Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 8312e2965612848d57f8f15e70f258c3075df2f3 / . / docs / LangRef.html
blob: fb052d85eb357fd6143790f932b348d25dd89108 [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1<!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>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]25 <li><a href="#namedtypes">Named Types</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]26 <li><a href="#globalvars">Global Variables</a></li>
27 <li><a href="#functionstructure">Functions</a></li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]28 <li><a href="#aliasstructure">Aliases</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]29 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]30 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]31 <li><a href="#gc">Garbage Collector Names</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]32 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
33 <li><a href="#datalayout">Data Layout</a></li>
34 </ol>
35 </li>
36 <li><a href="#typesystem">Type System</a>
37 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]38 <li><a href="#t_classifications">Type Classifications</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]39 <li><a href="#t_primitive">Primitive Types</a>
40 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]41 <li><a href="#t_floating">Floating Point Types</a></li>
42 <li><a href="#t_void">Void Type</a></li>
43 <li><a href="#t_label">Label Type</a></li>
Nick Lewycky29aaef82009-05-30 05:06:04 +0000[diff] [blame]44 <li><a href="#t_metadata">Metadata Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]45 </ol>
46 </li>
47 <li><a href="#t_derived">Derived Types</a>
48 <ol>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]49 <li><a href="#t_integer">Integer Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]50 <li><a href="#t_array">Array Type</a></li>
51 <li><a href="#t_function">Function Type</a></li>
52 <li><a href="#t_pointer">Pointer Type</a></li>
53 <li><a href="#t_struct">Structure Type</a></li>
54 <li><a href="#t_pstruct">Packed Structure Type</a></li>
55 <li><a href="#t_vector">Vector Type</a></li>
56 <li><a href="#t_opaque">Opaque Type</a></li>
57 </ol>
58 </li>
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]59 <li><a href="#t_uprefs">Type Up-references</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]60 </ol>
61 </li>
62 <li><a href="#constants">Constants</a>
63 <ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]64 <li><a href="#simpleconstants">Simple Constants</a></li>
Chris Lattner97063852009-02-28 18:32:25 +0000[diff] [blame]65 <li><a href="#complexconstants">Complex Constants</a></li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]66 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
67 <li><a href="#undefvalues">Undefined Values</a></li>
68 <li><a href="#constantexprs">Constant Expressions</a></li>
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame]69 <li><a href="#metadata">Embedded Metadata</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]70 </ol>
71 </li>
72 <li><a href="#othervalues">Other Values</a>
73 <ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]74 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]75 </ol>
76 </li>
77 <li><a href="#instref">Instruction Reference</a>
78 <ol>
79 <li><a href="#terminators">Terminator Instructions</a>
80 <ol>
81 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
82 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
83 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
84 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
85 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
86 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
87 </ol>
88 </li>
89 <li><a href="#binaryops">Binary Operations</a>
90 <ol>
91 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
92 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
93 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
94 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
95 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
96 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
97 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
98 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
99 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
100 </ol>
101 </li>
102 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
103 <ol>
104 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
105 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
106 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
107 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
108 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
109 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
110 </ol>
111 </li>
112 <li><a href="#vectorops">Vector Operations</a>
113 <ol>
114 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
115 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
116 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
117 </ol>
118 </li>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]119 <li><a href="#aggregateops">Aggregate Operations</a>
120 <ol>
121 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
122 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
123 </ol>
124 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]125 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
126 <ol>
127 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
128 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
129 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
130 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
131 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
132 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
133 </ol>
134 </li>
135 <li><a href="#convertops">Conversion Operations</a>
136 <ol>
137 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
138 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
139 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
140 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
141 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
142 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
143 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
144 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
145 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
146 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
147 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
148 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
149 </ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]150 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]151 <li><a href="#otherops">Other Operations</a>
152 <ol>
153 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
154 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]155 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
156 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]157 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
158 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
159 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
160 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
161 </ol>
162 </li>
163 </ol>
164 </li>
165 <li><a href="#intrinsics">Intrinsic Functions</a>
166 <ol>
167 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
168 <ol>
169 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
170 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
171 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
172 </ol>
173 </li>
174 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
175 <ol>
176 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
177 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
178 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
179 </ol>
180 </li>
181 <li><a href="#int_codegen">Code Generator Intrinsics</a>
182 <ol>
183 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
184 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
185 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
186 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
187 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
188 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
189 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
190 </ol>
191 </li>
192 <li><a href="#int_libc">Standard C Library Intrinsics</a>
193 <ol>
194 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
198 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]199 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
201 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]202 </ol>
203 </li>
204 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
205 <ol>
206 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
207 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
208 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
209 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
210 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
211 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
212 </ol>
213 </li>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]214 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
215 <ol>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]216 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
217 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
218 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
219 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
220 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingbda98b62009-02-08 23:00:09 +0000[diff] [blame]221 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]222 </ol>
223 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]224 <li><a href="#int_debugger">Debugger intrinsics</a></li>
225 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]226 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]227 <ol>
228 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]229 </ol>
230 </li>
Bill Wendling9127adb2008-11-18 22:10:53 +0000[diff] [blame]231 <li><a href="#int_atomics">Atomic intrinsics</a>
232 <ol>
233 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
234 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
235 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
236 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
237 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
238 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
239 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
240 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
241 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
242 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
243 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
244 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
245 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
246 </ol>
247 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]248 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]249 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]250 <li><a href="#int_var_annotation">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]251 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]252 <li><a href="#int_annotation">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]253 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]254 <li><a href="#int_trap">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]255 '<tt>llvm.trap</tt>' Intrinsic</a></li>
256 <li><a href="#int_stackprotector">
257 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]258 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]259 </li>
260 </ol>
261 </li>
262</ol>
263
264<div class="doc_author">
265 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
266 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
267</div>
268
269<!-- *********************************************************************** -->
270<div class="doc_section"> <a name="abstract">Abstract </a></div>
271<!-- *********************************************************************** -->
272
273<div class="doc_text">
274<p>This document is a reference manual for the LLVM assembly language.
Bill Wendlinge7846a52008-08-05 22:29:16 +0000[diff] [blame]275LLVM is a Static Single Assignment (SSA) based representation that provides
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]276type safety, low-level operations, flexibility, and the capability of
277representing 'all' high-level languages cleanly. It is the common code
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]278representation used throughout all phases of the LLVM compilation
279strategy.</p>
280</div>
281
282<!-- *********************************************************************** -->
283<div class="doc_section"> <a name="introduction">Introduction</a> </div>
284<!-- *********************************************************************** -->
285
286<div class="doc_text">
287
288<p>The LLVM code representation is designed to be used in three
289different forms: as an in-memory compiler IR, as an on-disk bitcode
290representation (suitable for fast loading by a Just-In-Time compiler),
291and as a human readable assembly language representation. This allows
292LLVM to provide a powerful intermediate representation for efficient
293compiler transformations and analysis, while providing a natural means
294to debug and visualize the transformations. The three different forms
295of LLVM are all equivalent. This document describes the human readable
296representation and notation.</p>
297
298<p>The LLVM representation aims to be light-weight and low-level
299while being expressive, typed, and extensible at the same time. It
300aims to be a "universal IR" of sorts, by being at a low enough level
301that high-level ideas may be cleanly mapped to it (similar to how
302microprocessors are "universal IR's", allowing many source languages to
303be mapped to them). By providing type information, LLVM can be used as
304the target of optimizations: for example, through pointer analysis, it
305can be proven that a C automatic variable is never accessed outside of
306the current function... allowing it to be promoted to a simple SSA
307value instead of a memory location.</p>
308
309</div>
310
311<!-- _______________________________________________________________________ -->
312<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
313
314<div class="doc_text">
315
316<p>It is important to note that this document describes 'well formed'
317LLVM assembly language. There is a difference between what the parser
318accepts and what is considered 'well formed'. For example, the
319following instruction is syntactically okay, but not well formed:</p>
320
321<div class="doc_code">
322<pre>
323%x = <a href="#i_add">add</a> i32 1, %x
324</pre>
325</div>
326
327<p>...because the definition of <tt>%x</tt> does not dominate all of
328its uses. The LLVM infrastructure provides a verification pass that may
329be used to verify that an LLVM module is well formed. This pass is
330automatically run by the parser after parsing input assembly and by
331the optimizer before it outputs bitcode. The violations pointed out
332by the verifier pass indicate bugs in transformation passes or input to
333the parser.</p>
334</div>
335
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]336<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]337
338<!-- *********************************************************************** -->
339<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
340<!-- *********************************************************************** -->
341
342<div class="doc_text">
343
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]344 <p>LLVM identifiers come in two basic types: global and local. Global
345 identifiers (functions, global variables) begin with the @ character. Local
346 identifiers (register names, types) begin with the % character. Additionally,
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]347 there are three different formats for identifiers, for different purposes:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]348
349<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]350 <li>Named values are represented as a string of characters with their prefix.
351 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
352 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]353 Identifiers which require other characters in their names can be surrounded
Daniel Dunbar73f9a772008-10-14 23:51:43 +0000[diff] [blame]354 with quotes. Special characters may be escaped using "\xx" where xx is the
355 ASCII code for the character in hexadecimal. In this way, any character can
356 be used in a name value, even quotes themselves.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]357
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]358 <li>Unnamed values are represented as an unsigned numeric value with their
359 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]360
361 <li>Constants, which are described in a <a href="#constants">section about
362 constants</a>, below.</li>
363</ol>
364
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]365<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]366don't need to worry about name clashes with reserved words, and the set of
367reserved words may be expanded in the future without penalty. Additionally,
368unnamed identifiers allow a compiler to quickly come up with a temporary
369variable without having to avoid symbol table conflicts.</p>
370
371<p>Reserved words in LLVM are very similar to reserved words in other
372languages. There are keywords for different opcodes
373('<tt><a href="#i_add">add</a></tt>',
374 '<tt><a href="#i_bitcast">bitcast</a></tt>',
375 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
376href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
377and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]378none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]379
380<p>Here is an example of LLVM code to multiply the integer variable
381'<tt>%X</tt>' by 8:</p>
382
383<p>The easy way:</p>
384
385<div class="doc_code">
386<pre>
387%result = <a href="#i_mul">mul</a> i32 %X, 8
388</pre>
389</div>
390
391<p>After strength reduction:</p>
392
393<div class="doc_code">
394<pre>
395%result = <a href="#i_shl">shl</a> i32 %X, i8 3
396</pre>
397</div>
398
399<p>And the hard way:</p>
400
401<div class="doc_code">
402<pre>
403<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
404<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
405%result = <a href="#i_add">add</a> i32 %1, %1
406</pre>
407</div>
408
409<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
410important lexical features of LLVM:</p>
411
412<ol>
413
414 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
415 line.</li>
416
417 <li>Unnamed temporaries are created when the result of a computation is not
418 assigned to a named value.</li>
419
420 <li>Unnamed temporaries are numbered sequentially</li>
421
422</ol>
423
424<p>...and it also shows a convention that we follow in this document. When
425demonstrating instructions, we will follow an instruction with a comment that
426defines the type and name of value produced. Comments are shown in italic
427text.</p>
428
429</div>
430
431<!-- *********************************************************************** -->
432<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
433<!-- *********************************************************************** -->
434
435<!-- ======================================================================= -->
436<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
437</div>
438
439<div class="doc_text">
440
441<p>LLVM programs are composed of "Module"s, each of which is a
442translation unit of the input programs. Each module consists of
443functions, global variables, and symbol table entries. Modules may be
444combined together with the LLVM linker, which merges function (and
445global variable) definitions, resolves forward declarations, and merges
446symbol table entries. Here is an example of the "hello world" module:</p>
447
448<div class="doc_code">
449<pre><i>; Declare the string constant as a global constant...</i>
450<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
451 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
452
453<i>; External declaration of the puts function</i>
454<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
455
456<i>; Definition of main function</i>
457define i32 @main() { <i>; i32()* </i>
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]458 <i>; Convert [13 x i8]* to i8 *...</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]459 %cast210 = <a
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]460 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]461
462 <i>; Call puts function to write out the string to stdout...</i>
463 <a
464 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
465 <a
466 href="#i_ret">ret</a> i32 0<br>}<br>
467</pre>
468</div>
469
470<p>This example is made up of a <a href="#globalvars">global variable</a>
471named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
472function, and a <a href="#functionstructure">function definition</a>
473for "<tt>main</tt>".</p>
474
475<p>In general, a module is made up of a list of global values,
476where both functions and global variables are global values. Global values are
477represented by a pointer to a memory location (in this case, a pointer to an
478array of char, and a pointer to a function), and have one of the following <a
479href="#linkage">linkage types</a>.</p>
480
481</div>
482
483<!-- ======================================================================= -->
484<div class="doc_subsection">
485 <a name="linkage">Linkage Types</a>
486</div>
487
488<div class="doc_text">
489
490<p>
491All Global Variables and Functions have one of the following types of linkage:
492</p>
493
494<dl>
495
Rafael Espindolaa168fc92009-01-15 20:18:42 +0000[diff] [blame]496 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
497
498 <dd>Global values with private linkage are only directly accessible by
499 objects in the current module. In particular, linking code into a module with
500 an private global value may cause the private to be renamed as necessary to
501 avoid collisions. Because the symbol is private to the module, all
502 references can be updated. This doesn't show up in any symbol table in the
503 object file.
504 </dd>
505
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]506 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]507
Duncan Sandsa75223a2009-01-16 09:29:46 +0000[diff] [blame]508 <dd> Similar to private, but the value shows as a local symbol (STB_LOCAL in
Rafael Espindolaa168fc92009-01-15 20:18:42 +0000[diff] [blame]509 the case of ELF) in the object file. This corresponds to the notion of the
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]510 '<tt>static</tt>' keyword in C.
511 </dd>
512
Chris Lattner68433442009-04-13 05:44:34 +0000[diff] [blame]513 <dt><tt><b><a name="available_externally">available_externally</a></b></tt>:
514 </dt>
515
516 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
517 into the object file corresponding to the LLVM module. They exist to
518 allow inlining and other optimizations to take place given knowledge of the
519 definition of the global, which is known to be somewhere outside the module.
520 Globals with <tt>available_externally</tt> linkage are allowed to be discarded
521 at will, and are otherwise the same as <tt>linkonce_odr</tt>. This linkage
522 type is only allowed on definitions, not declarations.</dd>
523
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]524 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
525
526 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
527 the same name when linkage occurs. This is typically used to implement
528 inline functions, templates, or other code which must be generated in each
529 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
530 allowed to be discarded.
531 </dd>
532
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]533 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
534
535 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
536 linkage, except that unreferenced <tt>common</tt> globals may not be
537 discarded. This is used for globals that may be emitted in multiple
538 translation units, but that are not guaranteed to be emitted into every
539 translation unit that uses them. One example of this is tentative
540 definitions in C, such as "<tt>int X;</tt>" at global scope.
541 </dd>
542
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]543 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
544
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]545 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
546 that some targets may choose to emit different assembly sequences for them
547 for target-dependent reasons. This is used for globals that are declared
548 "weak" in C source code.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]549 </dd>
550
551 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
552
553 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
554 pointer to array type. When two global variables with appending linkage are
555 linked together, the two global arrays are appended together. This is the
556 LLVM, typesafe, equivalent of having the system linker append together
557 "sections" with identical names when .o files are linked.
558 </dd>
559
560 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]561
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]562 <dd>The semantics of this linkage follow the ELF object file model: the
563 symbol is weak until linked, if not linked, the symbol becomes null instead
564 of being an undefined reference.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]565 </dd>
566
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]567 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]568 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
Chris Lattner68433442009-04-13 05:44:34 +0000[diff] [blame]569 <dd>Some languages allow differing globals to be merged, such as two
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]570 functions with different semantics. Other languages, such as <tt>C++</tt>,
571 ensure that only equivalent globals are ever merged (the "one definition
Chris Lattner68433442009-04-13 05:44:34 +0000[diff] [blame]572 rule" - "ODR"). Such languages can use the <tt>linkonce_odr</tt>
Duncan Sandsb95df792009-03-11 20:14:15 +0000[diff] [blame]573 and <tt>weak_odr</tt> linkage types to indicate that the global will only
574 be merged with equivalent globals. These linkage types are otherwise the
575 same as their non-<tt>odr</tt> versions.
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]576 </dd>
577
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]578 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
579
580 <dd>If none of the above identifiers are used, the global is externally
581 visible, meaning that it participates in linkage and can be used to resolve
582 external symbol references.
583 </dd>
584</dl>
585
586 <p>
587 The next two types of linkage are targeted for Microsoft Windows platform
588 only. They are designed to support importing (exporting) symbols from (to)
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]589 DLLs (Dynamic Link Libraries).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]590 </p>
591
592 <dl>
593 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
594
595 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
596 or variable via a global pointer to a pointer that is set up by the DLL
597 exporting the symbol. On Microsoft Windows targets, the pointer name is
Dan Gohman5ec99832009-01-12 21:35:55 +0000[diff] [blame]598 formed by combining <code>__imp_</code> and the function or variable name.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]599 </dd>
600
601 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
602
603 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
604 pointer to a pointer in a DLL, so that it can be referenced with the
605 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
Dan Gohman5ec99832009-01-12 21:35:55 +0000[diff] [blame]606 name is formed by combining <code>__imp_</code> and the function or variable
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]607 name.
608 </dd>
609
610</dl>
611
Dan Gohman4dfac702008-11-24 17:18:39 +0000[diff] [blame]612<p>For example, since the "<tt>.LC0</tt>"
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]613variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
614variable and was linked with this one, one of the two would be renamed,
615preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
616external (i.e., lacking any linkage declarations), they are accessible
617outside of the current module.</p>
618<p>It is illegal for a function <i>declaration</i>
Duncan Sands565f65d2009-03-11 08:08:06 +0000[diff] [blame]619to have any linkage type other than "externally visible", <tt>dllimport</tt>
620or <tt>extern_weak</tt>.</p>
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]621<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
622or <tt>weak_odr</tt> linkages.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]623</div>
624
625<!-- ======================================================================= -->
626<div class="doc_subsection">
627 <a name="callingconv">Calling Conventions</a>
628</div>
629
630<div class="doc_text">
631
632<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
633and <a href="#i_invoke">invokes</a> can all have an optional calling convention
634specified for the call. The calling convention of any pair of dynamic
635caller/callee must match, or the behavior of the program is undefined. The
636following calling conventions are supported by LLVM, and more may be added in
637the future:</p>
638
639<dl>
640 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
641
642 <dd>This calling convention (the default if no other calling convention is
643 specified) matches the target C calling conventions. This calling convention
644 supports varargs function calls and tolerates some mismatch in the declared
645 prototype and implemented declaration of the function (as does normal C).
646 </dd>
647
648 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
649
650 <dd>This calling convention attempts to make calls as fast as possible
651 (e.g. by passing things in registers). This calling convention allows the
652 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]653 without having to conform to an externally specified ABI (Application Binary
654 Interface). Implementations of this convention should allow arbitrary
Arnold Schwaighofer07444922008-05-14 09:17:12 +0000[diff] [blame]655 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
656 supported. This calling convention does not support varargs and requires the
657 prototype of all callees to exactly match the prototype of the function
658 definition.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]659 </dd>
660
661 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
662
663 <dd>This calling convention attempts to make code in the caller as efficient
664 as possible under the assumption that the call is not commonly executed. As
665 such, these calls often preserve all registers so that the call does not break
666 any live ranges in the caller side. This calling convention does not support
667 varargs and requires the prototype of all callees to exactly match the
668 prototype of the function definition.
669 </dd>
670
671 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
672
673 <dd>Any calling convention may be specified by number, allowing
674 target-specific calling conventions to be used. Target specific calling
675 conventions start at 64.
676 </dd>
677</dl>
678
679<p>More calling conventions can be added/defined on an as-needed basis, to
680support pascal conventions or any other well-known target-independent
681convention.</p>
682
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
692<p>
693All Global Variables and Functions have one of the following visibility styles:
694</p>
695
696<dl>
697 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
698
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]699 <dd>On targets that use the ELF object file format, default visibility means
700 that the declaration is visible to other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]701 modules and, in shared libraries, means that the declared entity may be
702 overridden. On Darwin, default visibility means that the declaration is
703 visible to other modules. Default visibility corresponds to "external
704 linkage" in the language.
705 </dd>
706
707 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
708
709 <dd>Two declarations of an object with hidden visibility refer to the same
710 object if they are in the same shared object. Usually, hidden visibility
711 indicates that the symbol will not be placed into the dynamic symbol table,
712 so no other module (executable or shared library) can reference it
713 directly.
714 </dd>
715
716 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
717
718 <dd>On ELF, protected visibility indicates that the symbol will be placed in
719 the dynamic symbol table, but that references within the defining module will
720 bind to the local symbol. That is, the symbol cannot be overridden by another
721 module.
722 </dd>
723</dl>
724
725</div>
726
727<!-- ======================================================================= -->
728<div class="doc_subsection">
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]729 <a name="namedtypes">Named Types</a>
730</div>
731
732<div class="doc_text">
733
734<p>LLVM IR allows you to specify name aliases for certain types. This can make
735it easier to read the IR and make the IR more condensed (particularly when
736recursive types are involved). An example of a name specification is:
737</p>
738
739<div class="doc_code">
740<pre>
741%mytype = type { %mytype*, i32 }
742</pre>
743</div>
744
745<p>You may give a name to any <a href="#typesystem">type</a> except "<a
746href="t_void">void</a>". Type name aliases may be used anywhere a type is
747expected with the syntax "%mytype".</p>
748
749<p>Note that type names are aliases for the structural type that they indicate,
750and that you can therefore specify multiple names for the same type. This often
751leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
752structural typing, the name is not part of the type. When printing out LLVM IR,
753the printer will pick <em>one name</em> to render all types of a particular
754shape. This means that if you have code where two different source types end up
755having the same LLVM type, that the dumper will sometimes print the "wrong" or
756unexpected type. This is an important design point and isn't going to
757change.</p>
758
759</div>
760
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]761<!-- ======================================================================= -->
762<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]763 <a name="globalvars">Global Variables</a>
764</div>
765
766<div class="doc_text">
767
768<p>Global variables define regions of memory allocated at compilation time
769instead of run-time. Global variables may optionally be initialized, may have
770an explicit section to be placed in, and may have an optional explicit alignment
771specified. A variable may be defined as "thread_local", which means that it
772will not be shared by threads (each thread will have a separated copy of the
773variable). A variable may be defined as a global "constant," which indicates
774that the contents of the variable will <b>never</b> be modified (enabling better
775optimization, allowing the global data to be placed in the read-only section of
776an executable, etc). Note that variables that need runtime initialization
777cannot be marked "constant" as there is a store to the variable.</p>
778
779<p>
780LLVM explicitly allows <em>declarations</em> of global variables to be marked
781constant, even if the final definition of the global is not. This capability
782can be used to enable slightly better optimization of the program, but requires
783the language definition to guarantee that optimizations based on the
784'constantness' are valid for the translation units that do not include the
785definition.
786</p>
787
788<p>As SSA values, global variables define pointer values that are in
789scope (i.e. they dominate) all basic blocks in the program. Global
790variables always define a pointer to their "content" type because they
791describe a region of memory, and all memory objects in LLVM are
792accessed through pointers.</p>
793
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]794<p>A global variable may be declared to reside in a target-specifc numbered
795address space. For targets that support them, address spaces may affect how
796optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]797the variable. The default address space is zero. The address space qualifier
798must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]799
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]800<p>LLVM allows an explicit section to be specified for globals. If the target
801supports it, it will emit globals to the section specified.</p>
802
803<p>An explicit alignment may be specified for a global. If not present, or if
804the alignment is set to zero, the alignment of the global is set by the target
805to whatever it feels convenient. If an explicit alignment is specified, the
806global is forced to have at least that much alignment. All alignments must be
807a power of 2.</p>
808
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]809<p>For example, the following defines a global in a numbered address space with
810an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]811
812<div class="doc_code">
813<pre>
Dan Gohman21ef02c2009-01-11 00:40:00 +0000[diff] [blame]814@G = addrspace(5) constant float 1.0, section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]815</pre>
816</div>
817
818</div>
819
820
821<!-- ======================================================================= -->
822<div class="doc_subsection">
823 <a name="functionstructure">Functions</a>
824</div>
825
826<div class="doc_text">
827
828<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
829an optional <a href="#linkage">linkage type</a>, an optional
830<a href="#visibility">visibility style</a>, an optional
831<a href="#callingconv">calling convention</a>, a return type, an optional
832<a href="#paramattrs">parameter attribute</a> for the return type, a function
833name, a (possibly empty) argument list (each with optional
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]834<a href="#paramattrs">parameter attributes</a>), optional
835<a href="#fnattrs">function attributes</a>, an optional section,
836an optional alignment, an optional <a href="#gc">garbage collector name</a>,
Chris Lattneree202542008-10-04 18:10:21 +0000[diff] [blame]837an opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]838
839LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
840optional <a href="#linkage">linkage type</a>, an optional
841<a href="#visibility">visibility style</a>, an optional
842<a href="#callingconv">calling convention</a>, a return type, an optional
843<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]844name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]845<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]846
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]847<p>A function definition contains a list of basic blocks, forming the CFG
848(Control Flow Graph) for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]849the function. Each basic block may optionally start with a label (giving the
850basic block a symbol table entry), contains a list of instructions, and ends
851with a <a href="#terminators">terminator</a> instruction (such as a branch or
852function return).</p>
853
854<p>The first basic block in a function is special in two ways: it is immediately
855executed on entrance to the function, and it is not allowed to have predecessor
856basic blocks (i.e. there can not be any branches to the entry block of a
857function). Because the block can have no predecessors, it also cannot have any
858<a href="#i_phi">PHI nodes</a>.</p>
859
860<p>LLVM allows an explicit section to be specified for functions. If the target
861supports it, it will emit functions to the section specified.</p>
862
863<p>An explicit alignment may be specified for a function. If not present, or if
864the alignment is set to zero, the alignment of the function is set by the target
865to whatever it feels convenient. If an explicit alignment is specified, the
866function is forced to have at least that much alignment. All alignments must be
867a power of 2.</p>
868
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]869 <h5>Syntax:</h5>
870
871<div class="doc_code">
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000[diff] [blame]872<tt>
873define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
874 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
875 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
876 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
877 [<a href="#gc">gc</a>] { ... }
878</tt>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]879</div>
880
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]881</div>
882
883
884<!-- ======================================================================= -->
885<div class="doc_subsection">
886 <a name="aliasstructure">Aliases</a>
887</div>
888<div class="doc_text">
889 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]890 function, global variable, another alias or bitcast of global value). Aliases
891 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]892 optional <a href="#visibility">visibility style</a>.</p>
893
894 <h5>Syntax:</h5>
895
896<div class="doc_code">
897<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000[diff] [blame]898@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]899</pre>
900</div>
901
902</div>
903
904
905
906<!-- ======================================================================= -->
907<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
908<div class="doc_text">
909 <p>The return type and each parameter of a function type may have a set of
910 <i>parameter attributes</i> associated with them. Parameter attributes are
911 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]912 a function. Parameter attributes are considered to be part of the function,
913 not of the function type, so functions with different parameter attributes
914 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]915
916 <p>Parameter attributes are simple keywords that follow the type specified. If
917 multiple parameter attributes are needed, they are space separated. For
918 example:</p>
919
920<div class="doc_code">
921<pre>
Nick Lewycky3022a742009-02-15 23:06:14 +0000[diff] [blame]922declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]923declare i32 @atoi(i8 zeroext)
924declare signext i8 @returns_signed_char()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]925</pre>
926</div>
927
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]928 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
929 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]930
931 <p>Currently, only the following parameter attributes are defined:</p>
932 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]933 <dt><tt>zeroext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]934 <dd>This indicates to the code generator that the parameter or return value
935 should be zero-extended to a 32-bit value by the caller (for a parameter)
936 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]937
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]938 <dt><tt>signext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]939 <dd>This indicates to the code generator that the parameter or return value
940 should be sign-extended to a 32-bit value by the caller (for a parameter)
941 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]942
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]943 <dt><tt>inreg</tt></dt>
Dale Johannesenc08a0e22008-09-25 20:47:45 +0000[diff] [blame]944 <dd>This indicates that this parameter or return value should be treated
945 in a special target-dependent fashion during while emitting code for a
946 function call or return (usually, by putting it in a register as opposed
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]947 to memory, though some targets use it to distinguish between two different
948 kinds of registers). Use of this attribute is target-specific.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]949
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]950 <dt><tt><a name="byval">byval</a></tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]951 <dd>This indicates that the pointer parameter should really be passed by
952 value to the function. The attribute implies that a hidden copy of the
953 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]954 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]955 pointer arguments. It is generally used to pass structs and arrays by
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]956 value, but is also valid on pointers to scalars. The copy is considered to
957 belong to the caller not the callee (for example,
958 <tt><a href="#readonly">readonly</a></tt> functions should not write to
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]959 <tt>byval</tt> parameters). This is not a valid attribute for return
Chris Lattnere39c5bf2009-02-05 05:42:28 +0000[diff] [blame]960 values. The byval attribute also supports specifying an alignment with the
961 align attribute. This has a target-specific effect on the code generator
962 that usually indicates a desired alignment for the synthesized stack
963 slot.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]964
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]965 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]966 <dd>This indicates that the pointer parameter specifies the address of a
967 structure that is the return value of the function in the source program.
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]968 This pointer must be guaranteed by the caller to be valid: loads and stores
969 to the structure may be assumed by the callee to not to trap. This may only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]970 be applied to the first parameter. This is not a valid attribute for
971 return values. </dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]972
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]973 <dt><tt>noalias</tt></dt>
Nick Lewyckyff384472008-11-24 03:41:24 +0000[diff] [blame]974 <dd>This indicates that the pointer does not alias any global or any other
975 parameter. The caller is responsible for ensuring that this is the
Nick Lewyckya604a942008-11-24 05:00:44 +0000[diff] [blame]976 case. On a function return value, <tt>noalias</tt> additionally indicates
977 that the pointer does not alias any other pointers visible to the
Nick Lewycky40fb6f22008-12-19 06:39:12 +0000[diff] [blame]978 caller. For further details, please see the discussion of the NoAlias
979 response in
980 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
981 analysis</a>.</dd>
982
983 <dt><tt>nocapture</tt></dt>
984 <dd>This indicates that the callee does not make any copies of the pointer
985 that outlive the callee itself. This is not a valid attribute for return
986 values.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]987
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]988 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]989 <dd>This indicates that the pointer parameter can be excised using the
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]990 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
991 attribute for return values.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]992 </dl>
993
994</div>
995
996<!-- ======================================================================= -->
997<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]998 <a name="gc">Garbage Collector Names</a>
999</div>
1000
1001<div class="doc_text">
1002<p>Each function may specify a garbage collector name, which is simply a
1003string.</p>
1004
1005<div class="doc_code"><pre
1006>define void @f() gc "name" { ...</pre></div>
1007
1008<p>The compiler declares the supported values of <i>name</i>. Specifying a
1009collector which will cause the compiler to alter its output in order to support
1010the named garbage collection algorithm.</p>
1011</div>
1012
1013<!-- ======================================================================= -->
1014<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1015 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1016</div>
1017
1018<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1019
1020<p>Function attributes are set to communicate additional information about
1021 a function. Function attributes are considered to be part of the function,
1022 not of the function type, so functions with different parameter attributes
1023 can have the same function type.</p>
1024
1025 <p>Function attributes are simple keywords that follow the type specified. If
1026 multiple attributes are needed, they are space separated. For
1027 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1028
1029<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1030<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1031define void @f() noinline { ... }
1032define void @f() alwaysinline { ... }
1033define void @f() alwaysinline optsize { ... }
1034define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1035</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1036</div>
1037
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1038<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1039<dt><tt>alwaysinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1040<dd>This attribute indicates that the inliner should attempt to inline this
1041function into callers whenever possible, ignoring any active inlining size
1042threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1043
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1044<dt><tt>noinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1045<dd>This attribute indicates that the inliner should never inline this function
Chris Lattner377f3ae2008-10-05 17:14:59 +0000[diff] [blame]1046in any situation. This attribute may not be used together with the
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1047<tt>alwaysinline</tt> attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1048
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1049<dt><tt>optsize</tt></dt>
Devang Patelc29099b2008-09-29 18:34:44 +0000[diff] [blame]1050<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1051make choices that keep the code size of this function low, and otherwise do
1052optimizations specifically to reduce code size.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1053
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1054<dt><tt>noreturn</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1055<dd>This function attribute indicates that the function never returns normally.
1056This produces undefined behavior at runtime if the function ever does
1057dynamically return.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1058
1059<dt><tt>nounwind</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1060<dd>This function attribute indicates that the function never returns with an
1061unwind or exceptional control flow. If the function does unwind, its runtime
1062behavior is undefined.</dd>
1063
1064<dt><tt>readnone</tt></dt>
Duncan Sands2f500832009-05-06 06:49:50 +0000[diff] [blame]1065<dd>This attribute indicates that the function computes its result (or decides to
1066unwind an exception) based strictly on its arguments, without dereferencing any
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]1067pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1068registers, etc) visible to caller functions. It does not write through any
1069pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
Duncan Sands2f500832009-05-06 06:49:50 +0000[diff] [blame]1070never changes any state visible to callers. This means that it cannot unwind
1071exceptions by calling the <tt>C++</tt> exception throwing methods, but could
1072use the <tt>unwind</tt> instruction.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1073
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]1074<dt><tt><a name="readonly">readonly</a></tt></dt>
1075<dd>This attribute indicates that the function does not write through any
1076pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1077or otherwise modify any state (e.g. memory, control registers, etc) visible to
1078caller functions. It may dereference pointer arguments and read state that may
Duncan Sands2f500832009-05-06 06:49:50 +0000[diff] [blame]1079be set in the caller. A readonly function always returns the same value (or
1080unwinds an exception identically) when called with the same set of arguments
1081and global state. It cannot unwind an exception by calling the <tt>C++</tt>
1082exception throwing methods, but may use the <tt>unwind</tt> instruction.</dd>
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1083
1084<dt><tt><a name="ssp">ssp</a></tt></dt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1085<dd>This attribute indicates that the function should emit a stack smashing
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1086protector. It is in the form of a "canary"&mdash;a random value placed on the
1087stack before the local variables that's checked upon return from the function to
1088see if it has been overwritten. A heuristic is used to determine if a function
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1089needs stack protectors or not.
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1090
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +0000[diff] [blame]1091<p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1092that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1093have an <tt>ssp</tt> attribute.</p></dd>
1094
1095<dt><tt>sspreq</tt></dt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1096<dd>This attribute indicates that the function should <em>always</em> emit a
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1097stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1098function attribute.
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +0000[diff] [blame]1099
1100<p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1101function that doesn't have an <tt>sspreq</tt> attribute or which has
1102an <tt>ssp</tt> attribute, then the resulting function will have
1103an <tt>sspreq</tt> attribute.</p></dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1104</dl>
1105
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1106</div>
1107
1108<!-- ======================================================================= -->
1109<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1110 <a name="moduleasm">Module-Level Inline Assembly</a>
1111</div>
1112
1113<div class="doc_text">
1114<p>
1115Modules may contain "module-level inline asm" blocks, which corresponds to the
1116GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1117LLVM and treated as a single unit, but may be separated in the .ll file if
1118desired. The syntax is very simple:
1119</p>
1120
1121<div class="doc_code">
1122<pre>
1123module asm "inline asm code goes here"
1124module asm "more can go here"
1125</pre>
1126</div>
1127
1128<p>The strings can contain any character by escaping non-printable characters.
1129 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1130 for the number.
1131</p>
1132
1133<p>
1134 The inline asm code is simply printed to the machine code .s file when
1135 assembly code is generated.
1136</p>
1137</div>
1138
1139<!-- ======================================================================= -->
1140<div class="doc_subsection">
1141 <a name="datalayout">Data Layout</a>
1142</div>
1143
1144<div class="doc_text">
1145<p>A module may specify a target specific data layout string that specifies how
1146data is to be laid out in memory. The syntax for the data layout is simply:</p>
1147<pre> target datalayout = "<i>layout specification</i>"</pre>
1148<p>The <i>layout specification</i> consists of a list of specifications
1149separated by the minus sign character ('-'). Each specification starts with a
1150letter and may include other information after the letter to define some
1151aspect of the data layout. The specifications accepted are as follows: </p>
1152<dl>
1153 <dt><tt>E</tt></dt>
1154 <dd>Specifies that the target lays out data in big-endian form. That is, the
1155 bits with the most significance have the lowest address location.</dd>
1156 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1157 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1158 the bits with the least significance have the lowest address location.</dd>
1159 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1160 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1161 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1162 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1163 too.</dd>
1164 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1165 <dd>This specifies the alignment for an integer type of a given bit
1166 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1167 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1168 <dd>This specifies the alignment for a vector type of a given bit
1169 <i>size</i>.</dd>
1170 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1171 <dd>This specifies the alignment for a floating point type of a given bit
1172 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1173 (double).</dd>
1174 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1175 <dd>This specifies the alignment for an aggregate type of a given bit
1176 <i>size</i>.</dd>
1177</dl>
1178<p>When constructing the data layout for a given target, LLVM starts with a
1179default set of specifications which are then (possibly) overriden by the
1180specifications in the <tt>datalayout</tt> keyword. The default specifications
1181are given in this list:</p>
1182<ul>
1183 <li><tt>E</tt> - big endian</li>
1184 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1185 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1186 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1187 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1188 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1189 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1190 alignment of 64-bits</li>
1191 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1192 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1193 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1194 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1195 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1196</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1197<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1198following rules:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1199<ol>
1200 <li>If the type sought is an exact match for one of the specifications, that
1201 specification is used.</li>
1202 <li>If no match is found, and the type sought is an integer type, then the
1203 smallest integer type that is larger than the bitwidth of the sought type is
1204 used. If none of the specifications are larger than the bitwidth then the the
1205 largest integer type is used. For example, given the default specifications
1206 above, the i7 type will use the alignment of i8 (next largest) while both
1207 i65 and i256 will use the alignment of i64 (largest specified).</li>
1208 <li>If no match is found, and the type sought is a vector type, then the
1209 largest vector type that is smaller than the sought vector type will be used
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1210 as a fall back. This happens because &lt;128 x double&gt; can be implemented
1211 in terms of 64 &lt;2 x double&gt;, for example.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1212</ol>
1213</div>
1214
1215<!-- *********************************************************************** -->
1216<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1217<!-- *********************************************************************** -->
1218
1219<div class="doc_text">
1220
1221<p>The LLVM type system is one of the most important features of the
1222intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1223optimizations to be performed on the intermediate representation directly,
1224without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1225extra analyses on the side before the transformation. A strong type
1226system makes it easier to read the generated code and enables novel
1227analyses and transformations that are not feasible to perform on normal
1228three address code representations.</p>
1229
1230</div>
1231
1232<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1233<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1234Classifications</a> </div>
1235<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1236<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1237classifications:</p>
1238
1239<table border="1" cellspacing="0" cellpadding="4">
1240 <tbody>
1241 <tr><th>Classification</th><th>Types</th></tr>
1242 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1243 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1244 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1245 </tr>
1246 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1247 <td><a href="#t_floating">floating point</a></td>
1248 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1249 </tr>
1250 <tr>
1251 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1252 <td><a href="#t_integer">integer</a>,
1253 <a href="#t_floating">floating point</a>,
1254 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1255 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1256 <a href="#t_struct">structure</a>,
1257 <a href="#t_array">array</a>,
Nick Lewycky29aaef82009-05-30 05:06:04 +0000[diff] [blame]1258 <a href="#t_label">label</a>,
1259 <a href="#t_metadata">metadata</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1260 </td>
1261 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1262 <tr>
1263 <td><a href="#t_primitive">primitive</a></td>
1264 <td><a href="#t_label">label</a>,
1265 <a href="#t_void">void</a>,
Nick Lewycky29aaef82009-05-30 05:06:04 +0000[diff] [blame]1266 <a href="#t_floating">floating point</a>,
1267 <a href="#t_metadata">metadata</a>.</td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1268 </tr>
1269 <tr>
1270 <td><a href="#t_derived">derived</a></td>
1271 <td><a href="#t_integer">integer</a>,
1272 <a href="#t_array">array</a>,
1273 <a href="#t_function">function</a>,
1274 <a href="#t_pointer">pointer</a>,
1275 <a href="#t_struct">structure</a>,
1276 <a href="#t_pstruct">packed structure</a>,
1277 <a href="#t_vector">vector</a>,
1278 <a href="#t_opaque">opaque</a>.
Dan Gohman032ba852008-10-14 16:32:04 +0000[diff] [blame]1279 </td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1280 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1281 </tbody>
1282</table>
1283
1284<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1285most important. Values of these types are the only ones which can be
1286produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1287instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1288</div>
1289
1290<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1291<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1292
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1293<div class="doc_text">
1294<p>The primitive types are the fundamental building blocks of the LLVM
1295system.</p>
1296
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1297</div>
1298
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1299<!-- _______________________________________________________________________ -->
1300<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1301
1302<div class="doc_text">
1303 <table>
1304 <tbody>
1305 <tr><th>Type</th><th>Description</th></tr>
1306 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1307 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1308 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1309 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1310 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1311 </tbody>
1312 </table>
1313</div>
1314
1315<!-- _______________________________________________________________________ -->
1316<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1317
1318<div class="doc_text">
1319<h5>Overview:</h5>
1320<p>The void type does not represent any value and has no size.</p>
1321
1322<h5>Syntax:</h5>
1323
1324<pre>
1325 void
1326</pre>
1327</div>
1328
1329<!-- _______________________________________________________________________ -->
1330<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1331
1332<div class="doc_text">
1333<h5>Overview:</h5>
1334<p>The label type represents code labels.</p>
1335
1336<h5>Syntax:</h5>
1337
1338<pre>
1339 label
1340</pre>
1341</div>
1342
Nick Lewycky29aaef82009-05-30 05:06:04 +0000[diff] [blame]1343<!-- _______________________________________________________________________ -->
1344<div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1345
1346<div class="doc_text">
1347<h5>Overview:</h5>
1348<p>The metadata type represents embedded metadata. The only derived type that
1349may contain metadata is <tt>metadata*</tt> or a function type that returns or
1350takes metadata typed parameters, but not pointer to metadata types.</p>
1351
1352<h5>Syntax:</h5>
1353
1354<pre>
1355 metadata
1356</pre>
1357</div>
1358
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1359
1360<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1361<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1362
1363<div class="doc_text">
1364
1365<p>The real power in LLVM comes from the derived types in the system.
1366This is what allows a programmer to represent arrays, functions,
1367pointers, and other useful types. Note that these derived types may be
1368recursive: For example, it is possible to have a two dimensional array.</p>
1369
1370</div>
1371
1372<!-- _______________________________________________________________________ -->
1373<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1374
1375<div class="doc_text">
1376
1377<h5>Overview:</h5>
1378<p>The integer type is a very simple derived type that simply specifies an
1379arbitrary bit width for the integer type desired. Any bit width from 1 bit to
13802^23-1 (about 8 million) can be specified.</p>
1381
1382<h5>Syntax:</h5>
1383
1384<pre>
1385 iN
1386</pre>
1387
1388<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1389value.</p>
1390
1391<h5>Examples:</h5>
1392<table class="layout">
Nick Lewycky39382d62009-05-24 02:46:06 +0000[diff] [blame]1393 <tr class="layout">
1394 <td class="left"><tt>i1</tt></td>
1395 <td class="left">a single-bit integer.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1396 </tr>
Nick Lewycky39382d62009-05-24 02:46:06 +0000[diff] [blame]1397 <tr class="layout">
1398 <td class="left"><tt>i32</tt></td>
1399 <td class="left">a 32-bit integer.</td>
1400 </tr>
1401 <tr class="layout">
1402 <td class="left"><tt>i1942652</tt></td>
1403 <td class="left">a really big integer of over 1 million bits.</td>
1404 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1405</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1406
1407<p>Note that the code generator does not yet support large integer types
1408to be used as function return types. The specific limit on how large a
1409return type the code generator can currently handle is target-dependent;
1410currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1411targets.</p>
1412
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1413</div>
1414
1415<!-- _______________________________________________________________________ -->
1416<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1417
1418<div class="doc_text">
1419
1420<h5>Overview:</h5>
1421
1422<p>The array type is a very simple derived type that arranges elements
1423sequentially in memory. The array type requires a size (number of
1424elements) and an underlying data type.</p>
1425
1426<h5>Syntax:</h5>
1427
1428<pre>
1429 [&lt;# elements&gt; x &lt;elementtype&gt;]
1430</pre>
1431
1432<p>The number of elements is a constant integer value; elementtype may
1433be any type with a size.</p>
1434
1435<h5>Examples:</h5>
1436<table class="layout">
1437 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1438 <td class="left"><tt>[40 x i32]</tt></td>
1439 <td class="left">Array of 40 32-bit integer values.</td>
1440 </tr>
1441 <tr class="layout">
1442 <td class="left"><tt>[41 x i32]</tt></td>
1443 <td class="left">Array of 41 32-bit integer values.</td>
1444 </tr>
1445 <tr class="layout">
1446 <td class="left"><tt>[4 x i8]</tt></td>
1447 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1448 </tr>
1449</table>
1450<p>Here are some examples of multidimensional arrays:</p>
1451<table class="layout">
1452 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1453 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1454 <td class="left">3x4 array of 32-bit integer values.</td>
1455 </tr>
1456 <tr class="layout">
1457 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1458 <td class="left">12x10 array of single precision floating point values.</td>
1459 </tr>
1460 <tr class="layout">
1461 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1462 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1463 </tr>
1464</table>
1465
1466<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1467length array. Normally, accesses past the end of an array are undefined in
1468LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1469As a special case, however, zero length arrays are recognized to be variable
1470length. This allows implementation of 'pascal style arrays' with the LLVM
1471type "{ i32, [0 x float]}", for example.</p>
1472
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1473<p>Note that the code generator does not yet support large aggregate types
1474to be used as function return types. The specific limit on how large an
1475aggregate return type the code generator can currently handle is
1476target-dependent, and also dependent on the aggregate element types.</p>
1477
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1478</div>
1479
1480<!-- _______________________________________________________________________ -->
1481<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1482<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1483
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1484<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1485
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1486<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1487consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1488return type of a function type is a scalar type, a void type, or a struct type.
Devang Pateld5404c02008-03-24 20:52:42 +0000[diff] [blame]1489If the return type is a struct type then all struct elements must be of first
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1490class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1491
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1492<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1493
1494<pre>
1495 &lt;returntype list&gt; (&lt;parameter list&gt;)
1496</pre>
1497
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1498<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1499specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1500which indicates that the function takes a variable number of arguments.
1501Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1502 href="#int_varargs">variable argument handling intrinsic</a> functions.
1503'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1504<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1505
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1506<h5>Examples:</h5>
1507<table class="layout">
1508 <tr class="layout">
1509 <td class="left"><tt>i32 (i32)</tt></td>
1510 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1511 </td>
1512 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1513 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1514 </tt></td>
1515 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1516 an <tt>i16</tt> that should be sign extended and a
1517 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1518 <tt>float</tt>.
1519 </td>
1520 </tr><tr class="layout">
1521 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1522 <td class="left">A vararg function that takes at least one
1523 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1524 which returns an integer. This is the signature for <tt>printf</tt> in
1525 LLVM.
1526 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1527 </tr><tr class="layout">
1528 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Misha Brukmanafc88b02008-11-27 06:41:20 +0000[diff] [blame]1529 <td class="left">A function taking an <tt>i32</tt>, returning two
1530 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1531 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1532 </tr>
1533</table>
1534
1535</div>
1536<!-- _______________________________________________________________________ -->
1537<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1538<div class="doc_text">
1539<h5>Overview:</h5>
1540<p>The structure type is used to represent a collection of data members
1541together in memory. The packing of the field types is defined to match
1542the ABI of the underlying processor. The elements of a structure may
1543be any type that has a size.</p>
1544<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1545and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1546field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1547instruction.</p>
1548<h5>Syntax:</h5>
1549<pre> { &lt;type list&gt; }<br></pre>
1550<h5>Examples:</h5>
1551<table class="layout">
1552 <tr class="layout">
1553 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1554 <td class="left">A triple of three <tt>i32</tt> values</td>
1555 </tr><tr class="layout">
1556 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1557 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1558 second element is a <a href="#t_pointer">pointer</a> to a
1559 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1560 an <tt>i32</tt>.</td>
1561 </tr>
1562</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1563
1564<p>Note that the code generator does not yet support large aggregate types
1565to be used as function return types. The specific limit on how large an
1566aggregate return type the code generator can currently handle is
1567target-dependent, and also dependent on the aggregate element types.</p>
1568
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1569</div>
1570
1571<!-- _______________________________________________________________________ -->
1572<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1573</div>
1574<div class="doc_text">
1575<h5>Overview:</h5>
1576<p>The packed structure type is used to represent a collection of data members
1577together in memory. There is no padding between fields. Further, the alignment
1578of a packed structure is 1 byte. The elements of a packed structure may
1579be any type that has a size.</p>
1580<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1581and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1582field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1583instruction.</p>
1584<h5>Syntax:</h5>
1585<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1586<h5>Examples:</h5>
1587<table class="layout">
1588 <tr class="layout">
1589 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1590 <td class="left">A triple of three <tt>i32</tt> values</td>
1591 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1592 <td class="left">
1593<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1594 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1595 second element is a <a href="#t_pointer">pointer</a> to a
1596 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1597 an <tt>i32</tt>.</td>
1598 </tr>
1599</table>
1600</div>
1601
1602<!-- _______________________________________________________________________ -->
1603<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1604<div class="doc_text">
1605<h5>Overview:</h5>
1606<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1607reference to another object, which must live in memory. Pointer types may have
1608an optional address space attribute defining the target-specific numbered
1609address space where the pointed-to object resides. The default address space is
1610zero.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +0000[diff] [blame]1611
1612<p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does
Chris Lattner0d879e02009-02-08 22:21:28 +0000[diff] [blame]1613it permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +0000[diff] [blame]1614
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1615<h5>Syntax:</h5>
1616<pre> &lt;type&gt; *<br></pre>
1617<h5>Examples:</h5>
1618<table class="layout">
1619 <tr class="layout">
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]1620 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1621 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1622 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1623 </tr>
1624 <tr class="layout">
1625 <td class="left"><tt>i32 (i32 *) *</tt></td>
1626 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1627 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1628 <tt>i32</tt>.</td>
1629 </tr>
1630 <tr class="layout">
1631 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1632 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1633 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1634 </tr>
1635</table>
1636</div>
1637
1638<!-- _______________________________________________________________________ -->
1639<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1640<div class="doc_text">
1641
1642<h5>Overview:</h5>
1643
1644<p>A vector type is a simple derived type that represents a vector
1645of elements. Vector types are used when multiple primitive data
1646are operated in parallel using a single instruction (SIMD).
1647A vector type requires a size (number of
1648elements) and an underlying primitive data type. Vectors must have a power
1649of two length (1, 2, 4, 8, 16 ...). Vector types are
1650considered <a href="#t_firstclass">first class</a>.</p>
1651
1652<h5>Syntax:</h5>
1653
1654<pre>
1655 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1656</pre>
1657
1658<p>The number of elements is a constant integer value; elementtype may
1659be any integer or floating point type.</p>
1660
1661<h5>Examples:</h5>
1662
1663<table class="layout">
1664 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1665 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1666 <td class="left">Vector of 4 32-bit integer values.</td>
1667 </tr>
1668 <tr class="layout">
1669 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1670 <td class="left">Vector of 8 32-bit floating-point values.</td>
1671 </tr>
1672 <tr class="layout">
1673 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1674 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1675 </tr>
1676</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1677
1678<p>Note that the code generator does not yet support large vector types
1679to be used as function return types. The specific limit on how large a
1680vector return type codegen can currently handle is target-dependent;
1681currently it's often a few times longer than a hardware vector register.</p>
1682
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1683</div>
1684
1685<!-- _______________________________________________________________________ -->
1686<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1687<div class="doc_text">
1688
1689<h5>Overview:</h5>
1690
1691<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1692corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1693In LLVM, opaque types can eventually be resolved to any type (not just a
1694structure type).</p>
1695
1696<h5>Syntax:</h5>
1697
1698<pre>
1699 opaque
1700</pre>
1701
1702<h5>Examples:</h5>
1703
1704<table class="layout">
1705 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1706 <td class="left"><tt>opaque</tt></td>
1707 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1708 </tr>
1709</table>
1710</div>
1711
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]1712<!-- ======================================================================= -->
1713<div class="doc_subsection">
1714 <a name="t_uprefs">Type Up-references</a>
1715</div>
1716
1717<div class="doc_text">
1718<h5>Overview:</h5>
1719<p>
1720An "up reference" allows you to refer to a lexically enclosing type without
1721requiring it to have a name. For instance, a structure declaration may contain a
1722pointer to any of the types it is lexically a member of. Example of up
1723references (with their equivalent as named type declarations) include:</p>
1724
1725<pre>
Chris Lattner5ad632d2009-02-09 10:00:56 +0000[diff] [blame]1726 { \2 * } %x = type { %x* }
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]1727 { \2 }* %y = type { %y }*
1728 \1* %z = type %z*
1729</pre>
1730
1731<p>
1732An up reference is needed by the asmprinter for printing out cyclic types when
1733there is no declared name for a type in the cycle. Because the asmprinter does
1734not want to print out an infinite type string, it needs a syntax to handle
1735recursive types that have no names (all names are optional in llvm IR).
1736</p>
1737
1738<h5>Syntax:</h5>
1739<pre>
1740 \&lt;level&gt;
1741</pre>
1742
1743<p>
1744The level is the count of the lexical type that is being referred to.
1745</p>
1746
1747<h5>Examples:</h5>
1748
1749<table class="layout">
1750 <tr class="layout">
1751 <td class="left"><tt>\1*</tt></td>
1752 <td class="left">Self-referential pointer.</td>
1753 </tr>
1754 <tr class="layout">
1755 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1756 <td class="left">Recursive structure where the upref refers to the out-most
1757 structure.</td>
1758 </tr>
1759</table>
1760</div>
1761
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1762
1763<!-- *********************************************************************** -->
1764<div class="doc_section"> <a name="constants">Constants</a> </div>
1765<!-- *********************************************************************** -->
1766
1767<div class="doc_text">
1768
1769<p>LLVM has several different basic types of constants. This section describes
1770them all and their syntax.</p>
1771
1772</div>
1773
1774<!-- ======================================================================= -->
1775<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1776
1777<div class="doc_text">
1778
1779<dl>
1780 <dt><b>Boolean constants</b></dt>
1781
1782 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1783 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1784 </dd>
1785
1786 <dt><b>Integer constants</b></dt>
1787
1788 <dd>Standard integers (such as '4') are constants of the <a
1789 href="#t_integer">integer</a> type. Negative numbers may be used with
1790 integer types.
1791 </dd>
1792
1793 <dt><b>Floating point constants</b></dt>
1794
1795 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1796 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1797 notation (see below). The assembler requires the exact decimal value of
1798 a floating-point constant. For example, the assembler accepts 1.25 but
1799 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1800 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1801
1802 <dt><b>Null pointer constants</b></dt>
1803
1804 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1805 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1806
1807</dl>
1808
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1809<p>The one non-intuitive notation for constants is the hexadecimal form
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1810of floating point constants. For example, the form '<tt>double
18110x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
18124.5e+15</tt>'. The only time hexadecimal floating point constants are required
1813(and the only time that they are generated by the disassembler) is when a
1814floating point constant must be emitted but it cannot be represented as a
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1815decimal floating point number in a reasonable number of digits. For example,
1816NaN's, infinities, and other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1817special values are represented in their IEEE hexadecimal format so that
1818assembly and disassembly do not cause any bits to change in the constants.</p>
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1819<p>When using the hexadecimal form, constants of types float and double are
1820represented using the 16-digit form shown above (which matches the IEEE754
1821representation for double); float values must, however, be exactly representable
1822as IEE754 single precision.
1823Hexadecimal format is always used for long
1824double, and there are three forms of long double. The 80-bit
1825format used by x86 is represented as <tt>0xK</tt>
1826followed by 20 hexadecimal digits.
1827The 128-bit format used by PowerPC (two adjacent doubles) is represented
1828by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit
1829format is represented
1830by <tt>0xL</tt> followed by 32 hexadecimal digits; no currently supported
1831target uses this format. Long doubles will only work if they match
1832the long double format on your target. All hexadecimal formats are big-endian
1833(sign bit at the left).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1834</div>
1835
1836<!-- ======================================================================= -->
Chris Lattner97063852009-02-28 18:32:25 +0000[diff] [blame]1837<div class="doc_subsection">
1838<a name="aggregateconstants"> <!-- old anchor -->
1839<a name="complexconstants">Complex Constants</a></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1840</div>
1841
1842<div class="doc_text">
Chris Lattner97063852009-02-28 18:32:25 +0000[diff] [blame]1843<p>Complex constants are a (potentially recursive) combination of simple
1844constants and smaller complex constants.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1845
1846<dl>
1847 <dt><b>Structure constants</b></dt>
1848
1849 <dd>Structure constants are represented with notation similar to structure
1850 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1851 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1852 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1853 must have <a href="#t_struct">structure type</a>, and the number and
1854 types of elements must match those specified by the type.
1855 </dd>
1856
1857 <dt><b>Array constants</b></dt>
1858
1859 <dd>Array constants are represented with notation similar to array type
1860 definitions (a comma separated list of elements, surrounded by square brackets
1861 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1862 constants must have <a href="#t_array">array type</a>, and the number and
1863 types of elements must match those specified by the type.
1864 </dd>
1865
1866 <dt><b>Vector constants</b></dt>
1867
1868 <dd>Vector constants are represented with notation similar to vector type
1869 definitions (a comma separated list of elements, surrounded by
1870 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1871 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1872 href="#t_vector">vector type</a>, and the number and types of elements must
1873 match those specified by the type.
1874 </dd>
1875
1876 <dt><b>Zero initialization</b></dt>
1877
1878 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1879 value to zero of <em>any</em> type, including scalar and aggregate types.
1880 This is often used to avoid having to print large zero initializers (e.g. for
1881 large arrays) and is always exactly equivalent to using explicit zero
1882 initializers.
1883 </dd>
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame]1884
1885 <dt><b>Metadata node</b></dt>
1886
Nick Lewyckyf122c7e2009-05-30 16:08:30 +0000[diff] [blame]1887 <dd>A metadata node is a structure-like constant with
1888 <a href="#t_metadata">metadata type</a>. For example:
1889 "<tt>metadata !{ i32 0, metadata !"test" }</tt>". Unlike other constants
1890 that are meant to be interpreted as part of the instruction stream, metadata
1891 is a place to attach additional information such as debug info.
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame]1892 </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1893</dl>
1894
1895</div>
1896
1897<!-- ======================================================================= -->
1898<div class="doc_subsection">
1899 <a name="globalconstants">Global Variable and Function Addresses</a>
1900</div>
1901
1902<div class="doc_text">
1903
1904<p>The addresses of <a href="#globalvars">global variables</a> and <a
1905href="#functionstructure">functions</a> are always implicitly valid (link-time)
1906constants. These constants are explicitly referenced when the <a
1907href="#identifiers">identifier for the global</a> is used and always have <a
1908href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1909file:</p>
1910
1911<div class="doc_code">
1912<pre>
1913@X = global i32 17
1914@Y = global i32 42
1915@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1916</pre>
1917</div>
1918
1919</div>
1920
1921<!-- ======================================================================= -->
1922<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1923<div class="doc_text">
1924 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1925 no specific value. Undefined values may be of any type and be used anywhere
1926 a constant is permitted.</p>
1927
1928 <p>Undefined values indicate to the compiler that the program is well defined
1929 no matter what value is used, giving the compiler more freedom to optimize.
1930 </p>
1931</div>
1932
1933<!-- ======================================================================= -->
1934<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1935</div>
1936
1937<div class="doc_text">
1938
1939<p>Constant expressions are used to allow expressions involving other constants
1940to be used as constants. Constant expressions may be of any <a
1941href="#t_firstclass">first class</a> type and may involve any LLVM operation
1942that does not have side effects (e.g. load and call are not supported). The
1943following is the syntax for constant expressions:</p>
1944
1945<dl>
1946 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1947 <dd>Truncate a constant to another type. The bit size of CST must be larger
1948 than the bit size of TYPE. Both types must be integers.</dd>
1949
1950 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1951 <dd>Zero extend a constant to another type. The bit size of CST must be
1952 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1953
1954 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1955 <dd>Sign extend a constant to another type. The bit size of CST must be
1956 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1957
1958 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1959 <dd>Truncate a floating point constant to another floating point type. The
1960 size of CST must be larger than the size of TYPE. Both types must be
1961 floating point.</dd>
1962
1963 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1964 <dd>Floating point extend a constant to another type. The size of CST must be
1965 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1966
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1967 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1968 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1969 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1970 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1971 of the same number of elements. If the value won't fit in the integer type,
1972 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1973
1974 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1975 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1976 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1977 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1978 of the same number of elements. If the value won't fit in the integer type,
1979 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1980
1981 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1982 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1983 constant. TYPE must be a scalar or vector floating point type. CST must be of
1984 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1985 of the same number of elements. If the value won't fit in the floating point
1986 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1987
1988 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1989 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1990 constant. TYPE must be a scalar or vector floating point type. CST must be of
1991 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1992 of the same number of elements. If the value won't fit in the floating point
1993 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1994
1995 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1996 <dd>Convert a pointer typed constant to the corresponding integer constant
1997 TYPE must be an integer type. CST must be of pointer type. The CST value is
1998 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1999
2000 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2001 <dd>Convert a integer constant to a pointer constant. TYPE must be a
2002 pointer type. CST must be of integer type. The CST value is zero extended,
2003 truncated, or unchanged to make it fit in a pointer size. This one is
2004 <i>really</i> dangerous!</dd>
2005
2006 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
Chris Lattner557bc5d2009-02-28 18:27:03 +0000[diff] [blame]2007 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2008 are the same as those for the <a href="#i_bitcast">bitcast
2009 instruction</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2010
2011 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2012
2013 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2014 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2015 instruction, the index list may have zero or more indexes, which are required
2016 to make sense for the type of "CSTPTR".</dd>
2017
2018 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2019
2020 <dd>Perform the <a href="#i_select">select operation</a> on
2021 constants.</dd>
2022
2023 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2024 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2025
2026 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2027 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2028
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]2029 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
2030 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
2031
2032 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
2033 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
2034
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2035 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2036
2037 <dd>Perform the <a href="#i_extractelement">extractelement
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2038 operation</a> on constants.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2039
2040 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2041
2042 <dd>Perform the <a href="#i_insertelement">insertelement
2043 operation</a> on constants.</dd>
2044
2045
2046 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2047
2048 <dd>Perform the <a href="#i_shufflevector">shufflevector
2049 operation</a> on constants.</dd>
2050
2051 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2052
2053 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2054 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
2055 binary</a> operations. The constraints on operands are the same as those for
2056 the corresponding instruction (e.g. no bitwise operations on floating point
2057 values are allowed).</dd>
2058</dl>
2059</div>
2060
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame]2061<!-- ======================================================================= -->
2062<div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2063</div>
2064
2065<div class="doc_text">
2066
2067<p>Embedded metadata provides a way to attach arbitrary data to the
2068instruction stream without affecting the behaviour of the program. There are
Nick Lewycky29aaef82009-05-30 05:06:04 +0000[diff] [blame]2069two metadata primitives, strings and nodes. All metadata has the
2070<tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2071point ('<tt>!</tt>').
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame]2072</p>
2073
2074<p>A metadata string is a string surrounded by double quotes. It can contain
2075any character by escaping non-printable characters with "\xx" where "xx" is
2076the two digit hex code. For example: "<tt>!"test\00"</tt>".
2077</p>
2078
2079<p>Metadata nodes are represented with notation similar to structure constants
2080(a comma separated list of elements, surrounded by braces and preceeded by an
Nick Lewycky29aaef82009-05-30 05:06:04 +0000[diff] [blame]2081exclamation point). For example: "<tt>!{ metadata !"test\00", i32 10}</tt>".
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame]2082</p>
2083
Nick Lewycky117f4382009-05-10 20:57:05 +0000[diff] [blame]2084<p>A metadata node will attempt to track changes to the values it holds. In
2085the event that a value is deleted, it will be replaced with a typeless
Nick Lewycky29aaef82009-05-30 05:06:04 +0000[diff] [blame]2086"<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
Nick Lewycky117f4382009-05-10 20:57:05 +0000[diff] [blame]2087
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame]2088<p>Optimizations may rely on metadata to provide additional information about
2089the program that isn't available in the instructions, or that isn't easily
2090computable. Similarly, the code generator may expect a certain metadata format
2091to be used to express debugging information.</p>
2092</div>
2093
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2094<!-- *********************************************************************** -->
2095<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2096<!-- *********************************************************************** -->
2097
2098<!-- ======================================================================= -->
2099<div class="doc_subsection">
2100<a name="inlineasm">Inline Assembler Expressions</a>
2101</div>
2102
2103<div class="doc_text">
2104
2105<p>
2106LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
2107Module-Level Inline Assembly</a>) through the use of a special value. This
2108value represents the inline assembler as a string (containing the instructions
2109to emit), a list of operand constraints (stored as a string), and a flag that
2110indicates whether or not the inline asm expression has side effects. An example
2111inline assembler expression is:
2112</p>
2113
2114<div class="doc_code">
2115<pre>
2116i32 (i32) asm "bswap $0", "=r,r"
2117</pre>
2118</div>
2119
2120<p>
2121Inline assembler expressions may <b>only</b> be used as the callee operand of
2122a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2123</p>
2124
2125<div class="doc_code">
2126<pre>
2127%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2128</pre>
2129</div>
2130
2131<p>
2132Inline asms with side effects not visible in the constraint list must be marked
2133as having side effects. This is done through the use of the
2134'<tt>sideeffect</tt>' keyword, like so:
2135</p>
2136
2137<div class="doc_code">
2138<pre>
2139call void asm sideeffect "eieio", ""()
2140</pre>
2141</div>
2142
2143<p>TODO: The format of the asm and constraints string still need to be
2144documented here. Constraints on what can be done (e.g. duplication, moving, etc
Chris Lattner1cdd64e2008-10-04 18:36:02 +0000[diff] [blame]2145need to be documented). This is probably best done by reference to another
2146document that covers inline asm from a holistic perspective.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2147</p>
2148
2149</div>
2150
2151<!-- *********************************************************************** -->
2152<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2153<!-- *********************************************************************** -->
2154
2155<div class="doc_text">
2156
2157<p>The LLVM instruction set consists of several different
2158classifications of instructions: <a href="#terminators">terminator
2159instructions</a>, <a href="#binaryops">binary instructions</a>,
2160<a href="#bitwiseops">bitwise binary instructions</a>, <a
2161 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2162instructions</a>.</p>
2163
2164</div>
2165
2166<!-- ======================================================================= -->
2167<div class="doc_subsection"> <a name="terminators">Terminator
2168Instructions</a> </div>
2169
2170<div class="doc_text">
2171
2172<p>As mentioned <a href="#functionstructure">previously</a>, every
2173basic block in a program ends with a "Terminator" instruction, which
2174indicates which block should be executed after the current block is
2175finished. These terminator instructions typically yield a '<tt>void</tt>'
2176value: they produce control flow, not values (the one exception being
2177the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2178<p>There are six different terminator instructions: the '<a
2179 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2180instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
2181the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2182 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2183 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2184
2185</div>
2186
2187<!-- _______________________________________________________________________ -->
2188<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2189Instruction</a> </div>
2190<div class="doc_text">
2191<h5>Syntax:</h5>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2192<pre>
2193 ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2194 ret void <i>; Return from void function</i>
2195</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2196
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2197<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2198
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2199<p>The '<tt>ret</tt>' instruction is used to return control flow (and
2200optionally a value) from a function back to the caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2201<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2202returns a value and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2203control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2204
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2205<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2206
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2207<p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2208the return value. The type of the return value must be a
2209'<a href="#t_firstclass">first class</a>' type.</p>
2210
2211<p>A function is not <a href="#wellformed">well formed</a> if
2212it it has a non-void return type and contains a '<tt>ret</tt>'
2213instruction with no return value or a return value with a type that
2214does not match its type, or if it has a void return type and contains
2215a '<tt>ret</tt>' instruction with a return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2216
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2217<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2218
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2219<p>When the '<tt>ret</tt>' instruction is executed, control flow
2220returns back to the calling function's context. If the caller is a "<a
2221 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2222the instruction after the call. If the caller was an "<a
2223 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2224at the beginning of the "normal" destination block. If the instruction
2225returns a value, that value shall set the call or invoke instruction's
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2226return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2227
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2228<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2229
2230<pre>
2231 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2232 ret void <i>; Return from a void function</i>
Bill Wendlingd163e2d2009-02-28 22:12:54 +0000[diff] [blame]2233 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2234</pre>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]2235
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]2236<p>Note that the code generator does not yet fully support large
2237 return values. The specific sizes that are currently supported are
2238 dependent on the target. For integers, on 32-bit targets the limit
2239 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2240 For aggregate types, the current limits are dependent on the element
2241 types; for example targets are often limited to 2 total integer
2242 elements and 2 total floating-point elements.</p>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]2243
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2244</div>
2245<!-- _______________________________________________________________________ -->
2246<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2247<div class="doc_text">
2248<h5>Syntax:</h5>
2249<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
2250</pre>
2251<h5>Overview:</h5>
2252<p>The '<tt>br</tt>' instruction is used to cause control flow to
2253transfer to a different basic block in the current function. There are
2254two forms of this instruction, corresponding to a conditional branch
2255and an unconditional branch.</p>
2256<h5>Arguments:</h5>
2257<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2258single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2259unconditional form of the '<tt>br</tt>' instruction takes a single
2260'<tt>label</tt>' value as a target.</p>
2261<h5>Semantics:</h5>
2262<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2263argument is evaluated. If the value is <tt>true</tt>, control flows
2264to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2265control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2266<h5>Example:</h5>
Chris Lattner95127832009-05-09 18:11:50 +0000[diff] [blame]2267<pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b<br> br i1 %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2268 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2269</div>
2270<!-- _______________________________________________________________________ -->
2271<div class="doc_subsubsection">
2272 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2273</div>
2274
2275<div class="doc_text">
2276<h5>Syntax:</h5>
2277
2278<pre>
2279 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
2280</pre>
2281
2282<h5>Overview:</h5>
2283
2284<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2285several different places. It is a generalization of the '<tt>br</tt>'
2286instruction, allowing a branch to occur to one of many possible
2287destinations.</p>
2288
2289
2290<h5>Arguments:</h5>
2291
2292<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2293comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2294an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2295table is not allowed to contain duplicate constant entries.</p>
2296
2297<h5>Semantics:</h5>
2298
2299<p>The <tt>switch</tt> instruction specifies a table of values and
2300destinations. When the '<tt>switch</tt>' instruction is executed, this
2301table is searched for the given value. If the value is found, control flow is
2302transfered to the corresponding destination; otherwise, control flow is
2303transfered to the default destination.</p>
2304
2305<h5>Implementation:</h5>
2306
2307<p>Depending on properties of the target machine and the particular
2308<tt>switch</tt> instruction, this instruction may be code generated in different
2309ways. For example, it could be generated as a series of chained conditional
2310branches or with a lookup table.</p>
2311
2312<h5>Example:</h5>
2313
2314<pre>
2315 <i>; Emulate a conditional br instruction</i>
2316 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]2317 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2318
2319 <i>; Emulate an unconditional br instruction</i>
2320 switch i32 0, label %dest [ ]
2321
2322 <i>; Implement a jump table:</i>
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]2323 switch i32 %val, label %otherwise [ i32 0, label %onzero
2324 i32 1, label %onone
2325 i32 2, label %ontwo ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2326</pre>
2327</div>
2328
2329<!-- _______________________________________________________________________ -->
2330<div class="doc_subsubsection">
2331 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2332</div>
2333
2334<div class="doc_text">
2335
2336<h5>Syntax:</h5>
2337
2338<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2339 &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 +0000[diff] [blame]2340 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2341</pre>
2342
2343<h5>Overview:</h5>
2344
2345<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2346function, with the possibility of control flow transfer to either the
2347'<tt>normal</tt>' label or the
2348'<tt>exception</tt>' label. If the callee function returns with the
2349"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2350"normal" label. If the callee (or any indirect callees) returns with the "<a
2351href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2352continued at the dynamically nearest "exception" label.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2353
2354<h5>Arguments:</h5>
2355
2356<p>This instruction requires several arguments:</p>
2357
2358<ol>
2359 <li>
2360 The optional "cconv" marker indicates which <a href="#callingconv">calling
2361 convention</a> the call should use. If none is specified, the call defaults
2362 to using C calling conventions.
2363 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2364
2365 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2366 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2367 and '<tt>inreg</tt>' attributes are valid here.</li>
2368
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2369 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2370 function value being invoked. In most cases, this is a direct function
2371 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2372 an arbitrary pointer to function value.
2373 </li>
2374
2375 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2376 function to be invoked. </li>
2377
2378 <li>'<tt>function args</tt>': argument list whose types match the function
2379 signature argument types. If the function signature indicates the function
2380 accepts a variable number of arguments, the extra arguments can be
2381 specified. </li>
2382
2383 <li>'<tt>normal label</tt>': the label reached when the called function
2384 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2385
2386 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2387 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2388
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2389 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2390 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2391 '<tt>readnone</tt>' attributes are valid here.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2392</ol>
2393
2394<h5>Semantics:</h5>
2395
2396<p>This instruction is designed to operate as a standard '<tt><a
2397href="#i_call">call</a></tt>' instruction in most regards. The primary
2398difference is that it establishes an association with a label, which is used by
2399the runtime library to unwind the stack.</p>
2400
2401<p>This instruction is used in languages with destructors to ensure that proper
2402cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2403exception. Additionally, this is important for implementation of
2404'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2405
Dan Gohman140ba5d2009-05-22 21:47:08 +0000[diff] [blame]2406<p>It is not valid to reference the return value of an invoke call from
2407anywhere not dominated by the normal label, since an unwind does not
2408provide a return value.</p>
2409
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2410<h5>Example:</h5>
2411<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2412 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2413 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2414 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2415 unwind label %TestCleanup <i>; {i32}:retval set</i>
2416</pre>
2417</div>
2418
2419
2420<!-- _______________________________________________________________________ -->
2421
2422<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2423Instruction</a> </div>
2424
2425<div class="doc_text">
2426
2427<h5>Syntax:</h5>
2428<pre>
2429 unwind
2430</pre>
2431
2432<h5>Overview:</h5>
2433
2434<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2435at the first callee in the dynamic call stack which used an <a
2436href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2437primarily used to implement exception handling.</p>
2438
2439<h5>Semantics:</h5>
2440
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2441<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2442immediately halt. The dynamic call stack is then searched for the first <a
2443href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2444execution continues at the "exceptional" destination block specified by the
2445<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2446dynamic call chain, undefined behavior results.</p>
2447</div>
2448
2449<!-- _______________________________________________________________________ -->
2450
2451<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2452Instruction</a> </div>
2453
2454<div class="doc_text">
2455
2456<h5>Syntax:</h5>
2457<pre>
2458 unreachable
2459</pre>
2460
2461<h5>Overview:</h5>
2462
2463<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2464instruction is used to inform the optimizer that a particular portion of the
2465code is not reachable. This can be used to indicate that the code after a
2466no-return function cannot be reached, and other facts.</p>
2467
2468<h5>Semantics:</h5>
2469
2470<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2471</div>
2472
2473
2474
2475<!-- ======================================================================= -->
2476<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2477<div class="doc_text">
2478<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2479program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2480produce a single value. The operands might represent
2481multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2482The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2483<p>There are several different binary operators:</p>
2484</div>
2485<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2486<div class="doc_subsubsection">
2487 <a name="i_add">'<tt>add</tt>' Instruction</a>
2488</div>
2489
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2490<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2491
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2492<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2493
2494<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2495 &lt;result&gt; = add &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2496</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2497
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2498<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2499
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2500<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2501
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2502<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2503
2504<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2505 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2506 <a href="#t_vector">vector</a> values. Both arguments must have identical
2507 types.</p>
2508
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2509<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2510
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2511<p>The value produced is the integer or floating point sum of the two
2512operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2513
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2514<p>If an integer sum has unsigned overflow, the result returned is the
2515mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2516the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2517
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2518<p>Because LLVM integers use a two's complement representation, this
2519instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2520
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2521<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2522
2523<pre>
2524 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2525</pre>
2526</div>
2527<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2528<div class="doc_subsubsection">
2529 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2530</div>
2531
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2532<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2533
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2534<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2535
2536<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2537 &lt;result&gt; = sub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2538</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2539
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2540<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2541
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2542<p>The '<tt>sub</tt>' instruction returns the difference of its two
2543operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2544
2545<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2546'<tt>neg</tt>' instruction present in most other intermediate
2547representations.</p>
2548
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2549<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2550
2551<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2552 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2553 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2554 types.</p>
2555
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2556<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2557
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2558<p>The value produced is the integer or floating point difference of
2559the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2560
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2561<p>If an integer difference has unsigned overflow, the result returned is the
2562mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2563the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2564
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2565<p>Because LLVM integers use a two's complement representation, this
2566instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2567
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2568<h5>Example:</h5>
2569<pre>
2570 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2571 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2572</pre>
2573</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2574
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2575<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2576<div class="doc_subsubsection">
2577 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2578</div>
2579
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2580<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2581
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2582<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2583<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2584</pre>
2585<h5>Overview:</h5>
2586<p>The '<tt>mul</tt>' instruction returns the product of its two
2587operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2588
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2589<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2590
2591<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2592href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2593or <a href="#t_vector">vector</a> values. Both arguments must have identical
2594types.</p>
2595
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2596<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2597
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2598<p>The value produced is the integer or floating point product of the
2599two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2600
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2601<p>If the result of an integer multiplication has unsigned overflow,
2602the result returned is the mathematical result modulo
26032<sup>n</sup>, where n is the bit width of the result.</p>
2604<p>Because LLVM integers use a two's complement representation, and the
2605result is the same width as the operands, this instruction returns the
2606correct result for both signed and unsigned integers. If a full product
2607(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2608should be sign-extended or zero-extended as appropriate to the
2609width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2610<h5>Example:</h5>
2611<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2612</pre>
2613</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2614
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2615<!-- _______________________________________________________________________ -->
2616<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2617</a></div>
2618<div class="doc_text">
2619<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2620<pre> &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 +0000[diff] [blame]2621</pre>
2622<h5>Overview:</h5>
2623<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2624operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2625
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2626<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2627
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2628<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2629<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2630values. Both arguments must have identical types.</p>
2631
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2632<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2633
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2634<p>The value produced is the unsigned integer quotient of the two operands.</p>
2635<p>Note that unsigned integer division and signed integer division are distinct
2636operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2637<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2638<h5>Example:</h5>
2639<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2640</pre>
2641</div>
2642<!-- _______________________________________________________________________ -->
2643<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2644</a> </div>
2645<div class="doc_text">
2646<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2647<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2648 &lt;result&gt; = sdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2649</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2650
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2651<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2652
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2653<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2654operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2655
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2656<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2657
2658<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2659<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2660values. Both arguments must have identical types.</p>
2661
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2662<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2663<p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2664<p>Note that signed integer division and unsigned integer division are distinct
2665operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2666<p>Division by zero leads to undefined behavior. Overflow also leads to
2667undefined behavior; this is a rare case, but can occur, for example,
2668by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2669<h5>Example:</h5>
2670<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2671</pre>
2672</div>
2673<!-- _______________________________________________________________________ -->
2674<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2675Instruction</a> </div>
2676<div class="doc_text">
2677<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2678<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2679 &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 +0000[diff] [blame]2680</pre>
2681<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2682
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2683<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2684operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2685
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2686<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2687
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2688<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2689<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2690of floating point values. Both arguments must have identical types.</p>
2691
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2692<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2693
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2694<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2695
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2696<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2697
2698<pre>
2699 &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2700</pre>
2701</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2702
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2703<!-- _______________________________________________________________________ -->
2704<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2705</div>
2706<div class="doc_text">
2707<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2708<pre> &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 +0000[diff] [blame]2709</pre>
2710<h5>Overview:</h5>
2711<p>The '<tt>urem</tt>' instruction returns the remainder from the
2712unsigned division of its two arguments.</p>
2713<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2714<p>The two arguments to the '<tt>urem</tt>' instruction must be
2715<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2716values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2717<h5>Semantics:</h5>
2718<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2719This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2720<p>Note that unsigned integer remainder and signed integer remainder are
2721distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2722<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2723<h5>Example:</h5>
2724<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2725</pre>
2726
2727</div>
2728<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2729<div class="doc_subsubsection">
2730 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2731</div>
2732
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2733<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2734
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2735<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2736
2737<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2738 &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 +0000[diff] [blame]2739</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2740
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2741<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2742
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2743<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2744signed division of its two operands. This instruction can also take
2745<a href="#t_vector">vector</a> versions of the values in which case
2746the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2747
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2748<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2749
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2750<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2751<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2752values. Both arguments must have identical types.</p>
2753
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2754<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2755
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2756<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2757has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2758operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2759a value. For more information about the difference, see <a
2760 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2761Math Forum</a>. For a table of how this is implemented in various languages,
2762please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2763Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2764<p>Note that signed integer remainder and unsigned integer remainder are
2765distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2766<p>Taking the remainder of a division by zero leads to undefined behavior.
2767Overflow also leads to undefined behavior; this is a rare case, but can occur,
2768for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2769(The remainder doesn't actually overflow, but this rule lets srem be
2770implemented using instructions that return both the result of the division
2771and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2772<h5>Example:</h5>
2773<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2774</pre>
2775
2776</div>
2777<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2778<div class="doc_subsubsection">
2779 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2780
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2781<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2782
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2783<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2784<pre> &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 +0000[diff] [blame]2785</pre>
2786<h5>Overview:</h5>
2787<p>The '<tt>frem</tt>' instruction returns the remainder from the
2788division of its two operands.</p>
2789<h5>Arguments:</h5>
2790<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2791<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2792of floating point values. Both arguments must have identical types.</p>
2793
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2794<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2795
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2796<p>This instruction returns the <i>remainder</i> of a division.
2797The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2798
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2799<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2800
2801<pre>
2802 &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2803</pre>
2804</div>
2805
2806<!-- ======================================================================= -->
2807<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2808Operations</a> </div>
2809<div class="doc_text">
2810<p>Bitwise binary operators are used to do various forms of
2811bit-twiddling in a program. They are generally very efficient
2812instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2813instructions. They require two operands of the same type, execute an operation on them,
2814and produce a single value. The resulting value is the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2815</div>
2816
2817<!-- _______________________________________________________________________ -->
2818<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2819Instruction</a> </div>
2820<div class="doc_text">
2821<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2822<pre> &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 +0000[diff] [blame]2823</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2824
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2825<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2826
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2827<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2828the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2829
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2830<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2831
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2832<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2833 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2834type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2835
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2836<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2837
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2838<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2839where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
Mon P Wangb0f51822008-12-10 08:55:09 +0000[diff] [blame]2840equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2841If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2842corresponding shift amount in <tt>op2</tt>.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2843
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2844<h5>Example:</h5><pre>
2845 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2846 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2847 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2848 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2849 &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 +0000[diff] [blame]2850</pre>
2851</div>
2852<!-- _______________________________________________________________________ -->
2853<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2854Instruction</a> </div>
2855<div class="doc_text">
2856<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2857<pre> &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 +0000[diff] [blame]2858</pre>
2859
2860<h5>Overview:</h5>
2861<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2862operand shifted to the right a specified number of bits with zero fill.</p>
2863
2864<h5>Arguments:</h5>
2865<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2866<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2867type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2868
2869<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2870
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2871<p>This instruction always performs a logical shift right operation. The most
2872significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2873shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
Mon P Wangb0f51822008-12-10 08:55:09 +0000[diff] [blame]2874the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2875vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2876amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2877
2878<h5>Example:</h5>
2879<pre>
2880 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2881 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2882 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2883 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2884 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2885 &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 +0000[diff] [blame]2886</pre>
2887</div>
2888
2889<!-- _______________________________________________________________________ -->
2890<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2891Instruction</a> </div>
2892<div class="doc_text">
2893
2894<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2895<pre> &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 +0000[diff] [blame]2896</pre>
2897
2898<h5>Overview:</h5>
2899<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2900operand shifted to the right a specified number of bits with sign extension.</p>
2901
2902<h5>Arguments:</h5>
2903<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2904<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2905type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2906
2907<h5>Semantics:</h5>
2908<p>This instruction always performs an arithmetic shift right operation,
2909The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2910of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
Mon P Wangb0f51822008-12-10 08:55:09 +0000[diff] [blame]2911larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2912arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2913corresponding shift amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2914
2915<h5>Example:</h5>
2916<pre>
2917 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2918 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2919 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2920 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2921 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2922 &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 +0000[diff] [blame]2923</pre>
2924</div>
2925
2926<!-- _______________________________________________________________________ -->
2927<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2928Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2929
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2930<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2931
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2932<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2933
2934<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2935 &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 +0000[diff] [blame]2936</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2937
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2938<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2939
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2940<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2941its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2942
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2943<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2944
2945<p>The two arguments to the '<tt>and</tt>' instruction must be
2946<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2947values. Both arguments must have identical types.</p>
2948
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2949<h5>Semantics:</h5>
2950<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2951<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2952<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2953<table border="1" cellspacing="0" cellpadding="4">
2954 <tbody>
2955 <tr>
2956 <td>In0</td>
2957 <td>In1</td>
2958 <td>Out</td>
2959 </tr>
2960 <tr>
2961 <td>0</td>
2962 <td>0</td>
2963 <td>0</td>
2964 </tr>
2965 <tr>
2966 <td>0</td>
2967 <td>1</td>
2968 <td>0</td>
2969 </tr>
2970 <tr>
2971 <td>1</td>
2972 <td>0</td>
2973 <td>0</td>
2974 </tr>
2975 <tr>
2976 <td>1</td>
2977 <td>1</td>
2978 <td>1</td>
2979 </tr>
2980 </tbody>
2981</table>
2982</div>
2983<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2984<pre>
2985 &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2986 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2987 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2988</pre>
2989</div>
2990<!-- _______________________________________________________________________ -->
2991<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2992<div class="doc_text">
2993<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2994<pre> &lt;result&gt; = or &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2995</pre>
2996<h5>Overview:</h5>
2997<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2998or of its two operands.</p>
2999<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3000
3001<p>The two arguments to the '<tt>or</tt>' instruction must be
3002<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3003values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3004<h5>Semantics:</h5>
3005<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3006<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]3007<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3008<table border="1" cellspacing="0" cellpadding="4">
3009 <tbody>
3010 <tr>
3011 <td>In0</td>
3012 <td>In1</td>
3013 <td>Out</td>
3014 </tr>
3015 <tr>
3016 <td>0</td>
3017 <td>0</td>
3018 <td>0</td>
3019 </tr>
3020 <tr>
3021 <td>0</td>
3022 <td>1</td>
3023 <td>1</td>
3024 </tr>
3025 <tr>
3026 <td>1</td>
3027 <td>0</td>
3028 <td>1</td>
3029 </tr>
3030 <tr>
3031 <td>1</td>
3032 <td>1</td>
3033 <td>1</td>
3034 </tr>
3035 </tbody>
3036</table>
3037</div>
3038<h5>Example:</h5>
3039<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3040 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3041 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3042</pre>
3043</div>
3044<!-- _______________________________________________________________________ -->
3045<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3046Instruction</a> </div>
3047<div class="doc_text">
3048<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3049<pre> &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 +0000[diff] [blame]3050</pre>
3051<h5>Overview:</h5>
3052<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
3053or of its two operands. The <tt>xor</tt> is used to implement the
3054"one's complement" operation, which is the "~" operator in C.</p>
3055<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3056<p>The two arguments to the '<tt>xor</tt>' instruction must be
3057<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3058values. Both arguments must have identical types.</p>
3059
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3060<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3061
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3062<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3063<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]3064<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3065<table border="1" cellspacing="0" cellpadding="4">
3066 <tbody>
3067 <tr>
3068 <td>In0</td>
3069 <td>In1</td>
3070 <td>Out</td>
3071 </tr>
3072 <tr>
3073 <td>0</td>
3074 <td>0</td>
3075 <td>0</td>
3076 </tr>
3077 <tr>
3078 <td>0</td>
3079 <td>1</td>
3080 <td>1</td>
3081 </tr>
3082 <tr>
3083 <td>1</td>
3084 <td>0</td>
3085 <td>1</td>
3086 </tr>
3087 <tr>
3088 <td>1</td>
3089 <td>1</td>
3090 <td>0</td>
3091 </tr>
3092 </tbody>
3093</table>
3094</div>
3095<p> </p>
3096<h5>Example:</h5>
3097<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3098 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3099 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3100 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3101</pre>
3102</div>
3103
3104<!-- ======================================================================= -->
3105<div class="doc_subsection">
3106 <a name="vectorops">Vector Operations</a>
3107</div>
3108
3109<div class="doc_text">
3110
3111<p>LLVM supports several instructions to represent vector operations in a
3112target-independent manner. These instructions cover the element-access and
3113vector-specific operations needed to process vectors effectively. While LLVM
3114does directly support these vector operations, many sophisticated algorithms
3115will want to use target-specific intrinsics to take full advantage of a specific
3116target.</p>
3117
3118</div>
3119
3120<!-- _______________________________________________________________________ -->
3121<div class="doc_subsubsection">
3122 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3123</div>
3124
3125<div class="doc_text">
3126
3127<h5>Syntax:</h5>
3128
3129<pre>
3130 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
3131</pre>
3132
3133<h5>Overview:</h5>
3134
3135<p>
3136The '<tt>extractelement</tt>' instruction extracts a single scalar
3137element from a vector at a specified index.
3138</p>
3139
3140
3141<h5>Arguments:</h5>
3142
3143<p>
3144The first operand of an '<tt>extractelement</tt>' instruction is a
3145value of <a href="#t_vector">vector</a> type. The second operand is
3146an index indicating the position from which to extract the element.
3147The index may be a variable.</p>
3148
3149<h5>Semantics:</h5>
3150
3151<p>
3152The result is a scalar of the same type as the element type of
3153<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3154<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3155results are undefined.
3156</p>
3157
3158<h5>Example:</h5>
3159
3160<pre>
3161 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
3162</pre>
3163</div>
3164
3165
3166<!-- _______________________________________________________________________ -->
3167<div class="doc_subsubsection">
3168 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3169</div>
3170
3171<div class="doc_text">
3172
3173<h5>Syntax:</h5>
3174
3175<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]3176 &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 +0000[diff] [blame]3177</pre>
3178
3179<h5>Overview:</h5>
3180
3181<p>
3182The '<tt>insertelement</tt>' instruction inserts a scalar
3183element into a vector at a specified index.
3184</p>
3185
3186
3187<h5>Arguments:</h5>
3188
3189<p>
3190The first operand of an '<tt>insertelement</tt>' instruction is a
3191value of <a href="#t_vector">vector</a> type. The second operand is a
3192scalar value whose type must equal the element type of the first
3193operand. The third operand is an index indicating the position at
3194which to insert the value. The index may be a variable.</p>
3195
3196<h5>Semantics:</h5>
3197
3198<p>
3199The result is a vector of the same type as <tt>val</tt>. Its
3200element values are those of <tt>val</tt> except at position
3201<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3202exceeds the length of <tt>val</tt>, the results are undefined.
3203</p>
3204
3205<h5>Example:</h5>
3206
3207<pre>
3208 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
3209</pre>
3210</div>
3211
3212<!-- _______________________________________________________________________ -->
3213<div class="doc_subsubsection">
3214 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3215</div>
3216
3217<div class="doc_text">
3218
3219<h5>Syntax:</h5>
3220
3221<pre>
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3222 &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 +0000[diff] [blame]3223</pre>
3224
3225<h5>Overview:</h5>
3226
3227<p>
3228The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3229from two input vectors, returning a vector with the same element type as
3230the input and length that is the same as the shuffle mask.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3231</p>
3232
3233<h5>Arguments:</h5>
3234
3235<p>
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3236The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3237with types that match each other. The third argument is a shuffle mask whose
3238element type is always 'i32'. The result of the instruction is a vector whose
3239length is the same as the shuffle mask and whose element type is the same as
3240the element type of the first two operands.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3241</p>
3242
3243<p>
3244The shuffle mask operand is required to be a constant vector with either
3245constant integer or undef values.
3246</p>
3247
3248<h5>Semantics:</h5>
3249
3250<p>
3251The elements of the two input vectors are numbered from left to right across
3252both of the vectors. The shuffle mask operand specifies, for each element of
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3253the result vector, which element of the two input vectors the result element
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3254gets. The element selector may be undef (meaning "don't care") and the second
3255operand may be undef if performing a shuffle from only one vector.
3256</p>
3257
3258<h5>Example:</h5>
3259
3260<pre>
3261 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3262 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
3263 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
3264 &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 +0000[diff] [blame]3265 %result = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
3266 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i>
3267 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3268 &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 +0000[diff] [blame]3269</pre>
3270</div>
3271
3272
3273<!-- ======================================================================= -->
3274<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3275 <a name="aggregateops">Aggregate Operations</a>
3276</div>
3277
3278<div class="doc_text">
3279
3280<p>LLVM supports several instructions for working with aggregate values.
3281</p>
3282
3283</div>
3284
3285<!-- _______________________________________________________________________ -->
3286<div class="doc_subsubsection">
3287 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3288</div>
3289
3290<div class="doc_text">
3291
3292<h5>Syntax:</h5>
3293
3294<pre>
3295 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
3296</pre>
3297
3298<h5>Overview:</h5>
3299
3300<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3301The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3302or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3303</p>
3304
3305
3306<h5>Arguments:</h5>
3307
3308<p>
3309The first operand of an '<tt>extractvalue</tt>' instruction is a
3310value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3311type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3312in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3313'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3314</p>
3315
3316<h5>Semantics:</h5>
3317
3318<p>
3319The result is the value at the position in the aggregate specified by
3320the index operands.
3321</p>
3322
3323<h5>Example:</h5>
3324
3325<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3326 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3327</pre>
3328</div>
3329
3330
3331<!-- _______________________________________________________________________ -->
3332<div class="doc_subsubsection">
3333 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3334</div>
3335
3336<div class="doc_text">
3337
3338<h5>Syntax:</h5>
3339
3340<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3341 &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 +0000[diff] [blame]3342</pre>
3343
3344<h5>Overview:</h5>
3345
3346<p>
3347The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3348into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3349</p>
3350
3351
3352<h5>Arguments:</h5>
3353
3354<p>
3355The first operand of an '<tt>insertvalue</tt>' instruction is a
3356value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3357The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3358The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3359indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3360indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3361'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3362The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3363by the indices.
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3364</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3365
3366<h5>Semantics:</h5>
3367
3368<p>
3369The result is an aggregate of the same type as <tt>val</tt>. Its
3370value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3371specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3372</p>
3373
3374<h5>Example:</h5>
3375
3376<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3377 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3378</pre>
3379</div>
3380
3381
3382<!-- ======================================================================= -->
3383<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3384 <a name="memoryops">Memory Access and Addressing Operations</a>
3385</div>
3386
3387<div class="doc_text">
3388
3389<p>A key design point of an SSA-based representation is how it
3390represents memory. In LLVM, no memory locations are in SSA form, which
3391makes things very simple. This section describes how to read, write,
3392allocate, and free memory in LLVM.</p>
3393
3394</div>
3395
3396<!-- _______________________________________________________________________ -->
3397<div class="doc_subsubsection">
3398 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3399</div>
3400
3401<div class="doc_text">
3402
3403<h5>Syntax:</h5>
3404
3405<pre>
3406 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3407</pre>
3408
3409<h5>Overview:</h5>
3410
3411<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3412heap and returns a pointer to it. The object is always allocated in the generic
3413address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3414
3415<h5>Arguments:</h5>
3416
3417<p>The '<tt>malloc</tt>' instruction allocates
3418<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3419bytes of memory from the operating system and returns a pointer of the
3420appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3421number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3422If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3423be aligned to at least that boundary. If not specified, or if zero, the target can
3424choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3425
3426<p>'<tt>type</tt>' must be a sized type.</p>
3427
3428<h5>Semantics:</h5>
3429
3430<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Nick Lewyckyff384472008-11-24 03:41:24 +0000[diff] [blame]3431a pointer is returned. The result of a zero byte allocation is undefined. The
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3432result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3433
3434<h5>Example:</h5>
3435
3436<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3437 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3438
3439 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3440 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3441 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3442 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3443 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3444</pre>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]3445
3446<p>Note that the code generator does not yet respect the
3447 alignment value.</p>
3448
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3449</div>
3450
3451<!-- _______________________________________________________________________ -->
3452<div class="doc_subsubsection">
3453 <a name="i_free">'<tt>free</tt>' Instruction</a>
3454</div>
3455
3456<div class="doc_text">
3457
3458<h5>Syntax:</h5>
3459
3460<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3461 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3462</pre>
3463
3464<h5>Overview:</h5>
3465
3466<p>The '<tt>free</tt>' instruction returns memory back to the unused
3467memory heap to be reallocated in the future.</p>
3468
3469<h5>Arguments:</h5>
3470
3471<p>'<tt>value</tt>' shall be a pointer value that points to a value
3472that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3473instruction.</p>
3474
3475<h5>Semantics:</h5>
3476
3477<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3478after this instruction executes. If the pointer is null, the operation
3479is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3480
3481<h5>Example:</h5>
3482
3483<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3484 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3485 free [4 x i8]* %array
3486</pre>
3487</div>
3488
3489<!-- _______________________________________________________________________ -->
3490<div class="doc_subsubsection">
3491 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3492</div>
3493
3494<div class="doc_text">
3495
3496<h5>Syntax:</h5>
3497
3498<pre>
3499 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3500</pre>
3501
3502<h5>Overview:</h5>
3503
3504<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3505currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3506returns to its caller. The object is always allocated in the generic address
3507space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3508
3509<h5>Arguments:</h5>
3510
3511<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3512bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3513appropriate type to the program. If "NumElements" is specified, it is the
3514number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3515If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3516to be aligned to at least that boundary. If not specified, or if zero, the target
3517can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3518
3519<p>'<tt>type</tt>' may be any sized type.</p>
3520
3521<h5>Semantics:</h5>
3522
Bill Wendling2a454572009-05-08 20:49:29 +0000[diff] [blame]3523<p>Memory is allocated; a pointer is returned. The operation is undefined if
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3524there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3525memory is automatically released when the function returns. The '<tt>alloca</tt>'
3526instruction is commonly used to represent automatic variables that must
3527have an address available. When the function returns (either with the <tt><a
3528 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3529instructions), the memory is reclaimed. Allocating zero bytes
3530is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3531
3532<h5>Example:</h5>
3533
3534<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3535 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3536 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3537 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3538 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3539</pre>
3540</div>
3541
3542<!-- _______________________________________________________________________ -->
3543<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3544Instruction</a> </div>
3545<div class="doc_text">
3546<h5>Syntax:</h5>
3547<pre> &lt;result&gt; = load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;]<br> &lt;result&gt; = volatile load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;]<br></pre>
3548<h5>Overview:</h5>
3549<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3550<h5>Arguments:</h5>
3551<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3552address from which to load. The pointer must point to a <a
3553 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3554marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3555the number or order of execution of this <tt>load</tt> with other
3556volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3557instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3558<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3559The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3560(that is, the alignment of the memory address). A value of 0 or an
3561omitted "align" argument means that the operation has the preferential
3562alignment for the target. It is the responsibility of the code emitter
3563to ensure that the alignment information is correct. Overestimating
3564the alignment results in an undefined behavior. Underestimating the
3565alignment may produce less efficient code. An alignment of 1 is always
3566safe.
3567</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3568<h5>Semantics:</h5>
Duncan Sandsfe87fff2009-03-22 11:33:16 +0000[diff] [blame]3569<p>The location of memory pointed to is loaded. If the value being loaded
3570is of scalar type then the number of bytes read does not exceed the minimum
3571number of bytes needed to hold all bits of the type. For example, loading an
3572<tt>i24</tt> reads at most three bytes. When loading a value of a type like
3573<tt>i20</tt> with a size that is not an integral number of bytes, the result
3574is undefined if the value was not originally written using a store of the
3575same type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3576<h5>Examples:</h5>
3577<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3578 <a
3579 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3580 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3581</pre>
3582</div>
3583<!-- _______________________________________________________________________ -->
3584<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3585Instruction</a> </div>
3586<div class="doc_text">
3587<h5>Syntax:</h5>
3588<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3589 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3590</pre>
3591<h5>Overview:</h5>
3592<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3593<h5>Arguments:</h5>
3594<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3595to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3596operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3597of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3598operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3599optimizer is not allowed to modify the number or order of execution of
3600this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3601 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3602<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3603The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3604(that is, the alignment of the memory address). A value of 0 or an
3605omitted "align" argument means that the operation has the preferential
3606alignment for the target. It is the responsibility of the code emitter
3607to ensure that the alignment information is correct. Overestimating
3608the alignment results in an undefined behavior. Underestimating the
3609alignment may produce less efficient code. An alignment of 1 is always
3610safe.
3611</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3612<h5>Semantics:</h5>
3613<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
Duncan Sandsfe87fff2009-03-22 11:33:16 +0000[diff] [blame]3614at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.
3615If '<tt>&lt;value&gt;</tt>' is of scalar type then the number of bytes
3616written does not exceed the minimum number of bytes needed to hold all
3617bits of the type. For example, storing an <tt>i24</tt> writes at most
3618three bytes. When writing a value of a type like <tt>i20</tt> with a
3619size that is not an integral number of bytes, it is unspecified what
3620happens to the extra bits that do not belong to the type, but they will
3621typically be overwritten.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3622<h5>Example:</h5>
3623<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3624 store i32 3, i32* %ptr <i>; yields {void}</i>
3625 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3626</pre>
3627</div>
3628
3629<!-- _______________________________________________________________________ -->
3630<div class="doc_subsubsection">
3631 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3632</div>
3633
3634<div class="doc_text">
3635<h5>Syntax:</h5>
3636<pre>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3637 &lt;result&gt; = getelementptr &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3638</pre>
3639
3640<h5>Overview:</h5>
3641
3642<p>
3643The '<tt>getelementptr</tt>' instruction is used to get the address of a
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3644subelement of an aggregate data structure. It performs address calculation only
3645and does not access memory.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3646
3647<h5>Arguments:</h5>
3648
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3649<p>The first argument is always a pointer, and forms the basis of the
3650calculation. The remaining arguments are indices, that indicate which of the
3651elements of the aggregate object are indexed. The interpretation of each index
3652is dependent on the type being indexed into. The first index always indexes the
3653pointer value given as the first argument, the second index indexes a value of
3654the type pointed to (not necessarily the value directly pointed to, since the
3655first index can be non-zero), etc. The first type indexed into must be a pointer
3656value, subsequent types can be arrays, vectors and structs. Note that subsequent
3657types being indexed into can never be pointers, since that would require loading
3658the pointer before continuing calculation.</p>
3659
3660<p>The type of each index argument depends on the type it is indexing into.
3661When indexing into a (packed) structure, only <tt>i32</tt> integer
3662<b>constants</b> are allowed. When indexing into an array, pointer or vector,
Sanjiv Guptaa2e72d92009-04-27 03:21:00 +0000[diff] [blame]3663integers of any width are allowed (also non-constants).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3664
3665<p>For example, let's consider a C code fragment and how it gets
3666compiled to LLVM:</p>
3667
3668<div class="doc_code">
3669<pre>
3670struct RT {
3671 char A;
3672 int B[10][20];
3673 char C;
3674};
3675struct ST {
3676 int X;
3677 double Y;
3678 struct RT Z;
3679};
3680
3681int *foo(struct ST *s) {
3682 return &amp;s[1].Z.B[5][13];
3683}
3684</pre>
3685</div>
3686
3687<p>The LLVM code generated by the GCC frontend is:</p>
3688
3689<div class="doc_code">
3690<pre>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]3691%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3692%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3693
3694define i32* %foo(%ST* %s) {
3695entry:
3696 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3697 ret i32* %reg
3698}
3699</pre>
3700</div>
3701
3702<h5>Semantics:</h5>
3703
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3704<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3705type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3706}</tt>' type, a structure. The second index indexes into the third element of
3707the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3708i8 }</tt>' type, another structure. The third index indexes into the second
3709element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3710array. The two dimensions of the array are subscripted into, yielding an
3711'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3712to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3713
3714<p>Note that it is perfectly legal to index partially through a
3715structure, returning a pointer to an inner element. Because of this,
3716the LLVM code for the given testcase is equivalent to:</p>
3717
3718<pre>
3719 define i32* %foo(%ST* %s) {
3720 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3721 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3722 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3723 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3724 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3725 ret i32* %t5
3726 }
3727</pre>
3728
Chris Lattner50609942009-03-09 20:55:18 +0000[diff] [blame]3729<p>Note that it is undefined to access an array out of bounds: array
3730and pointer indexes must always be within the defined bounds of the
3731array type when accessed with an instruction that dereferences the
3732pointer (e.g. a load or store instruction). The one exception for
3733this rule is zero length arrays. These arrays are defined to be
3734accessible as variable length arrays, which requires access beyond the
3735zero'th element.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3736
3737<p>The getelementptr instruction is often confusing. For some more insight
3738into how it works, see <a href="GetElementPtr.html">the getelementptr
3739FAQ</a>.</p>
3740
3741<h5>Example:</h5>
3742
3743<pre>
3744 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3745 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3746 <i>; yields i8*:vptr</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3747 %vptr = getelementptr {i32, &lt;2 x i8&gt;}* %svptr, i64 0, i32 1, i32 1
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3748 <i>; yields i8*:eptr</i>
3749 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Sanjiv Gupta4f9a0dc2009-04-25 07:27:44 +0000[diff] [blame]3750 <i>; yields i32*:iptr</i>
Sanjiv Gupta1e46c582009-04-24 16:38:13 +0000[diff] [blame]3751 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3752</pre>
3753</div>
3754
3755<!-- ======================================================================= -->
3756<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3757</div>
3758<div class="doc_text">
3759<p>The instructions in this category are the conversion instructions (casting)
3760which all take a single operand and a type. They perform various bit conversions
3761on the operand.</p>
3762</div>
3763
3764<!-- _______________________________________________________________________ -->
3765<div class="doc_subsubsection">
3766 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3767</div>
3768<div class="doc_text">
3769
3770<h5>Syntax:</h5>
3771<pre>
3772 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3773</pre>
3774
3775<h5>Overview:</h5>
3776<p>
3777The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3778</p>
3779
3780<h5>Arguments:</h5>
3781<p>
3782The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3783be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3784and type of the result, which must be an <a href="#t_integer">integer</a>
3785type. The bit size of <tt>value</tt> must be larger than the bit size of
3786<tt>ty2</tt>. Equal sized types are not allowed.</p>
3787
3788<h5>Semantics:</h5>
3789<p>
3790The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3791and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3792larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3793It will always truncate bits.</p>
3794
3795<h5>Example:</h5>
3796<pre>
3797 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3798 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3799 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3800</pre>
3801</div>
3802
3803<!-- _______________________________________________________________________ -->
3804<div class="doc_subsubsection">
3805 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3806</div>
3807<div class="doc_text">
3808
3809<h5>Syntax:</h5>
3810<pre>
3811 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3812</pre>
3813
3814<h5>Overview:</h5>
3815<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3816<tt>ty2</tt>.</p>
3817
3818
3819<h5>Arguments:</h5>
3820<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3821<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3822also be of <a href="#t_integer">integer</a> type. The bit size of the
3823<tt>value</tt> must be smaller than the bit size of the destination type,
3824<tt>ty2</tt>.</p>
3825
3826<h5>Semantics:</h5>
3827<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3828bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3829
3830<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3831
3832<h5>Example:</h5>
3833<pre>
3834 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3835 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3836</pre>
3837</div>
3838
3839<!-- _______________________________________________________________________ -->
3840<div class="doc_subsubsection">
3841 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3842</div>
3843<div class="doc_text">
3844
3845<h5>Syntax:</h5>
3846<pre>
3847 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3848</pre>
3849
3850<h5>Overview:</h5>
3851<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3852
3853<h5>Arguments:</h5>
3854<p>
3855The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3856<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3857also be of <a href="#t_integer">integer</a> type. The bit size of the
3858<tt>value</tt> must be smaller than the bit size of the destination type,
3859<tt>ty2</tt>.</p>
3860
3861<h5>Semantics:</h5>
3862<p>
3863The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3864bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3865the type <tt>ty2</tt>.</p>
3866
3867<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3868
3869<h5>Example:</h5>
3870<pre>
3871 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3872 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3873</pre>
3874</div>
3875
3876<!-- _______________________________________________________________________ -->
3877<div class="doc_subsubsection">
3878 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3879</div>
3880
3881<div class="doc_text">
3882
3883<h5>Syntax:</h5>
3884
3885<pre>
3886 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3887</pre>
3888
3889<h5>Overview:</h5>
3890<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3891<tt>ty2</tt>.</p>
3892
3893
3894<h5>Arguments:</h5>
3895<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3896 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3897cast it to. The size of <tt>value</tt> must be larger than the size of
3898<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3899<i>no-op cast</i>.</p>
3900
3901<h5>Semantics:</h5>
3902<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3903<a href="#t_floating">floating point</a> type to a smaller
3904<a href="#t_floating">floating point</a> type. If the value cannot fit within
3905the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3906
3907<h5>Example:</h5>
3908<pre>
3909 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3910 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3911</pre>
3912</div>
3913
3914<!-- _______________________________________________________________________ -->
3915<div class="doc_subsubsection">
3916 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3917</div>
3918<div class="doc_text">
3919
3920<h5>Syntax:</h5>
3921<pre>
3922 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3923</pre>
3924
3925<h5>Overview:</h5>
3926<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3927floating point value.</p>
3928
3929<h5>Arguments:</h5>
3930<p>The '<tt>fpext</tt>' instruction takes a
3931<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3932and a <a href="#t_floating">floating point</a> type to cast it to. The source
3933type must be smaller than the destination type.</p>
3934
3935<h5>Semantics:</h5>
3936<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3937<a href="#t_floating">floating point</a> type to a larger
3938<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3939used to make a <i>no-op cast</i> because it always changes bits. Use
3940<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3941
3942<h5>Example:</h5>
3943<pre>
3944 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3945 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3946</pre>
3947</div>
3948
3949<!-- _______________________________________________________________________ -->
3950<div class="doc_subsubsection">
3951 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3952</div>
3953<div class="doc_text">
3954
3955<h5>Syntax:</h5>
3956<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3957 &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 +0000[diff] [blame]3958</pre>
3959
3960<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3961<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3962unsigned integer equivalent of type <tt>ty2</tt>.
3963</p>
3964
3965<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3966<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3967scalar or vector <a href="#t_floating">floating point</a> value, and a type
3968to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3969type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3970vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3971
3972<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3973<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3974<a href="#t_floating">floating point</a> operand into the nearest (rounding
3975towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3976the results are undefined.</p>
3977
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3978<h5>Example:</h5>
3979<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3980 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3981 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3982 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3983</pre>
3984</div>
3985
3986<!-- _______________________________________________________________________ -->
3987<div class="doc_subsubsection">
3988 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3989</div>
3990<div class="doc_text">
3991
3992<h5>Syntax:</h5>
3993<pre>
3994 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3995</pre>
3996
3997<h5>Overview:</h5>
3998<p>The '<tt>fptosi</tt>' instruction converts
3999<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
4000</p>
4001
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4002<h5>Arguments:</h5>
4003<p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]4004scalar or vector <a href="#t_floating">floating point</a> value, and a type
4005to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4006type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4007vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4008
4009<h5>Semantics:</h5>
4010<p>The '<tt>fptosi</tt>' instruction converts its
4011<a href="#t_floating">floating point</a> operand into the nearest (rounding
4012towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4013the results are undefined.</p>
4014
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4015<h5>Example:</h5>
4016<pre>
4017 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]4018 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4019 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4020</pre>
4021</div>
4022
4023<!-- _______________________________________________________________________ -->
4024<div class="doc_subsubsection">
4025 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4026</div>
4027<div class="doc_text">
4028
4029<h5>Syntax:</h5>
4030<pre>
4031 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4032</pre>
4033
4034<h5>Overview:</h5>
4035<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4036integer and converts that value to the <tt>ty2</tt> type.</p>
4037
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4038<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]4039<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4040scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4041to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4042type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4043floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4044
4045<h5>Semantics:</h5>
4046<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4047integer quantity and converts it to the corresponding floating point value. If
4048the value cannot fit in the floating point value, the results are undefined.</p>
4049
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4050<h5>Example:</h5>
4051<pre>
4052 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4053 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4054</pre>
4055</div>
4056
4057<!-- _______________________________________________________________________ -->
4058<div class="doc_subsubsection">
4059 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4060</div>
4061<div class="doc_text">
4062
4063<h5>Syntax:</h5>
4064<pre>
4065 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4066</pre>
4067
4068<h5>Overview:</h5>
4069<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
4070integer and converts that value to the <tt>ty2</tt> type.</p>
4071
4072<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]4073<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4074scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4075to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4076type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4077floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4078
4079<h5>Semantics:</h5>
4080<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
4081integer quantity and converts it to the corresponding floating point value. If
4082the value cannot fit in the floating point value, the results are undefined.</p>
4083
4084<h5>Example:</h5>
4085<pre>
4086 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4087 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4088</pre>
4089</div>
4090
4091<!-- _______________________________________________________________________ -->
4092<div class="doc_subsubsection">
4093 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4094</div>
4095<div class="doc_text">
4096
4097<h5>Syntax:</h5>
4098<pre>
4099 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4100</pre>
4101
4102<h5>Overview:</h5>
4103<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4104the integer type <tt>ty2</tt>.</p>
4105
4106<h5>Arguments:</h5>
4107<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4108must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4109<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4110
4111<h5>Semantics:</h5>
4112<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4113<tt>ty2</tt> by interpreting the pointer value as an integer and either
4114truncating or zero extending that value to the size of the integer type. If
4115<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4116<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4117are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4118change.</p>
4119
4120<h5>Example:</h5>
4121<pre>
4122 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4123 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4124</pre>
4125</div>
4126
4127<!-- _______________________________________________________________________ -->
4128<div class="doc_subsubsection">
4129 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4130</div>
4131<div class="doc_text">
4132
4133<h5>Syntax:</h5>
4134<pre>
4135 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4136</pre>
4137
4138<h5>Overview:</h5>
4139<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4140a pointer type, <tt>ty2</tt>.</p>
4141
4142<h5>Arguments:</h5>
4143<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4144value to cast, and a type to cast it to, which must be a
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4145<a href="#t_pointer">pointer</a> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4146
4147<h5>Semantics:</h5>
4148<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4149<tt>ty2</tt> by applying either a zero extension or a truncation depending on
4150the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4151size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4152the size of a pointer then a zero extension is done. If they are the same size,
4153nothing is done (<i>no-op cast</i>).</p>
4154
4155<h5>Example:</h5>
4156<pre>
4157 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4158 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4159 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4160</pre>
4161</div>
4162
4163<!-- _______________________________________________________________________ -->
4164<div class="doc_subsubsection">
4165 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4166</div>
4167<div class="doc_text">
4168
4169<h5>Syntax:</h5>
4170<pre>
4171 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4172</pre>
4173
4174<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4175
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4176<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4177<tt>ty2</tt> without changing any bits.</p>
4178
4179<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4180
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4181<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]4182a non-aggregate first class value, and a type to cast it to, which must also be
4183a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4184<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4185and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4186type is a pointer, the destination type must also be a pointer. This
4187instruction supports bitwise conversion of vectors to integers and to vectors
4188of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4189
4190<h5>Semantics:</h5>
4191<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4192<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4193this conversion. The conversion is done as if the <tt>value</tt> had been
4194stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4195converted to other pointer types with this instruction. To convert pointers to
4196other types, use the <a href="#i_inttoptr">inttoptr</a> or
4197<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4198
4199<h5>Example:</h5>
4200<pre>
4201 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4202 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4203 %Z = bitcast &lt;2 x int&gt; %V to i64; <i>; yields i64: %V</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4204</pre>
4205</div>
4206
4207<!-- ======================================================================= -->
4208<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4209<div class="doc_text">
4210<p>The instructions in this category are the "miscellaneous"
4211instructions, which defy better classification.</p>
4212</div>
4213
4214<!-- _______________________________________________________________________ -->
4215<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4216</div>
4217<div class="doc_text">
4218<h5>Syntax:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4219<pre> &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 +0000[diff] [blame]4220</pre>
4221<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4222<p>The '<tt>icmp</tt>' instruction returns a boolean value or
4223a vector of boolean values based on comparison
4224of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4225<h5>Arguments:</h5>
4226<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4227the condition code indicating the kind of comparison to perform. It is not
4228a value, just a keyword. The possible condition code are:
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4229</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4230<ol>
4231 <li><tt>eq</tt>: equal</li>
4232 <li><tt>ne</tt>: not equal </li>
4233 <li><tt>ugt</tt>: unsigned greater than</li>
4234 <li><tt>uge</tt>: unsigned greater or equal</li>
4235 <li><tt>ult</tt>: unsigned less than</li>
4236 <li><tt>ule</tt>: unsigned less or equal</li>
4237 <li><tt>sgt</tt>: signed greater than</li>
4238 <li><tt>sge</tt>: signed greater or equal</li>
4239 <li><tt>slt</tt>: signed less than</li>
4240 <li><tt>sle</tt>: signed less or equal</li>
4241</ol>
4242<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4243<a href="#t_pointer">pointer</a>
4244or integer <a href="#t_vector">vector</a> typed.
4245They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4246<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4247<p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4248the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4249yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4250</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4251<ol>
4252 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4253 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4254 </li>
4255 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4256 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4257 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4258 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4259 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4260 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4261 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4262 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4263 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4264 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4265 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4266 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4267 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4268 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4269 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4270 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4271 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4272 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4273</ol>
4274<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4275values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4276<p>If the operands are integer vectors, then they are compared
4277element by element. The result is an <tt>i1</tt> vector with
4278the same number of elements as the values being compared.
4279Otherwise, the result is an <tt>i1</tt>.
4280</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4281
4282<h5>Example:</h5>
4283<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
4284 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
4285 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
4286 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4287 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4288 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
4289</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4290
4291<p>Note that the code generator does not yet support vector types with
4292 the <tt>icmp</tt> instruction.</p>
4293
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4294</div>
4295
4296<!-- _______________________________________________________________________ -->
4297<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4298</div>
4299<div class="doc_text">
4300<h5>Syntax:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4301<pre> &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 +0000[diff] [blame]4302</pre>
4303<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4304<p>The '<tt>fcmp</tt>' instruction returns a boolean value
4305or vector of boolean values based on comparison
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4306of its operands.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4307<p>
4308If the operands are floating point scalars, then the result
4309type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4310</p>
4311<p>If the operands are floating point vectors, then the result type
4312is a vector of boolean with the same number of elements as the
4313operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4314<h5>Arguments:</h5>
4315<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4316the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4317a value, just a keyword. The possible condition code are:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4318<ol>
4319 <li><tt>false</tt>: no comparison, always returns false</li>
4320 <li><tt>oeq</tt>: ordered and equal</li>
4321 <li><tt>ogt</tt>: ordered and greater than </li>
4322 <li><tt>oge</tt>: ordered and greater than or equal</li>
4323 <li><tt>olt</tt>: ordered and less than </li>
4324 <li><tt>ole</tt>: ordered and less than or equal</li>
4325 <li><tt>one</tt>: ordered and not equal</li>
4326 <li><tt>ord</tt>: ordered (no nans)</li>
4327 <li><tt>ueq</tt>: unordered or equal</li>
4328 <li><tt>ugt</tt>: unordered or greater than </li>
4329 <li><tt>uge</tt>: unordered or greater than or equal</li>
4330 <li><tt>ult</tt>: unordered or less than </li>
4331 <li><tt>ule</tt>: unordered or less than or equal</li>
4332 <li><tt>une</tt>: unordered or not equal</li>
4333 <li><tt>uno</tt>: unordered (either nans)</li>
4334 <li><tt>true</tt>: no comparison, always returns true</li>
4335</ol>
4336<p><i>Ordered</i> means that neither operand is a QNAN while
4337<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4338<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4339either a <a href="#t_floating">floating point</a> type
4340or a <a href="#t_vector">vector</a> of floating point type.
4341They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4342<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4343<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4344according to the condition code given as <tt>cond</tt>.
4345If the operands are vectors, then the vectors are compared
4346element by element.
4347Each comparison performed
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4348always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4349<ol>
4350 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4351 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4352 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4353 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4354 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4355 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4356 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4357 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4358 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4359 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4360 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4361 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4362 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4363 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4364 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4365 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4366 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4367 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4368 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4369 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4370 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4371 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4372 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4373 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4374 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4375 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4376 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4377 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4378</ol>
4379
4380<h5>Example:</h5>
4381<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4382 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4383 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4384 &lt;result&gt; = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4385</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4386
4387<p>Note that the code generator does not yet support vector types with
4388 the <tt>fcmp</tt> instruction.</p>
4389
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4390</div>
4391
4392<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4393<div class="doc_subsubsection">
4394 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4395</div>
4396<div class="doc_text">
4397<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4398<pre> &lt;result&gt; = vicmp &lt;cond&gt; &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4399</pre>
4400<h5>Overview:</h5>
4401<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4402element-wise comparison of its two integer vector operands.</p>
4403<h5>Arguments:</h5>
4404<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4405the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4406a value, just a keyword. The possible condition code are:</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4407<ol>
4408 <li><tt>eq</tt>: equal</li>
4409 <li><tt>ne</tt>: not equal </li>
4410 <li><tt>ugt</tt>: unsigned greater than</li>
4411 <li><tt>uge</tt>: unsigned greater or equal</li>
4412 <li><tt>ult</tt>: unsigned less than</li>
4413 <li><tt>ule</tt>: unsigned less or equal</li>
4414 <li><tt>sgt</tt>: signed greater than</li>
4415 <li><tt>sge</tt>: signed greater or equal</li>
4416 <li><tt>slt</tt>: signed less than</li>
4417 <li><tt>sle</tt>: signed less or equal</li>
4418</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4419<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4420<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4421<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4422<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4423according to the condition code given as <tt>cond</tt>. The comparison yields a
4424<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4425identical type as the values being compared. The most significant bit in each
4426element is 1 if the element-wise comparison evaluates to true, and is 0
4427otherwise. All other bits of the result are undefined. The condition codes
4428are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4429instruction</a>.</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4430
4431<h5>Example:</h5>
4432<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4433 &lt;result&gt; = vicmp eq &lt;2 x i32&gt; &lt; i32 4, i32 0&gt;, &lt; i32 5, i32 0&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4434 &lt;result&gt; = vicmp ult &lt;2 x i8 &gt; &lt; i8 1, i8 2&gt;, &lt; i8 2, i8 2 &gt; <i>; yields: result=&lt;2 x i8&gt; &lt; i8 -1, i8 0 &gt;</i>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4435</pre>
4436</div>
4437
4438<!-- _______________________________________________________________________ -->
4439<div class="doc_subsubsection">
4440 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4441</div>
4442<div class="doc_text">
4443<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4444<pre> &lt;result&gt; = vfcmp &lt;cond&gt; &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt;</pre>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4445<h5>Overview:</h5>
4446<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4447element-wise comparison of its two floating point vector operands. The output
4448elements have the same width as the input elements.</p>
4449<h5>Arguments:</h5>
4450<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4451the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4452a value, just a keyword. The possible condition code are:</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4453<ol>
4454 <li><tt>false</tt>: no comparison, always returns false</li>
4455 <li><tt>oeq</tt>: ordered and equal</li>
4456 <li><tt>ogt</tt>: ordered and greater than </li>
4457 <li><tt>oge</tt>: ordered and greater than or equal</li>
4458 <li><tt>olt</tt>: ordered and less than </li>
4459 <li><tt>ole</tt>: ordered and less than or equal</li>
4460 <li><tt>one</tt>: ordered and not equal</li>
4461 <li><tt>ord</tt>: ordered (no nans)</li>
4462 <li><tt>ueq</tt>: unordered or equal</li>
4463 <li><tt>ugt</tt>: unordered or greater than </li>
4464 <li><tt>uge</tt>: unordered or greater than or equal</li>
4465 <li><tt>ult</tt>: unordered or less than </li>
4466 <li><tt>ule</tt>: unordered or less than or equal</li>
4467 <li><tt>une</tt>: unordered or not equal</li>
4468 <li><tt>uno</tt>: unordered (either nans)</li>
4469 <li><tt>true</tt>: no comparison, always returns true</li>
4470</ol>
4471<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4472<a href="#t_floating">floating point</a> typed. They must also be identical
4473types.</p>
4474<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4475<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4476according to the condition code given as <tt>cond</tt>. The comparison yields a
4477<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4478an identical number of elements as the values being compared, and each element
4479having identical with to the width of the floating point elements. The most
4480significant bit in each element is 1 if the element-wise comparison evaluates to
4481true, and is 0 otherwise. All other bits of the result are undefined. The
4482condition codes are evaluated identically to the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4483<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4484
4485<h5>Example:</h5>
4486<pre>
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000[diff] [blame]4487 <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4488 &lt;result&gt; = vfcmp oeq &lt;2 x float&gt; &lt; float 4, float 0 &gt;, &lt; float 5, float 0 &gt;
4489
4490 <i>; yields: result=&lt;2 x i64&gt; &lt; i64 -1, i64 0 &gt;</i>
4491 &lt;result&gt; = vfcmp ult &lt;2 x double&gt; &lt; double 1, double 2 &gt;, &lt; double 2, double 2&gt;
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4492</pre>
4493</div>
4494
4495<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4496<div class="doc_subsubsection">
4497 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4498</div>
4499
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4500<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4501
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4502<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4503
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4504<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4505<h5>Overview:</h5>
4506<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4507the SSA graph representing the function.</p>
4508<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4509
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4510<p>The type of the incoming values is specified with the first type
4511field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4512as arguments, with one pair for each predecessor basic block of the
4513current block. Only values of <a href="#t_firstclass">first class</a>
4514type may be used as the value arguments to the PHI node. Only labels
4515may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4516
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4517<p>There must be no non-phi instructions between the start of a basic
4518block and the PHI instructions: i.e. PHI instructions must be first in
4519a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4520
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4521<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4522
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4523<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4524specified by the pair corresponding to the predecessor basic block that executed
4525just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4526
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4527<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4528<pre>
4529Loop: ; Infinite loop that counts from 0 on up...
4530 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4531 %nextindvar = add i32 %indvar, 1
4532 br label %Loop
4533</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4534</div>
4535
4536<!-- _______________________________________________________________________ -->
4537<div class="doc_subsubsection">
4538 <a name="i_select">'<tt>select</tt>' Instruction</a>
4539</div>
4540
4541<div class="doc_text">
4542
4543<h5>Syntax:</h5>
4544
4545<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4546 &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>
4547
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4548 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4549</pre>
4550
4551<h5>Overview:</h5>
4552
4553<p>
4554The '<tt>select</tt>' instruction is used to choose one value based on a
4555condition, without branching.
4556</p>
4557
4558
4559<h5>Arguments:</h5>
4560
4561<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4562The '<tt>select</tt>' instruction requires an 'i1' value or
4563a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4564condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4565type. If the val1/val2 are vectors and
4566the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4567individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4568</p>
4569
4570<h5>Semantics:</h5>
4571
4572<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4573If the condition is an i1 and it evaluates to 1, the instruction returns the first
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4574value argument; otherwise, it returns the second value argument.
4575</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4576<p>
4577If the condition is a vector of i1, then the value arguments must
4578be vectors of the same size, and the selection is done element
4579by element.
4580</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4581
4582<h5>Example:</h5>
4583
4584<pre>
4585 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4586</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4587
4588<p>Note that the code generator does not yet support conditions
4589 with vector type.</p>
4590
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4591</div>
4592
4593
4594<!-- _______________________________________________________________________ -->
4595<div class="doc_subsubsection">
4596 <a name="i_call">'<tt>call</tt>' Instruction</a>
4597</div>
4598
4599<div class="doc_text">
4600
4601<h5>Syntax:</h5>
4602<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4603 &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 +0000[diff] [blame]4604</pre>
4605
4606<h5>Overview:</h5>
4607
4608<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4609
4610<h5>Arguments:</h5>
4611
4612<p>This instruction requires several arguments:</p>
4613
4614<ol>
4615 <li>
4616 <p>The optional "tail" marker indicates whether the callee function accesses
4617 any allocas or varargs in the caller. If the "tail" marker is present, the
4618 function call is eligible for tail call optimization. Note that calls may
4619 be marked "tail" even if they do not occur before a <a
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4620 href="#i_ret"><tt>ret</tt></a> instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4621 </li>
4622 <li>
4623 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4624 convention</a> the call should use. If none is specified, the call defaults
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4625 to using C calling conventions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4626 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4627
4628 <li>
4629 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4630 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4631 and '<tt>inreg</tt>' attributes are valid here.</p>
4632 </li>
4633
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4634 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4635 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4636 the type of the return value. Functions that return no value are marked
4637 <tt><a href="#t_void">void</a></tt>.</p>
4638 </li>
4639 <li>
4640 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4641 value being invoked. The argument types must match the types implied by
4642 this signature. This type can be omitted if the function is not varargs
4643 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4644 </li>
4645 <li>
4646 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4647 be invoked. In most cases, this is a direct function invocation, but
4648 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4649 to function value.</p>
4650 </li>
4651 <li>
4652 <p>'<tt>function args</tt>': argument list whose types match the
4653 function signature argument types. All arguments must be of
4654 <a href="#t_firstclass">first class</a> type. If the function signature
4655 indicates the function accepts a variable number of arguments, the extra
4656 arguments can be specified.</p>
4657 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4658 <li>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4659 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4660 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4661 '<tt>readnone</tt>' attributes are valid here.</p>
4662 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4663</ol>
4664
4665<h5>Semantics:</h5>
4666
4667<p>The '<tt>call</tt>' instruction is used to cause control flow to
4668transfer to a specified function, with its incoming arguments bound to
4669the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4670instruction in the called function, control flow continues with the
4671instruction after the function call, and the return value of the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4672function is bound to the result argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4673
4674<h5>Example:</h5>
4675
4676<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4677 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4678 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4679 %X = tail call i32 @foo() <i>; yields i32</i>
4680 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4681 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4682
4683 %struct.A = type { i32, i8 }
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4684 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4685 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4686 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattnerac454b32008-10-08 06:26:11 +0000[diff] [blame]4687 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijman2c4e05a2008-10-07 10:03:45 +0000[diff] [blame]4688 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4689</pre>
4690
4691</div>
4692
4693<!-- _______________________________________________________________________ -->
4694<div class="doc_subsubsection">
4695 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4696</div>
4697
4698<div class="doc_text">
4699
4700<h5>Syntax:</h5>
4701
4702<pre>
4703 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4704</pre>
4705
4706<h5>Overview:</h5>
4707
4708<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4709the "variable argument" area of a function call. It is used to implement the
4710<tt>va_arg</tt> macro in C.</p>
4711
4712<h5>Arguments:</h5>
4713
4714<p>This instruction takes a <tt>va_list*</tt> value and the type of
4715the argument. It returns a value of the specified argument type and
4716increments the <tt>va_list</tt> to point to the next argument. The
4717actual type of <tt>va_list</tt> is target specific.</p>
4718
4719<h5>Semantics:</h5>
4720
4721<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4722type from the specified <tt>va_list</tt> and causes the
4723<tt>va_list</tt> to point to the next argument. For more information,
4724see the variable argument handling <a href="#int_varargs">Intrinsic
4725Functions</a>.</p>
4726
4727<p>It is legal for this instruction to be called in a function which does not
4728take a variable number of arguments, for example, the <tt>vfprintf</tt>
4729function.</p>
4730
4731<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4732href="#intrinsics">intrinsic function</a> because it takes a type as an
4733argument.</p>
4734
4735<h5>Example:</h5>
4736
4737<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4738
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]4739<p>Note that the code generator does not yet fully support va_arg
4740 on many targets. Also, it does not currently support va_arg with
4741 aggregate types on any target.</p>
4742
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4743</div>
4744
4745<!-- *********************************************************************** -->
4746<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4747<!-- *********************************************************************** -->
4748
4749<div class="doc_text">
4750
4751<p>LLVM supports the notion of an "intrinsic function". These functions have
4752well known names and semantics and are required to follow certain restrictions.
4753Overall, these intrinsics represent an extension mechanism for the LLVM
4754language that does not require changing all of the transformations in LLVM when
4755adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4756
4757<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4758prefix is reserved in LLVM for intrinsic names; thus, function names may not
4759begin with this prefix. Intrinsic functions must always be external functions:
4760you cannot define the body of intrinsic functions. Intrinsic functions may
4761only be used in call or invoke instructions: it is illegal to take the address
4762of an intrinsic function. Additionally, because intrinsic functions are part
4763of the LLVM language, it is required if any are added that they be documented
4764here.</p>
4765
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4766<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4767a family of functions that perform the same operation but on different data
4768types. Because LLVM can represent over 8 million different integer types,
4769overloading is used commonly to allow an intrinsic function to operate on any
4770integer type. One or more of the argument types or the result type can be
4771overloaded to accept any integer type. Argument types may also be defined as
4772exactly matching a previous argument's type or the result type. This allows an
4773intrinsic function which accepts multiple arguments, but needs all of them to
4774be of the same type, to only be overloaded with respect to a single argument or
4775the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4776
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4777<p>Overloaded intrinsics will have the names of its overloaded argument types
4778encoded into its function name, each preceded by a period. Only those types
4779which are overloaded result in a name suffix. Arguments whose type is matched
4780against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4781take an integer of any width and returns an integer of exactly the same integer
4782width. This leads to a family of functions such as
4783<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4784Only one type, the return type, is overloaded, and only one type suffix is
4785required. Because the argument's type is matched against the return type, it
4786does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4787
4788<p>To learn how to add an intrinsic function, please see the
4789<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4790</p>
4791
4792</div>
4793
4794<!-- ======================================================================= -->
4795<div class="doc_subsection">
4796 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4797</div>
4798
4799<div class="doc_text">
4800
4801<p>Variable argument support is defined in LLVM with the <a
4802 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4803intrinsic functions. These functions are related to the similarly
4804named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4805
4806<p>All of these functions operate on arguments that use a
4807target-specific value type "<tt>va_list</tt>". The LLVM assembly
4808language reference manual does not define what this type is, so all
4809transformations should be prepared to handle these functions regardless of
4810the type used.</p>
4811
4812<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4813instruction and the variable argument handling intrinsic functions are
4814used.</p>
4815
4816<div class="doc_code">
4817<pre>
4818define i32 @test(i32 %X, ...) {
4819 ; Initialize variable argument processing
4820 %ap = alloca i8*
4821 %ap2 = bitcast i8** %ap to i8*
4822 call void @llvm.va_start(i8* %ap2)
4823
4824 ; Read a single integer argument
4825 %tmp = va_arg i8** %ap, i32
4826
4827 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4828 %aq = alloca i8*
4829 %aq2 = bitcast i8** %aq to i8*
4830 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4831 call void @llvm.va_end(i8* %aq2)
4832
4833 ; Stop processing of arguments.
4834 call void @llvm.va_end(i8* %ap2)
4835 ret i32 %tmp
4836}
4837
4838declare void @llvm.va_start(i8*)
4839declare void @llvm.va_copy(i8*, i8*)
4840declare void @llvm.va_end(i8*)
4841</pre>
4842</div>
4843
4844</div>
4845
4846<!-- _______________________________________________________________________ -->
4847<div class="doc_subsubsection">
4848 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4849</div>
4850
4851
4852<div class="doc_text">
4853<h5>Syntax:</h5>
4854<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4855<h5>Overview:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4856<p>The '<tt>llvm.va_start</tt>' intrinsic initializes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4857<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4858href="#i_va_arg">va_arg</a></tt>.</p>
4859
4860<h5>Arguments:</h5>
4861
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4862<p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4863
4864<h5>Semantics:</h5>
4865
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4866<p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4867macro available in C. In a target-dependent way, it initializes the
4868<tt>va_list</tt> element to which the argument points, so that the next call to
4869<tt>va_arg</tt> will produce the first variable argument passed to the function.
4870Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4871last argument of the function as the compiler can figure that out.</p>
4872
4873</div>
4874
4875<!-- _______________________________________________________________________ -->
4876<div class="doc_subsubsection">
4877 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4878</div>
4879
4880<div class="doc_text">
4881<h5>Syntax:</h5>
4882<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4883<h5>Overview:</h5>
4884
4885<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4886which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4887or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4888
4889<h5>Arguments:</h5>
4890
4891<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4892
4893<h5>Semantics:</h5>
4894
4895<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4896macro available in C. In a target-dependent way, it destroys the
4897<tt>va_list</tt> element to which the argument points. Calls to <a
4898href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4899<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4900<tt>llvm.va_end</tt>.</p>
4901
4902</div>
4903
4904<!-- _______________________________________________________________________ -->
4905<div class="doc_subsubsection">
4906 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4907</div>
4908
4909<div class="doc_text">
4910
4911<h5>Syntax:</h5>
4912
4913<pre>
4914 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4915</pre>
4916
4917<h5>Overview:</h5>
4918
4919<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4920from the source argument list to the destination argument list.</p>
4921
4922<h5>Arguments:</h5>
4923
4924<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4925The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4926
4927
4928<h5>Semantics:</h5>
4929
4930<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4931macro available in C. In a target-dependent way, it copies the source
4932<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4933intrinsic is necessary because the <tt><a href="#int_va_start">
4934llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4935example, memory allocation.</p>
4936
4937</div>
4938
4939<!-- ======================================================================= -->
4940<div class="doc_subsection">
4941 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4942</div>
4943
4944<div class="doc_text">
4945
4946<p>
4947LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4948Collection</a> (GC) requires the implementation and generation of these
4949intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4950These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4951stack</a>, as well as garbage collector implementations that require <a
4952href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4953Front-ends for type-safe garbage collected languages should generate these
4954intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4955href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4956</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4957
4958<p>The garbage collection intrinsics only operate on objects in the generic
4959 address space (address space zero).</p>
4960
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4961</div>
4962
4963<!-- _______________________________________________________________________ -->
4964<div class="doc_subsubsection">
4965 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4966</div>
4967
4968<div class="doc_text">
4969
4970<h5>Syntax:</h5>
4971
4972<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4973 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4974</pre>
4975
4976<h5>Overview:</h5>
4977
4978<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4979the code generator, and allows some metadata to be associated with it.</p>
4980
4981<h5>Arguments:</h5>
4982
4983<p>The first argument specifies the address of a stack object that contains the
4984root pointer. The second pointer (which must be either a constant or a global
4985value address) contains the meta-data to be associated with the root.</p>
4986
4987<h5>Semantics:</h5>
4988
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4989<p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4990location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4991the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4992intrinsic may only be used in a function which <a href="#gc">specifies a GC
4993algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4994
4995</div>
4996
4997
4998<!-- _______________________________________________________________________ -->
4999<div class="doc_subsubsection">
5000 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5001</div>
5002
5003<div class="doc_text">
5004
5005<h5>Syntax:</h5>
5006
5007<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5008 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5009</pre>
5010
5011<h5>Overview:</h5>
5012
5013<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5014locations, allowing garbage collector implementations that require read
5015barriers.</p>
5016
5017<h5>Arguments:</h5>
5018
5019<p>The second argument is the address to read from, which should be an address
5020allocated from the garbage collector. The first object is a pointer to the
5021start of the referenced object, if needed by the language runtime (otherwise
5022null).</p>
5023
5024<h5>Semantics:</h5>
5025
5026<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5027instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]5028garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5029may only be used in a function which <a href="#gc">specifies a GC
5030algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5031
5032</div>
5033
5034
5035<!-- _______________________________________________________________________ -->
5036<div class="doc_subsubsection">
5037 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5038</div>
5039
5040<div class="doc_text">
5041
5042<h5>Syntax:</h5>
5043
5044<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5045 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5046</pre>
5047
5048<h5>Overview:</h5>
5049
5050<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5051locations, allowing garbage collector implementations that require write
5052barriers (such as generational or reference counting collectors).</p>
5053
5054<h5>Arguments:</h5>
5055
5056<p>The first argument is the reference to store, the second is the start of the
5057object to store it to, and the third is the address of the field of Obj to
5058store to. If the runtime does not require a pointer to the object, Obj may be
5059null.</p>
5060
5061<h5>Semantics:</h5>
5062
5063<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5064instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]5065garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5066may only be used in a function which <a href="#gc">specifies a GC
5067algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5068
5069</div>
5070
5071
5072
5073<!-- ======================================================================= -->
5074<div class="doc_subsection">
5075 <a name="int_codegen">Code Generator Intrinsics</a>
5076</div>
5077
5078<div class="doc_text">
5079<p>
5080These intrinsics are provided by LLVM to expose special features that may only
5081be implemented with code generator support.
5082</p>
5083
5084</div>
5085
5086<!-- _______________________________________________________________________ -->
5087<div class="doc_subsubsection">
5088 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5089</div>
5090
5091<div class="doc_text">
5092
5093<h5>Syntax:</h5>
5094<pre>
5095 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
5096</pre>
5097
5098<h5>Overview:</h5>
5099
5100<p>
5101The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5102target-specific value indicating the return address of the current function
5103or one of its callers.
5104</p>
5105
5106<h5>Arguments:</h5>
5107
5108<p>
5109The argument to this intrinsic indicates which function to return the address
5110for. Zero indicates the calling function, one indicates its caller, etc. The
5111argument is <b>required</b> to be a constant integer value.
5112</p>
5113
5114<h5>Semantics:</h5>
5115
5116<p>
5117The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
5118the return address of the specified call frame, or zero if it cannot be
5119identified. The value returned by this intrinsic is likely to be incorrect or 0
5120for arguments other than zero, so it should only be used for debugging purposes.
5121</p>
5122
5123<p>
5124Note that calling this intrinsic does not prevent function inlining or other
5125aggressive transformations, so the value returned may not be that of the obvious
5126source-language caller.
5127</p>
5128</div>
5129
5130
5131<!-- _______________________________________________________________________ -->
5132<div class="doc_subsubsection">
5133 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5134</div>
5135
5136<div class="doc_text">
5137
5138<h5>Syntax:</h5>
5139<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5140 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5141</pre>
5142
5143<h5>Overview:</h5>
5144
5145<p>
5146The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5147target-specific frame pointer value for the specified stack frame.
5148</p>
5149
5150<h5>Arguments:</h5>
5151
5152<p>
5153The argument to this intrinsic indicates which function to return the frame
5154pointer for. Zero indicates the calling function, one indicates its caller,
5155etc. The argument is <b>required</b> to be a constant integer value.
5156</p>
5157
5158<h5>Semantics:</h5>
5159
5160<p>
5161The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5162the frame address of the specified call frame, or zero if it cannot be
5163identified. The value returned by this intrinsic is likely to be incorrect or 0
5164for arguments other than zero, so it should only be used for debugging purposes.
5165</p>
5166
5167<p>
5168Note that calling this intrinsic does not prevent function inlining or other
5169aggressive transformations, so the value returned may not be that of the obvious
5170source-language caller.
5171</p>
5172</div>
5173
5174<!-- _______________________________________________________________________ -->
5175<div class="doc_subsubsection">
5176 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5177</div>
5178
5179<div class="doc_text">
5180
5181<h5>Syntax:</h5>
5182<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5183 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5184</pre>
5185
5186<h5>Overview:</h5>
5187
5188<p>
5189The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
5190the function stack, for use with <a href="#int_stackrestore">
5191<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5192features like scoped automatic variable sized arrays in C99.
5193</p>
5194
5195<h5>Semantics:</h5>
5196
5197<p>
5198This intrinsic returns a opaque pointer value that can be passed to <a
5199href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
5200<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5201<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5202state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5203practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5204that were allocated after the <tt>llvm.stacksave</tt> was executed.
5205</p>
5206
5207</div>
5208
5209<!-- _______________________________________________________________________ -->
5210<div class="doc_subsubsection">
5211 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5212</div>
5213
5214<div class="doc_text">
5215
5216<h5>Syntax:</h5>
5217<pre>
5218 declare void @llvm.stackrestore(i8 * %ptr)
5219</pre>
5220
5221<h5>Overview:</h5>
5222
5223<p>
5224The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5225the function stack to the state it was in when the corresponding <a
5226href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
5227useful for implementing language features like scoped automatic variable sized
5228arrays in C99.
5229</p>
5230
5231<h5>Semantics:</h5>
5232
5233<p>
5234See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
5235</p>
5236
5237</div>
5238
5239
5240<!-- _______________________________________________________________________ -->
5241<div class="doc_subsubsection">
5242 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5243</div>
5244
5245<div class="doc_text">
5246
5247<h5>Syntax:</h5>
5248<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5249 declare void @llvm.prefetch(i8* &lt;address&gt;, i32 &lt;rw&gt;, i32 &lt;locality&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5250</pre>
5251
5252<h5>Overview:</h5>
5253
5254
5255<p>
5256The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5257a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5258no
5259effect on the behavior of the program but can change its performance
5260characteristics.
5261</p>
5262
5263<h5>Arguments:</h5>
5264
5265<p>
5266<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5267determining if the fetch should be for a read (0) or write (1), and
5268<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5269locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5270<tt>locality</tt> arguments must be constant integers.
5271</p>
5272
5273<h5>Semantics:</h5>
5274
5275<p>
5276This intrinsic does not modify the behavior of the program. In particular,
5277prefetches cannot trap and do not produce a value. On targets that support this
5278intrinsic, the prefetch can provide hints to the processor cache for better
5279performance.
5280</p>
5281
5282</div>
5283
5284<!-- _______________________________________________________________________ -->
5285<div class="doc_subsubsection">
5286 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5287</div>
5288
5289<div class="doc_text">
5290
5291<h5>Syntax:</h5>
5292<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5293 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5294</pre>
5295
5296<h5>Overview:</h5>
5297
5298
5299<p>
5300The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5301(PC) in a region of
5302code to simulators and other tools. The method is target specific, but it is
5303expected that the marker will use exported symbols to transmit the PC of the
5304marker.
5305The marker makes no guarantees that it will remain with any specific instruction
5306after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5307optimizations. The intended use is to be inserted after optimizations to allow
5308correlations of simulation runs.
5309</p>
5310
5311<h5>Arguments:</h5>
5312
5313<p>
5314<tt>id</tt> is a numerical id identifying the marker.
5315</p>
5316
5317<h5>Semantics:</h5>
5318
5319<p>
5320This intrinsic does not modify the behavior of the program. Backends that do not
5321support this intrinisic may ignore it.
5322</p>
5323
5324</div>
5325
5326<!-- _______________________________________________________________________ -->
5327<div class="doc_subsubsection">
5328 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5329</div>
5330
5331<div class="doc_text">
5332
5333<h5>Syntax:</h5>
5334<pre>
5335 declare i64 @llvm.readcyclecounter( )
5336</pre>
5337
5338<h5>Overview:</h5>
5339
5340
5341<p>
5342The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5343counter register (or similar low latency, high accuracy clocks) on those targets
5344that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5345As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5346should only be used for small timings.
5347</p>
5348
5349<h5>Semantics:</h5>
5350
5351<p>
5352When directly supported, reading the cycle counter should not modify any memory.
5353Implementations are allowed to either return a application specific value or a
5354system wide value. On backends without support, this is lowered to a constant 0.
5355</p>
5356
5357</div>
5358
5359<!-- ======================================================================= -->
5360<div class="doc_subsection">
5361 <a name="int_libc">Standard C Library Intrinsics</a>
5362</div>
5363
5364<div class="doc_text">
5365<p>
5366LLVM provides intrinsics for a few important standard C library functions.
5367These intrinsics allow source-language front-ends to pass information about the
5368alignment of the pointer arguments to the code generator, providing opportunity
5369for more efficient code generation.
5370</p>
5371
5372</div>
5373
5374<!-- _______________________________________________________________________ -->
5375<div class="doc_subsubsection">
5376 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5377</div>
5378
5379<div class="doc_text">
5380
5381<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5382<p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5383width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5384<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5385 declare void @llvm.memcpy.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5386 i8 &lt;len&gt;, i32 &lt;align&gt;)
5387 declare void @llvm.memcpy.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5388 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5389 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5390 i32 &lt;len&gt;, i32 &lt;align&gt;)
5391 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5392 i64 &lt;len&gt;, i32 &lt;align&gt;)
5393</pre>
5394
5395<h5>Overview:</h5>
5396
5397<p>
5398The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5399location to the destination location.
5400</p>
5401
5402<p>
5403Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5404intrinsics do not return a value, and takes an extra alignment argument.
5405</p>
5406
5407<h5>Arguments:</h5>
5408
5409<p>
5410The first argument is a pointer to the destination, the second is a pointer to
5411the source. The third argument is an integer argument
5412specifying the number of bytes to copy, and the fourth argument is the alignment
5413of the source and destination locations.
5414</p>
5415
5416<p>
5417If the call to this intrinisic has an alignment value that is not 0 or 1, then
5418the caller guarantees that both the source and destination pointers are aligned
5419to that boundary.
5420</p>
5421
5422<h5>Semantics:</h5>
5423
5424<p>
5425The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5426location to the destination location, which are not allowed to overlap. It
5427copies "len" bytes of memory over. If the argument is known to be aligned to
5428some boundary, this can be specified as the fourth argument, otherwise it should
5429be set to 0 or 1.
5430</p>
5431</div>
5432
5433
5434<!-- _______________________________________________________________________ -->
5435<div class="doc_subsubsection">
5436 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5437</div>
5438
5439<div class="doc_text">
5440
5441<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5442<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5443width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5444<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5445 declare void @llvm.memmove.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5446 i8 &lt;len&gt;, i32 &lt;align&gt;)
5447 declare void @llvm.memmove.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5448 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5449 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5450 i32 &lt;len&gt;, i32 &lt;align&gt;)
5451 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5452 i64 &lt;len&gt;, i32 &lt;align&gt;)
5453</pre>
5454
5455<h5>Overview:</h5>
5456
5457<p>
5458The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5459location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5460'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5461</p>
5462
5463<p>
5464Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5465intrinsics do not return a value, and takes an extra alignment argument.
5466</p>
5467
5468<h5>Arguments:</h5>
5469
5470<p>
5471The first argument is a pointer to the destination, the second is a pointer to
5472the source. The third argument is an integer argument
5473specifying the number of bytes to copy, and the fourth argument is the alignment
5474of the source and destination locations.
5475</p>
5476
5477<p>
5478If the call to this intrinisic has an alignment value that is not 0 or 1, then
5479the caller guarantees that the source and destination pointers are aligned to
5480that boundary.
5481</p>
5482
5483<h5>Semantics:</h5>
5484
5485<p>
5486The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5487location to the destination location, which may overlap. It
5488copies "len" bytes of memory over. If the argument is known to be aligned to
5489some boundary, this can be specified as the fourth argument, otherwise it should
5490be set to 0 or 1.
5491</p>
5492</div>
5493
5494
5495<!-- _______________________________________________________________________ -->
5496<div class="doc_subsubsection">
5497 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5498</div>
5499
5500<div class="doc_text">
5501
5502<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5503<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5504width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5505<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5506 declare void @llvm.memset.i8(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5507 i8 &lt;len&gt;, i32 &lt;align&gt;)
5508 declare void @llvm.memset.i16(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5509 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5510 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5511 i32 &lt;len&gt;, i32 &lt;align&gt;)
5512 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5513 i64 &lt;len&gt;, i32 &lt;align&gt;)
5514</pre>
5515
5516<h5>Overview:</h5>
5517
5518<p>
5519The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5520byte value.
5521</p>
5522
5523<p>
5524Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5525does not return a value, and takes an extra alignment argument.
5526</p>
5527
5528<h5>Arguments:</h5>
5529
5530<p>
5531The first argument is a pointer to the destination to fill, the second is the
5532byte value to fill it with, the third argument is an integer
5533argument specifying the number of bytes to fill, and the fourth argument is the
5534known alignment of destination location.
5535</p>
5536
5537<p>
5538If the call to this intrinisic has an alignment value that is not 0 or 1, then
5539the caller guarantees that the destination pointer is aligned to that boundary.
5540</p>
5541
5542<h5>Semantics:</h5>
5543
5544<p>
5545The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5546the
5547destination location. If the argument is known to be aligned to some boundary,
5548this can be specified as the fourth argument, otherwise it should be set to 0 or
55491.
5550</p>
5551</div>
5552
5553
5554<!-- _______________________________________________________________________ -->
5555<div class="doc_subsubsection">
5556 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5557</div>
5558
5559<div class="doc_text">
5560
5561<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5562<p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5563floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5564types however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5565<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5566 declare float @llvm.sqrt.f32(float %Val)
5567 declare double @llvm.sqrt.f64(double %Val)
5568 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5569 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5570 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5571</pre>
5572
5573<h5>Overview:</h5>
5574
5575<p>
5576The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5577returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5578<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5579negative numbers other than -0.0 (which allows for better optimization, because
5580there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5581defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5582</p>
5583
5584<h5>Arguments:</h5>
5585
5586<p>
5587The argument and return value are floating point numbers of the same type.
5588</p>
5589
5590<h5>Semantics:</h5>
5591
5592<p>
5593This function returns the sqrt of the specified operand if it is a nonnegative
5594floating point number.
5595</p>
5596</div>
5597
5598<!-- _______________________________________________________________________ -->
5599<div class="doc_subsubsection">
5600 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5601</div>
5602
5603<div class="doc_text">
5604
5605<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5606<p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5607floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5608types however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5609<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5610 declare float @llvm.powi.f32(float %Val, i32 %power)
5611 declare double @llvm.powi.f64(double %Val, i32 %power)
5612 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5613 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5614 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5615</pre>
5616
5617<h5>Overview:</h5>
5618
5619<p>
5620The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5621specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5622multiplications is not defined. When a vector of floating point type is
5623used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5624</p>
5625
5626<h5>Arguments:</h5>
5627
5628<p>
5629The second argument is an integer power, and the first is a value to raise to
5630that power.
5631</p>
5632
5633<h5>Semantics:</h5>
5634
5635<p>
5636This function returns the first value raised to the second power with an
5637unspecified sequence of rounding operations.</p>
5638</div>
5639
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5640<!-- _______________________________________________________________________ -->
5641<div class="doc_subsubsection">
5642 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5643</div>
5644
5645<div class="doc_text">
5646
5647<h5>Syntax:</h5>
5648<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5649floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5650types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5651<pre>
5652 declare float @llvm.sin.f32(float %Val)
5653 declare double @llvm.sin.f64(double %Val)
5654 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5655 declare fp128 @llvm.sin.f128(fp128 %Val)
5656 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5657</pre>
5658
5659<h5>Overview:</h5>
5660
5661<p>
5662The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5663</p>
5664
5665<h5>Arguments:</h5>
5666
5667<p>
5668The argument and return value are floating point numbers of the same type.
5669</p>
5670
5671<h5>Semantics:</h5>
5672
5673<p>
5674This function returns the sine of the specified operand, returning the
5675same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5676conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5677</div>
5678
5679<!-- _______________________________________________________________________ -->
5680<div class="doc_subsubsection">
5681 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5682</div>
5683
5684<div class="doc_text">
5685
5686<h5>Syntax:</h5>
5687<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5688floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5689types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5690<pre>
5691 declare float @llvm.cos.f32(float %Val)
5692 declare double @llvm.cos.f64(double %Val)
5693 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5694 declare fp128 @llvm.cos.f128(fp128 %Val)
5695 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5696</pre>
5697
5698<h5>Overview:</h5>
5699
5700<p>
5701The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5702</p>
5703
5704<h5>Arguments:</h5>
5705
5706<p>
5707The argument and return value are floating point numbers of the same type.
5708</p>
5709
5710<h5>Semantics:</h5>
5711
5712<p>
5713This function returns the cosine of the specified operand, returning the
5714same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5715conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5716</div>
5717
5718<!-- _______________________________________________________________________ -->
5719<div class="doc_subsubsection">
5720 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5721</div>
5722
5723<div class="doc_text">
5724
5725<h5>Syntax:</h5>
5726<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5727floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5728types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5729<pre>
5730 declare float @llvm.pow.f32(float %Val, float %Power)
5731 declare double @llvm.pow.f64(double %Val, double %Power)
5732 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5733 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5734 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5735</pre>
5736
5737<h5>Overview:</h5>
5738
5739<p>
5740The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5741specified (positive or negative) power.
5742</p>
5743
5744<h5>Arguments:</h5>
5745
5746<p>
5747The second argument is a floating point power, and the first is a value to
5748raise to that power.
5749</p>
5750
5751<h5>Semantics:</h5>
5752
5753<p>
5754This function returns the first value raised to the second power,
5755returning the
5756same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5757conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5758</div>
5759
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5760
5761<!-- ======================================================================= -->
5762<div class="doc_subsection">
5763 <a name="int_manip">Bit Manipulation Intrinsics</a>
5764</div>
5765
5766<div class="doc_text">
5767<p>
5768LLVM provides intrinsics for a few important bit manipulation operations.
5769These allow efficient code generation for some algorithms.
5770</p>
5771
5772</div>
5773
5774<!-- _______________________________________________________________________ -->
5775<div class="doc_subsubsection">
5776 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5777</div>
5778
5779<div class="doc_text">
5780
5781<h5>Syntax:</h5>
5782<p>This is an overloaded intrinsic function. You can use bswap on any integer
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5783type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5784<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5785 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5786 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5787 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5788</pre>
5789
5790<h5>Overview:</h5>
5791
5792<p>
5793The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5794values with an even number of bytes (positive multiple of 16 bits). These are
5795useful for performing operations on data that is not in the target's native
5796byte order.
5797</p>
5798
5799<h5>Semantics:</h5>
5800
5801<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5802The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5803and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5804intrinsic returns an i32 value that has the four bytes of the input i32
5805swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5806i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5807<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5808additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5809</p>
5810
5811</div>
5812
5813<!-- _______________________________________________________________________ -->
5814<div class="doc_subsubsection">
5815 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5816</div>
5817
5818<div class="doc_text">
5819
5820<h5>Syntax:</h5>
5821<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5822width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5823<pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5824 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5825 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5826 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5827 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5828 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5829</pre>
5830
5831<h5>Overview:</h5>
5832
5833<p>
5834The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5835value.
5836</p>
5837
5838<h5>Arguments:</h5>
5839
5840<p>
5841The only argument is the value to be counted. The argument may be of any
5842integer type. The return type must match the argument type.
5843</p>
5844
5845<h5>Semantics:</h5>
5846
5847<p>
5848The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5849</p>
5850</div>
5851
5852<!-- _______________________________________________________________________ -->
5853<div class="doc_subsubsection">
5854 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5855</div>
5856
5857<div class="doc_text">
5858
5859<h5>Syntax:</h5>
5860<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5861integer bit width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5862<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5863 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5864 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5865 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5866 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5867 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5868</pre>
5869
5870<h5>Overview:</h5>
5871
5872<p>
5873The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5874leading zeros in a variable.
5875</p>
5876
5877<h5>Arguments:</h5>
5878
5879<p>
5880The only argument is the value to be counted. The argument may be of any
5881integer type. The return type must match the argument type.
5882</p>
5883
5884<h5>Semantics:</h5>
5885
5886<p>
5887The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5888in a variable. If the src == 0 then the result is the size in bits of the type
5889of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5890</p>
5891</div>
5892
5893
5894
5895<!-- _______________________________________________________________________ -->
5896<div class="doc_subsubsection">
5897 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5898</div>
5899
5900<div class="doc_text">
5901
5902<h5>Syntax:</h5>
5903<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5904integer bit width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5905<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5906 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5907 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5908 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5909 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5910 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5911</pre>
5912
5913<h5>Overview:</h5>
5914
5915<p>
5916The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5917trailing zeros.
5918</p>
5919
5920<h5>Arguments:</h5>
5921
5922<p>
5923The only argument is the value to be counted. The argument may be of any
5924integer type. The return type must match the argument type.
5925</p>
5926
5927<h5>Semantics:</h5>
5928
5929<p>
5930The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5931in a variable. If the src == 0 then the result is the size in bits of the type
5932of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5933</p>
5934</div>
5935
5936<!-- _______________________________________________________________________ -->
5937<div class="doc_subsubsection">
5938 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5939</div>
5940
5941<div class="doc_text">
5942
5943<h5>Syntax:</h5>
5944<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5945on any integer bit width.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5946<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5947 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5948 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5949</pre>
5950
5951<h5>Overview:</h5>
5952<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5953range of bits from an integer value and returns them in the same bit width as
5954the original value.</p>
5955
5956<h5>Arguments:</h5>
5957<p>The first argument, <tt>%val</tt> and the result may be integer types of
5958any bit width but they must have the same bit width. The second and third
5959arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5960
5961<h5>Semantics:</h5>
5962<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5963of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5964<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5965operates in forward mode.</p>
5966<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5967right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5968only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5969<ol>
5970 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5971 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5972 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5973 to determine the number of bits to retain.</li>
5974 <li>A mask of the retained bits is created by shifting a -1 value.</li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5975 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5976</ol>
5977<p>In reverse mode, a similar computation is made except that the bits are
5978returned in the reverse order. So, for example, if <tt>X</tt> has the value
5979<tt>i16 0x0ACF (101011001111)</tt> and we apply
5980<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5981<tt>i16 0x0026 (000000100110)</tt>.</p>
5982</div>
5983
5984<div class="doc_subsubsection">
5985 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5986</div>
5987
5988<div class="doc_text">
5989
5990<h5>Syntax:</h5>
5991<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5992on any integer bit width.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5993<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5994 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5995 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5996</pre>
5997
5998<h5>Overview:</h5>
5999<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
6000of bits in an integer value with another integer value. It returns the integer
6001with the replaced bits.</p>
6002
6003<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6004<p>The first argument, <tt>%val</tt>, and the result may be integer types of
6005any bit width, but they must have the same bit width. <tt>%val</tt> is the value
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6006whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
6007integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
6008type since they specify only a bit index.</p>
6009
6010<h5>Semantics:</h5>
6011<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
6012of operation: forwards and reverse. If <tt>%lo</tt> is greater than
6013<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
6014operates in forward mode.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6015
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6016<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
6017truncating it down to the size of the replacement area or zero extending it
6018up to that size.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6019
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6020<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
6021are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
6022in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6023to the <tt>%hi</tt>th bit.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6024
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6025<p>In reverse mode, a similar computation is made except that the bits are
6026reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6027<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6028
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6029<h5>Examples:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6030
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6031<pre>
6032 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
6033 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
6034 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
6035 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
6036 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
6037</pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6038
6039</div>
6040
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6041<!-- ======================================================================= -->
6042<div class="doc_subsection">
6043 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6044</div>
6045
6046<div class="doc_text">
6047<p>
6048LLVM provides intrinsics for some arithmetic with overflow operations.
6049</p>
6050
6051</div>
6052
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6053<!-- _______________________________________________________________________ -->
6054<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6055 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6056</div>
6057
6058<div class="doc_text">
6059
6060<h5>Syntax:</h5>
6061
6062<p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6063on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6064
6065<pre>
6066 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6067 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6068 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6069</pre>
6070
6071<h5>Overview:</h5>
6072
6073<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6074a signed addition of the two arguments, and indicate whether an overflow
6075occurred during the signed summation.</p>
6076
6077<h5>Arguments:</h5>
6078
6079<p>The arguments (%a and %b) and the first element of the result structure may
6080be of integer types of any bit width, but they must have the same bit width. The
6081second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6082and <tt>%b</tt> are the two values that will undergo signed addition.</p>
6083
6084<h5>Semantics:</h5>
6085
6086<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6087a signed addition of the two variables. They return a structure &mdash; the
6088first element of which is the signed summation, and the second element of which
6089is a bit specifying if the signed summation resulted in an overflow.</p>
6090
6091<h5>Examples:</h5>
6092<pre>
6093 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6094 %sum = extractvalue {i32, i1} %res, 0
6095 %obit = extractvalue {i32, i1} %res, 1
6096 br i1 %obit, label %overflow, label %normal
6097</pre>
6098
6099</div>
6100
6101<!-- _______________________________________________________________________ -->
6102<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6103 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6104</div>
6105
6106<div class="doc_text">
6107
6108<h5>Syntax:</h5>
6109
6110<p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6111on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6112
6113<pre>
6114 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6115 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6116 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6117</pre>
6118
6119<h5>Overview:</h5>
6120
6121<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6122an unsigned addition of the two arguments, and indicate whether a carry occurred
6123during the unsigned summation.</p>
6124
6125<h5>Arguments:</h5>
6126
6127<p>The arguments (%a and %b) and the first element of the result structure may
6128be of integer types of any bit width, but they must have the same bit width. The
6129second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6130and <tt>%b</tt> are the two values that will undergo unsigned addition.</p>
6131
6132<h5>Semantics:</h5>
6133
6134<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6135an unsigned addition of the two arguments. They return a structure &mdash; the
6136first element of which is the sum, and the second element of which is a bit
6137specifying if the unsigned summation resulted in a carry.</p>
6138
6139<h5>Examples:</h5>
6140<pre>
6141 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6142 %sum = extractvalue {i32, i1} %res, 0
6143 %obit = extractvalue {i32, i1} %res, 1
6144 br i1 %obit, label %carry, label %normal
6145</pre>
6146
6147</div>
6148
6149<!-- _______________________________________________________________________ -->
6150<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6151 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6152</div>
6153
6154<div class="doc_text">
6155
6156<h5>Syntax:</h5>
6157
6158<p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6159on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6160
6161<pre>
6162 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6163 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6164 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6165</pre>
6166
6167<h5>Overview:</h5>
6168
6169<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6170a signed subtraction of the two arguments, and indicate whether an overflow
6171occurred during the signed subtraction.</p>
6172
6173<h5>Arguments:</h5>
6174
6175<p>The arguments (%a and %b) and the first element of the result structure may
6176be of integer types of any bit width, but they must have the same bit width. The
6177second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6178and <tt>%b</tt> are the two values that will undergo signed subtraction.</p>
6179
6180<h5>Semantics:</h5>
6181
6182<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6183a signed subtraction of the two arguments. They return a structure &mdash; the
6184first element of which is the subtraction, and the second element of which is a bit
6185specifying if the signed subtraction resulted in an overflow.</p>
6186
6187<h5>Examples:</h5>
6188<pre>
6189 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6190 %sum = extractvalue {i32, i1} %res, 0
6191 %obit = extractvalue {i32, i1} %res, 1
6192 br i1 %obit, label %overflow, label %normal
6193</pre>
6194
6195</div>
6196
6197<!-- _______________________________________________________________________ -->
6198<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6199 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6200</div>
6201
6202<div class="doc_text">
6203
6204<h5>Syntax:</h5>
6205
6206<p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6207on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6208
6209<pre>
6210 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6211 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6212 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6213</pre>
6214
6215<h5>Overview:</h5>
6216
6217<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6218an unsigned subtraction of the two arguments, and indicate whether an overflow
6219occurred during the unsigned subtraction.</p>
6220
6221<h5>Arguments:</h5>
6222
6223<p>The arguments (%a and %b) and the first element of the result structure may
6224be of integer types of any bit width, but they must have the same bit width. The
6225second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6226and <tt>%b</tt> are the two values that will undergo unsigned subtraction.</p>
6227
6228<h5>Semantics:</h5>
6229
6230<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6231an unsigned subtraction of the two arguments. They return a structure &mdash; the
6232first element of which is the subtraction, and the second element of which is a bit
6233specifying if the unsigned subtraction resulted in an overflow.</p>
6234
6235<h5>Examples:</h5>
6236<pre>
6237 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6238 %sum = extractvalue {i32, i1} %res, 0
6239 %obit = extractvalue {i32, i1} %res, 1
6240 br i1 %obit, label %overflow, label %normal
6241</pre>
6242
6243</div>
6244
6245<!-- _______________________________________________________________________ -->
6246<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6247 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6248</div>
6249
6250<div class="doc_text">
6251
6252<h5>Syntax:</h5>
6253
6254<p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6255on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6256
6257<pre>
6258 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6259 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6260 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6261</pre>
6262
6263<h5>Overview:</h5>
6264
6265<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6266a signed multiplication of the two arguments, and indicate whether an overflow
6267occurred during the signed multiplication.</p>
6268
6269<h5>Arguments:</h5>
6270
6271<p>The arguments (%a and %b) and the first element of the result structure may
6272be of integer types of any bit width, but they must have the same bit width. The
6273second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6274and <tt>%b</tt> are the two values that will undergo signed multiplication.</p>
6275
6276<h5>Semantics:</h5>
6277
6278<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6279a signed multiplication of the two arguments. They return a structure &mdash;
6280the first element of which is the multiplication, and the second element of
6281which is a bit specifying if the signed multiplication resulted in an
6282overflow.</p>
6283
6284<h5>Examples:</h5>
6285<pre>
6286 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6287 %sum = extractvalue {i32, i1} %res, 0
6288 %obit = extractvalue {i32, i1} %res, 1
6289 br i1 %obit, label %overflow, label %normal
6290</pre>
6291
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6292</div>
6293
Bill Wendlingbda98b62009-02-08 23:00:09 +0000[diff] [blame]6294<!-- _______________________________________________________________________ -->
6295<div class="doc_subsubsection">
6296 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6297</div>
6298
6299<div class="doc_text">
6300
6301<h5>Syntax:</h5>
6302
6303<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6304on any integer bit width.</p>
6305
6306<pre>
6307 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6308 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6309 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6310</pre>
6311
6312<h5>Overview:</h5>
6313
6314<p><i><b>Warning:</b> '<tt>llvm.umul.with.overflow</tt>' is badly broken. It is
6315actively being fixed, but it should not currently be used!</i></p>
6316
6317<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6318a unsigned multiplication of the two arguments, and indicate whether an overflow
6319occurred during the unsigned multiplication.</p>
6320
6321<h5>Arguments:</h5>
6322
6323<p>The arguments (%a and %b) and the first element of the result structure may
6324be of integer types of any bit width, but they must have the same bit width. The
6325second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6326and <tt>%b</tt> are the two values that will undergo unsigned
6327multiplication.</p>
6328
6329<h5>Semantics:</h5>
6330
6331<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6332an unsigned multiplication of the two arguments. They return a structure &mdash;
6333the first element of which is the multiplication, and the second element of
6334which is a bit specifying if the unsigned multiplication resulted in an
6335overflow.</p>
6336
6337<h5>Examples:</h5>
6338<pre>
6339 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6340 %sum = extractvalue {i32, i1} %res, 0
6341 %obit = extractvalue {i32, i1} %res, 1
6342 br i1 %obit, label %overflow, label %normal
6343</pre>
6344
6345</div>
6346
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6347<!-- ======================================================================= -->
6348<div class="doc_subsection">
6349 <a name="int_debugger">Debugger Intrinsics</a>
6350</div>
6351
6352<div class="doc_text">
6353<p>
6354The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
6355are described in the <a
6356href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
6357Debugging</a> document.
6358</p>
6359</div>
6360
6361
6362<!-- ======================================================================= -->
6363<div class="doc_subsection">
6364 <a name="int_eh">Exception Handling Intrinsics</a>
6365</div>
6366
6367<div class="doc_text">
6368<p> The LLVM exception handling intrinsics (which all start with
6369<tt>llvm.eh.</tt> prefix), are described in the <a
6370href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6371Handling</a> document. </p>
6372</div>
6373
6374<!-- ======================================================================= -->
6375<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6376 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6377</div>
6378
6379<div class="doc_text">
6380<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6381 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6382 the <tt>nest</tt> attribute, from a function. The result is a callable
6383 function pointer lacking the nest parameter - the caller does not need
6384 to provide a value for it. Instead, the value to use is stored in
6385 advance in a "trampoline", a block of memory usually allocated
6386 on the stack, which also contains code to splice the nest value into the
6387 argument list. This is used to implement the GCC nested function address
6388 extension.
6389</p>
6390<p>
6391 For example, if the function is
6392 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]6393 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6394<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6395 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6396 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6397 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6398 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6399</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]6400 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6401 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6402</div>
6403
6404<!-- _______________________________________________________________________ -->
6405<div class="doc_subsubsection">
6406 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6407</div>
6408<div class="doc_text">
6409<h5>Syntax:</h5>
6410<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6411declare i8* @llvm.init.trampoline(i8* &lt;tramp&gt;, i8* &lt;func&gt;, i8* &lt;nval&gt;)
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6412</pre>
6413<h5>Overview:</h5>
6414<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6415 This fills the memory pointed to by <tt>tramp</tt> with code
6416 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6417</p>
6418<h5>Arguments:</h5>
6419<p>
6420 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6421 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
6422 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]6423 intrinsic. Note that the size and the alignment are target-specific - LLVM
6424 currently provides no portable way of determining them, so a front-end that
6425 generates this intrinsic needs to have some target-specific knowledge.
6426 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6427</p>
6428<h5>Semantics:</h5>
6429<p>
6430 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6431 dependent code, turning it into a function. A pointer to this function is
6432 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6433 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6434 before being called. The new function's signature is the same as that of
6435 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
6436 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
6437 of pointer type. Calling the new function is equivalent to calling
6438 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
6439 missing <tt>nest</tt> argument. If, after calling
6440 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
6441 modified, then the effect of any later call to the returned function pointer is
6442 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6443</p>
6444</div>
6445
6446<!-- ======================================================================= -->
6447<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6448 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6449</div>
6450
6451<div class="doc_text">
6452<p>
6453 These intrinsic functions expand the "universal IR" of LLVM to represent
6454 hardware constructs for atomic operations and memory synchronization. This
6455 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]6456 is aimed at a low enough level to allow any programming models or APIs
6457 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6458 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6459 hardware behavior. Just as hardware provides a "universal IR" for source
6460 languages, it also provides a starting point for developing a "universal"
6461 atomic operation and synchronization IR.
6462</p>
6463<p>
6464 These do <em>not</em> form an API such as high-level threading libraries,
6465 software transaction memory systems, atomic primitives, and intrinsic
6466 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6467 application libraries. The hardware interface provided by LLVM should allow
6468 a clean implementation of all of these APIs and parallel programming models.
6469 No one model or paradigm should be selected above others unless the hardware
6470 itself ubiquitously does so.
6471
6472</p>
6473</div>
6474
6475<!-- _______________________________________________________________________ -->
6476<div class="doc_subsubsection">
6477 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6478</div>
6479<div class="doc_text">
6480<h5>Syntax:</h5>
6481<pre>
6482declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
6483i1 &lt;device&gt; )
6484
6485</pre>
6486<h5>Overview:</h5>
6487<p>
6488 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6489 specific pairs of memory access types.
6490</p>
6491<h5>Arguments:</h5>
6492<p>
6493 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6494 The first four arguments enables a specific barrier as listed below. The fith
6495 argument specifies that the barrier applies to io or device or uncached memory.
6496
6497</p>
6498 <ul>
6499 <li><tt>ll</tt>: load-load barrier</li>
6500 <li><tt>ls</tt>: load-store barrier</li>
6501 <li><tt>sl</tt>: store-load barrier</li>
6502 <li><tt>ss</tt>: store-store barrier</li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6503 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6504 </ul>
6505<h5>Semantics:</h5>
6506<p>
6507 This intrinsic causes the system to enforce some ordering constraints upon
6508 the loads and stores of the program. This barrier does not indicate
6509 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6510 which they occur. For any of the specified pairs of load and store operations
6511 (f.ex. load-load, or store-load), all of the first operations preceding the
6512 barrier will complete before any of the second operations succeeding the
6513 barrier begin. Specifically the semantics for each pairing is as follows:
6514</p>
6515 <ul>
6516 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6517 after the barrier begins.</li>
6518
6519 <li><tt>ls</tt>: All loads before the barrier must complete before any
6520 store after the barrier begins.</li>
6521 <li><tt>ss</tt>: All stores before the barrier must complete before any
6522 store after the barrier begins.</li>
6523 <li><tt>sl</tt>: All stores before the barrier must complete before any
6524 load after the barrier begins.</li>
6525 </ul>
6526<p>
6527 These semantics are applied with a logical "and" behavior when more than one
6528 is enabled in a single memory barrier intrinsic.
6529</p>
6530<p>
6531 Backends may implement stronger barriers than those requested when they do not
6532 support as fine grained a barrier as requested. Some architectures do not
6533 need all types of barriers and on such architectures, these become noops.
6534</p>
6535<h5>Example:</h5>
6536<pre>
6537%ptr = malloc i32
6538 store i32 4, %ptr
6539
6540%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6541 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6542 <i>; guarantee the above finishes</i>
6543 store i32 8, %ptr <i>; before this begins</i>
6544</pre>
6545</div>
6546
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6547<!-- _______________________________________________________________________ -->
6548<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6549 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6550</div>
6551<div class="doc_text">
6552<h5>Syntax:</h5>
6553<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6554 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6555 any integer bit width and for different address spaces. Not all targets
6556 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6557
6558<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6559declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
6560declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
6561declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
6562declare 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 +0000[diff] [blame]6563
6564</pre>
6565<h5>Overview:</h5>
6566<p>
6567 This loads a value in memory and compares it to a given value. If they are
6568 equal, it stores a new value into the memory.
6569</p>
6570<h5>Arguments:</h5>
6571<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6572 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6573 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6574 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6575 this integer type. While any bit width integer may be used, targets may only
6576 lower representations they support in hardware.
6577
6578</p>
6579<h5>Semantics:</h5>
6580<p>
6581 This entire intrinsic must be executed atomically. It first loads the value
6582 in memory pointed to by <tt>ptr</tt> and compares it with the value
6583 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6584 loaded value is yielded in all cases. This provides the equivalent of an
6585 atomic compare-and-swap operation within the SSA framework.
6586</p>
6587<h5>Examples:</h5>
6588
6589<pre>
6590%ptr = malloc i32
6591 store i32 4, %ptr
6592
6593%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6594%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6595 <i>; yields {i32}:result1 = 4</i>
6596%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6597%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6598
6599%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6600%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6601 <i>; yields {i32}:result2 = 8</i>
6602%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6603
6604%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6605</pre>
6606</div>
6607
6608<!-- _______________________________________________________________________ -->
6609<div class="doc_subsubsection">
6610 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6611</div>
6612<div class="doc_text">
6613<h5>Syntax:</h5>
6614
6615<p>
6616 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6617 integer bit width. Not all targets support all bit widths however.</p>
6618<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6619declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
6620declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
6621declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
6622declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6623
6624</pre>
6625<h5>Overview:</h5>
6626<p>
6627 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6628 the value from memory. It then stores the value in <tt>val</tt> in the memory
6629 at <tt>ptr</tt>.
6630</p>
6631<h5>Arguments:</h5>
6632
6633<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6634 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6635 <tt>val</tt> argument and the result must be integers of the same bit width.
6636 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6637 integer type. The targets may only lower integer representations they
6638 support.
6639</p>
6640<h5>Semantics:</h5>
6641<p>
6642 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6643 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6644 equivalent of an atomic swap operation within the SSA framework.
6645
6646</p>
6647<h5>Examples:</h5>
6648<pre>
6649%ptr = malloc i32
6650 store i32 4, %ptr
6651
6652%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6653%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6654 <i>; yields {i32}:result1 = 4</i>
6655%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6656%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6657
6658%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6659%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6660 <i>; yields {i32}:result2 = 8</i>
6661
6662%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6663%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6664</pre>
6665</div>
6666
6667<!-- _______________________________________________________________________ -->
6668<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6669 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6670
6671</div>
6672<div class="doc_text">
6673<h5>Syntax:</h5>
6674<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6675 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6676 integer bit width. Not all targets support all bit widths however.</p>
6677<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6678declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6679declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6680declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6681declare i64 @llvm.atomic.load.add.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6682
6683</pre>
6684<h5>Overview:</h5>
6685<p>
6686 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6687 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6688</p>
6689<h5>Arguments:</h5>
6690<p>
6691
6692 The intrinsic takes two arguments, the first a pointer to an integer value
6693 and the second an integer value. The result is also an integer value. These
6694 integer types can have any bit width, but they must all have the same bit
6695 width. The targets may only lower integer representations they support.
6696</p>
6697<h5>Semantics:</h5>
6698<p>
6699 This intrinsic does a series of operations atomically. It first loads the
6700 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6701 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6702</p>
6703
6704<h5>Examples:</h5>
6705<pre>
6706%ptr = malloc i32
6707 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6708%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6709 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6710%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6711 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6712%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6713 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6714%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6715</pre>
6716</div>
6717
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6718<!-- _______________________________________________________________________ -->
6719<div class="doc_subsubsection">
6720 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6721
6722</div>
6723<div class="doc_text">
6724<h5>Syntax:</h5>
6725<p>
6726 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6727 any integer bit width and for different address spaces. Not all targets
6728 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6729<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6730declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6731declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6732declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6733declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6734
6735</pre>
6736<h5>Overview:</h5>
6737<p>
6738 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6739 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6740</p>
6741<h5>Arguments:</h5>
6742<p>
6743
6744 The intrinsic takes two arguments, the first a pointer to an integer value
6745 and the second an integer value. The result is also an integer value. These
6746 integer types can have any bit width, but they must all have the same bit
6747 width. The targets may only lower integer representations they support.
6748</p>
6749<h5>Semantics:</h5>
6750<p>
6751 This intrinsic does a series of operations atomically. It first loads the
6752 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6753 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6754</p>
6755
6756<h5>Examples:</h5>
6757<pre>
6758%ptr = malloc i32
6759 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6760%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6761 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6762%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6763 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6764%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6765 <i>; yields {i32}:result3 = 2</i>
6766%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6767</pre>
6768</div>
6769
6770<!-- _______________________________________________________________________ -->
6771<div class="doc_subsubsection">
6772 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6773 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6774 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6775 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6776
6777</div>
6778<div class="doc_text">
6779<h5>Syntax:</h5>
6780<p>
6781 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6782 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6783 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6784 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6785<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6786declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6787declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6788declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6789declare i64 @llvm.atomic.load.and.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6790
6791</pre>
6792
6793<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6794declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6795declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6796declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6797declare i64 @llvm.atomic.load.or.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6798
6799</pre>
6800
6801<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6802declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6803declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6804declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6805declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6806
6807</pre>
6808
6809<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6810declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6811declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6812declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6813declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6814
6815</pre>
6816<h5>Overview:</h5>
6817<p>
6818 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6819 the value stored in memory at <tt>ptr</tt>. It yields the original value
6820 at <tt>ptr</tt>.
6821</p>
6822<h5>Arguments:</h5>
6823<p>
6824
6825 These intrinsics take two arguments, the first a pointer to an integer value
6826 and the second an integer value. The result is also an integer value. These
6827 integer types can have any bit width, but they must all have the same bit
6828 width. The targets may only lower integer representations they support.
6829</p>
6830<h5>Semantics:</h5>
6831<p>
6832 These intrinsics does a series of operations atomically. They first load the
6833 value stored at <tt>ptr</tt>. They then do the bitwise operation
6834 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6835 value stored at <tt>ptr</tt>.
6836</p>
6837
6838<h5>Examples:</h5>
6839<pre>
6840%ptr = malloc i32
6841 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6842%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6843 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6844%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6845 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6846%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6847 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6848%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6849 <i>; yields {i32}:result3 = FF</i>
6850%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6851</pre>
6852</div>
6853
6854
6855<!-- _______________________________________________________________________ -->
6856<div class="doc_subsubsection">
6857 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6858 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6859 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6860 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6861
6862</div>
6863<div class="doc_text">
6864<h5>Syntax:</h5>
6865<p>
6866 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6867 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6868 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6869 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6870 support all bit widths however.</p>
6871<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6872declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6873declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6874declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6875declare i64 @llvm.atomic.load.max.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6876
6877</pre>
6878
6879<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6880declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6881declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6882declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6883declare i64 @llvm.atomic.load.min.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6884
6885</pre>
6886
6887<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6888declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6889declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6890declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6891declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6892
6893</pre>
6894
6895<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6896declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6897declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6898declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6899declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6900
6901</pre>
6902<h5>Overview:</h5>
6903<p>
6904 These intrinsics takes the signed or unsigned minimum or maximum of
6905 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6906 original value at <tt>ptr</tt>.
6907</p>
6908<h5>Arguments:</h5>
6909<p>
6910
6911 These intrinsics take two arguments, the first a pointer to an integer value
6912 and the second an integer value. The result is also an integer value. These
6913 integer types can have any bit width, but they must all have the same bit
6914 width. The targets may only lower integer representations they support.
6915</p>
6916<h5>Semantics:</h5>
6917<p>
6918 These intrinsics does a series of operations atomically. They first load the
6919 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6920 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6921 the original value stored at <tt>ptr</tt>.
6922</p>
6923
6924<h5>Examples:</h5>
6925<pre>
6926%ptr = malloc i32
6927 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6928%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6929 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6930%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6931 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6932%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6933 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6934%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6935 <i>; yields {i32}:result3 = 8</i>
6936%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6937</pre>
6938</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6939
6940<!-- ======================================================================= -->
6941<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6942 <a name="int_general">General Intrinsics</a>
6943</div>
6944
6945<div class="doc_text">
6946<p> This class of intrinsics is designed to be generic and has
6947no specific purpose. </p>
6948</div>
6949
6950<!-- _______________________________________________________________________ -->
6951<div class="doc_subsubsection">
6952 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6953</div>
6954
6955<div class="doc_text">
6956
6957<h5>Syntax:</h5>
6958<pre>
6959 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6960</pre>
6961
6962<h5>Overview:</h5>
6963
6964<p>
6965The '<tt>llvm.var.annotation</tt>' intrinsic
6966</p>
6967
6968<h5>Arguments:</h5>
6969
6970<p>
6971The first argument is a pointer to a value, the second is a pointer to a
6972global string, the third is a pointer to a global string which is the source
6973file name, and the last argument is the line number.
6974</p>
6975
6976<h5>Semantics:</h5>
6977
6978<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6979This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6980This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6981annotations. These have no other defined use, they are ignored by code
6982generation and optimization.
6983</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6984</div>
6985
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6986<!-- _______________________________________________________________________ -->
6987<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6988 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6989</div>
6990
6991<div class="doc_text">
6992
6993<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6994<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6995any integer bit width.
6996</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6997<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6998 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6999 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
7000 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
7001 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
7002 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 +0000[diff] [blame]7003</pre>
7004
7005<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]7006
7007<p>
7008The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]7009</p>
7010
7011<h5>Arguments:</h5>
7012
7013<p>
7014The first argument is an integer value (result of some expression),
7015the second is a pointer to a global string, the third is a pointer to a global
7016string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]7017It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]7018</p>
7019
7020<h5>Semantics:</h5>
7021
7022<p>
7023This intrinsic allows annotations to be put on arbitrary expressions
7024with arbitrary strings. This can be useful for special purpose optimizations
7025that want to look for these annotations. These have no other defined use, they
7026are ignored by code generation and optimization.
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]7027</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]7028</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7029
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]7030<!-- _______________________________________________________________________ -->
7031<div class="doc_subsubsection">
7032 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7033</div>
7034
7035<div class="doc_text">
7036
7037<h5>Syntax:</h5>
7038<pre>
7039 declare void @llvm.trap()
7040</pre>
7041
7042<h5>Overview:</h5>
7043
7044<p>
7045The '<tt>llvm.trap</tt>' intrinsic
7046</p>
7047
7048<h5>Arguments:</h5>
7049
7050<p>
7051None
7052</p>
7053
7054<h5>Semantics:</h5>
7055
7056<p>
7057This intrinsics is lowered to the target dependent trap instruction. If the
7058target does not have a trap instruction, this intrinsic will be lowered to the
7059call of the abort() function.
7060</p>
7061</div>
7062
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]7063<!-- _______________________________________________________________________ -->
7064<div class="doc_subsubsection">
Misha Brukman5dd7f4d2008-11-22 23:55:29 +0000[diff] [blame]7065 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]7066</div>
7067<div class="doc_text">
7068<h5>Syntax:</h5>
7069<pre>
7070declare void @llvm.stackprotector( i8* &lt;guard&gt;, i8** &lt;slot&gt; )
7071
7072</pre>
7073<h5>Overview:</h5>
7074<p>
7075 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
7076 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
7077 it is placed on the stack before local variables.
7078</p>
7079<h5>Arguments:</h5>
7080<p>
7081 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
7082 first argument is the value loaded from the stack guard
7083 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
7084 has enough space to hold the value of the guard.
7085</p>
7086<h5>Semantics:</h5>
7087<p>
7088 This intrinsic causes the prologue/epilogue inserter to force the position of
7089 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7090 stack. This is to ensure that if a local variable on the stack is overwritten,
7091 it will destroy the value of the guard. When the function exits, the guard on
7092 the stack is checked against the original guard. If they're different, then
7093 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
7094</p>
7095</div>
7096
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7097<!-- *********************************************************************** -->
7098<hr>
7099<address>
7100 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]7101 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7102 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]7103 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7104
7105 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7106 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7107 Last modified: $Date$
7108</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]7109
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7110</body>
7111</html>