first a pointer to an integer value
6570 and the second an integer value. The result is also an integer value. These
6571 integer types can have any bit width, but they must all have the same bit
6572 width. The targets may only lower integer representations they support.</p>
6573
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006574<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006575<p>This intrinsic does a series of operations atomically. It first loads the
6576 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6577 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006578
6579<h5>Examples:</h5>
6580<pre>
6581%ptr = malloc i32
6582 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +00006583%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006584 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006585%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006586 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006587%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006588 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006589%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006590</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006591
Andrew Lenharthe44f3902008-02-21 06:45:13 +00006592</div>
6593
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006594<!-- _______________________________________________________________________ -->
6595<div class="doc_subsubsection">
6596 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6597
6598</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006599
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006600<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006601
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006602<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006603<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6604 any integer bit width and for different address spaces. Not all targets
6605 support all bit widths however.</p>
6606
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006607<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006608 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6609 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6610 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6611 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006612</pre>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006613
Bill Wendlingf85859d2009-07-20 02:29:24 +00006614<h5>Overview:</h5>
6615<p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6616 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6617
6618<h5>Arguments:</h5>
6619<p>The intrinsic takes two arguments, the first a pointer to an integer value
6620 and the second an integer value. The result is also an integer value. These
6621 integer types can have any bit width, but they must all have the same bit
6622 width. The targets may only lower integer representations they support.</p>
6623
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006624<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006625<p>This intrinsic does a series of operations atomically. It first loads the
6626 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6627 result to <tt>ptr</tt>. It yields the original value stored
6628 at <tt>ptr</tt>.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006629
6630<h5>Examples:</h5>
6631<pre>
6632%ptr = malloc i32
6633 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +00006634%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006635 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006636%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006637 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006638%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006639 <i>; yields {i32}:result3 = 2</i>
6640%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6641</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006642
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006643</div>
6644
6645<!-- _______________________________________________________________________ -->
6646<div class="doc_subsubsection">
6647 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6648 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6649 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6650 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006651</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006652
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006653<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006654
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006655<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006656<p>These are overloaded intrinsics. You can
6657 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
6658 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
6659 bit width and for different address spaces. Not all targets support all bit
6660 widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006661
Bill Wendlingf85859d2009-07-20 02:29:24 +00006662<pre>
6663 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6664 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6665 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6666 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006667</pre>
6668
6669<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006670 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6671 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6672 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6673 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006674</pre>
6675
6676<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006677 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6678 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6679 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6680 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006681</pre>
6682
6683<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006684 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6685 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6686 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6687 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006688</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006689
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006690<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006691<p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6692 the value stored in memory at <tt>ptr</tt>. It yields the original value
6693 at <tt>ptr</tt>.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006694
Bill Wendlingf85859d2009-07-20 02:29:24 +00006695<h5>Arguments:</h5>
6696<p>These intrinsics take two arguments, the first a pointer to an integer value
6697 and the second an integer value. The result is also an integer value. These
6698 integer types can have any bit width, but they must all have the same bit
6699 width. The targets may only lower integer representations they support.</p>
6700
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006701<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006702<p>These intrinsics does a series of operations atomically. They first load the
6703 value stored at <tt>ptr</tt>. They then do the bitwise
6704 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
6705 original value stored at <tt>ptr</tt>.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006706
6707<h5>Examples:</h5>
6708<pre>
6709%ptr = malloc i32
6710 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +00006711%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006712 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006713%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006714 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006715%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006716 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006717%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006718 <i>; yields {i32}:result3 = FF</i>
6719%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6720</pre>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006721
Bill Wendlingf85859d2009-07-20 02:29:24 +00006722</div>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006723
6724<!-- _______________________________________________________________________ -->
6725<div class="doc_subsubsection">
6726 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6727 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6728 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6729 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006730</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006731
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006732<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006733
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006734<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006735<p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6736 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6737 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6738 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006739
Bill Wendlingf85859d2009-07-20 02:29:24 +00006740<pre>
6741 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6742 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6743 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6744 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006745</pre>
6746
6747<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006748 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6749 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6750 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6751 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006752</pre>
6753
6754<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006755 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6756 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6757 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6758 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006759</pre>
6760
6761<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006762 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6763 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6764 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6765 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006766</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006767
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006768<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006769<p>These intrinsics takes the signed or unsigned minimum or maximum of
6770 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6771 original value at <tt>ptr</tt>.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006772
Bill Wendlingf85859d2009-07-20 02:29:24 +00006773<h5>Arguments:</h5>
6774<p>These intrinsics take two arguments, the first a pointer to an integer value
6775 and the second an integer value. The result is also an integer value. These
6776 integer types can have any bit width, but they must all have the same bit
6777 width. The targets may only lower integer representations they support.</p>
6778
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006779<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006780<p>These intrinsics does a series of operations atomically. They first load the
6781 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
6782 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
6783 yield the original value stored at <tt>ptr</tt>.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006784
6785<h5>Examples:</h5>
6786<pre>
6787%ptr = malloc i32
6788 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +00006789%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006790 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006791%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006792 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006793%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006794 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +00006795%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006796 <i>; yields {i32}:result3 = 8</i>
6797%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6798</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006799
Mon P Wang6bde9ec2008-06-25 08:15:39 +00006800</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +00006801
6802<!-- ======================================================================= -->
6803<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006804 <a name="int_general">General Intrinsics</a>
6805</div>
6806
6807<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006808
6809<p>This class of intrinsics is designed to be generic and has no specific
6810 purpose.</p>
6811
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006812</div>
6813
6814<!-- _______________________________________________________________________ -->
6815<div class="doc_subsubsection">
6816 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6817</div>
6818
6819<div class="doc_text">
6820
6821<h5>Syntax:</h5>
6822<pre>
6823 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6824</pre>
6825
6826<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006827<p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006828
6829<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006830<p>The first argument is a pointer to a value, the second is a pointer to a
6831 global string, the third is a pointer to a global string which is the source
6832 file name, and the last argument is the line number.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006833
6834<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006835<p>This intrinsic allows annotation of local variables with arbitrary strings.
6836 This can be useful for special purpose optimizations that want to look for
6837 these annotations. These have no other defined use, they are ignored by code
6838 generation and optimization.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006839
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006840</div>
6841
Tanya Lattnerb306a9e2007-09-21 22:59:12 +00006842<!-- _______________________________________________________________________ -->
6843<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +00006844 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +00006845</div>
6846
6847<div class="doc_text">
6848
6849<h5>Syntax:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006850<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6851 any integer bit width.</p>
6852
Tanya Lattnerb306a9e2007-09-21 22:59:12 +00006853<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +00006854 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6855 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6856 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6857 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6858 declare i256 @llvm.annotation.i256(i256 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
Tanya Lattnerb306a9e2007-09-21 22:59:12 +00006859</pre>
6860
6861<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006862<p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +00006863
6864<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006865<p>The first argument is an integer value (result of some expression), the
6866 second is a pointer to a global string, the third is a pointer to a global
6867 string which is the source file name, and the last argument is the line
6868 number. It returns the value of the first argument.</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +00006869
6870<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006871<p>This intrinsic allows annotations to be put on arbitrary expressions with
6872 arbitrary strings. This can be useful for special purpose optimizations that
6873 want to look for these annotations. These have no other defined use, they
6874 are ignored by code generation and optimization.</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +00006875
Tanya Lattnerb306a9e2007-09-21 22:59:12 +00006876</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006877
Anton Korobeynikove6e764f2008-01-15 22:31:34 +00006878<!-- _______________________________________________________________________ -->
6879<div class="doc_subsubsection">
6880 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6881</div>
6882
6883<div class="doc_text">
6884
6885<h5>Syntax:</h5>
6886<pre>
6887 declare void @llvm.trap()
6888</pre>
6889
6890<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006891<p>The '<tt>llvm.trap</tt>' intrinsic.</p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +00006892
6893<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006894<p>None.</p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +00006895
6896<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006897<p>This intrinsics is lowered to the target dependent trap instruction. If the
6898 target does not have a trap instruction, this intrinsic will be lowered to
6899 the call of the <tt>abort()</tt> function.</p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +00006900
Anton Korobeynikove6e764f2008-01-15 22:31:34 +00006901</div>
6902
Bill Wendlinge4164592008-11-19 05:56:17 +00006903<!-- _______________________________________________________________________ -->
6904<div class="doc_subsubsection">
Misha Brukman5dd7f4d2008-11-22 23:55:29 +00006905 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
Bill Wendlinge4164592008-11-19 05:56:17 +00006906</div>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006907
Bill Wendlinge4164592008-11-19 05:56:17 +00006908<div class="doc_text">
Bill Wendlingf85859d2009-07-20 02:29:24 +00006909
Bill Wendlinge4164592008-11-19 05:56:17 +00006910<h5>Syntax:</h5>
6911<pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006912 declare void @llvm.stackprotector( i8* &lt;guard&gt;, i8** &lt;slot&gt; )
Bill Wendlinge4164592008-11-19 05:56:17 +00006913</pre>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006914
Bill Wendlinge4164592008-11-19 05:56:17 +00006915<h5>Overview:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006916<p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
6917 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
6918 ensure that it is placed on the stack before local variables.</p>
6919
Bill Wendlinge4164592008-11-19 05:56:17 +00006920<h5>Arguments:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006921<p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
6922 arguments. The first argument is the value loaded from the stack
6923 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
6924 that has enough space to hold the value of the guard.</p>
6925
Bill Wendlinge4164592008-11-19 05:56:17 +00006926<h5>Semantics:</h5>
Bill Wendlingf85859d2009-07-20 02:29:24 +00006927<p>This intrinsic causes the prologue/epilogue inserter to force the position of
6928 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6929 stack. This is to ensure that if a local variable on the stack is
6930 overwritten, it will destroy the value of the guard. When the function exits,
6931 the guard on the stack is checked against the original guard. If they're
6932 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
6933 function.</p>
6934
Bill Wendlinge4164592008-11-19 05:56:17 +00006935</div>
6936
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006937<!-- *********************************************************************** -->
6938<hr>
6939<address>
6940 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +00006941 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006942 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +00006943 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006944
6945 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6946 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6947 Last modified: $Date$
6948</address>
Chris Lattner08497ce2008-01-04 04:33:49 +00006949
Dan Gohmanf17a25c2007-07-18 16:29:46 +00006950</body>
6951</html>