Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / llvm / 32a080e1e42fd011b0055bc4f29d7124e217a077 / . / docs / LangRef.html
blob: 99eb163f8caef6f343441b6fe157e58adf1757de [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>
25 <li><a href="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#aliasstructure">Aliases</a>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]29 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]30 <li><a href="#gc">Garbage Collector Names</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]31 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
32 <li><a href="#datalayout">Data Layout</a></li>
33 </ol>
34 </li>
35 <li><a href="#typesystem">Type System</a>
36 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]37 <li><a href="#t_classifications">Type Classifications</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]38 <li><a href="#t_primitive">Primitive Types</a>
39 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]40 <li><a href="#t_floating">Floating Point Types</a></li>
41 <li><a href="#t_void">Void Type</a></li>
42 <li><a href="#t_label">Label Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]43 </ol>
44 </li>
45 <li><a href="#t_derived">Derived Types</a>
46 <ol>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]47 <li><a href="#t_integer">Integer Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]48 <li><a href="#t_array">Array Type</a></li>
49 <li><a href="#t_function">Function Type</a></li>
50 <li><a href="#t_pointer">Pointer Type</a></li>
51 <li><a href="#t_struct">Structure Type</a></li>
52 <li><a href="#t_pstruct">Packed Structure Type</a></li>
53 <li><a href="#t_vector">Vector Type</a></li>
54 <li><a href="#t_opaque">Opaque Type</a></li>
55 </ol>
56 </li>
57 </ol>
58 </li>
59 <li><a href="#constants">Constants</a>
60 <ol>
61 <li><a href="#simpleconstants">Simple Constants</a>
62 <li><a href="#aggregateconstants">Aggregate Constants</a>
63 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
64 <li><a href="#undefvalues">Undefined Values</a>
65 <li><a href="#constantexprs">Constant Expressions</a>
66 </ol>
67 </li>
68 <li><a href="#othervalues">Other Values</a>
69 <ol>
70 <li><a href="#inlineasm">Inline Assembler Expressions</a>
71 </ol>
72 </li>
73 <li><a href="#instref">Instruction Reference</a>
74 <ol>
75 <li><a href="#terminators">Terminator Instructions</a>
76 <ol>
77 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
78 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
79 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
80 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
81 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
82 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
83 </ol>
84 </li>
85 <li><a href="#binaryops">Binary Operations</a>
86 <ol>
87 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
88 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
89 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
90 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
91 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
92 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
93 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
94 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
95 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
96 </ol>
97 </li>
98 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
99 <ol>
100 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
101 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
102 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
103 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
104 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
105 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
106 </ol>
107 </li>
108 <li><a href="#vectorops">Vector Operations</a>
109 <ol>
110 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
111 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
112 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
113 </ol>
114 </li>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]115 <li><a href="#aggregateops">Aggregate Operations</a>
116 <ol>
117 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
118 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
119 </ol>
120 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]121 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
122 <ol>
123 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
124 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
125 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
126 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
127 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
128 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
129 </ol>
130 </li>
131 <li><a href="#convertops">Conversion Operations</a>
132 <ol>
133 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
134 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
135 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
136 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
137 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
140 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
141 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
142 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
143 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
144 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
145 </ol>
146 <li><a href="#otherops">Other Operations</a>
147 <ol>
148 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
149 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]150 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
151 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]152 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
153 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
154 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
155 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
156 </ol>
157 </li>
158 </ol>
159 </li>
160 <li><a href="#intrinsics">Intrinsic Functions</a>
161 <ol>
162 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
163 <ol>
164 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
165 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
166 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
167 </ol>
168 </li>
169 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
170 <ol>
171 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
172 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
173 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
174 </ol>
175 </li>
176 <li><a href="#int_codegen">Code Generator Intrinsics</a>
177 <ol>
178 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
179 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
180 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
181 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
182 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
183 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
184 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
185 </ol>
186 </li>
187 <li><a href="#int_libc">Standard C Library Intrinsics</a>
188 <ol>
189 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
190 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
191 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
192 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]194 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]197 </ol>
198 </li>
199 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
200 <ol>
201 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
202 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
203 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
204 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
205 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
207 </ol>
208 </li>
209 <li><a href="#int_debugger">Debugger intrinsics</a></li>
210 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]211 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]212 <ol>
213 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]214 </ol>
215 </li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]216 <li><a href="#int_atomics">Atomic intrinsics</a>
217 <ol>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]218 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]219 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]220 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]221 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
222 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
223 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
224 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
225 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
226 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
227 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
228 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
229 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
230 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]231 </ol>
232 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]233 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]234 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]235 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]236 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]237 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]238 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]239 <li><a href="#int_trap">
240 <tt>llvm.trap</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]241 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]242 </li>
243 </ol>
244 </li>
245</ol>
246
247<div class="doc_author">
248 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
249 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
250</div>
251
252<!-- *********************************************************************** -->
253<div class="doc_section"> <a name="abstract">Abstract </a></div>
254<!-- *********************************************************************** -->
255
256<div class="doc_text">
257<p>This document is a reference manual for the LLVM assembly language.
Bill Wendlinge7846a52008-08-05 22:29:16 +0000[diff] [blame]258LLVM is a Static Single Assignment (SSA) based representation that provides
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]259type safety, low-level operations, flexibility, and the capability of
260representing 'all' high-level languages cleanly. It is the common code
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]261representation used throughout all phases of the LLVM compilation
262strategy.</p>
263</div>
264
265<!-- *********************************************************************** -->
266<div class="doc_section"> <a name="introduction">Introduction</a> </div>
267<!-- *********************************************************************** -->
268
269<div class="doc_text">
270
271<p>The LLVM code representation is designed to be used in three
272different forms: as an in-memory compiler IR, as an on-disk bitcode
273representation (suitable for fast loading by a Just-In-Time compiler),
274and as a human readable assembly language representation. This allows
275LLVM to provide a powerful intermediate representation for efficient
276compiler transformations and analysis, while providing a natural means
277to debug and visualize the transformations. The three different forms
278of LLVM are all equivalent. This document describes the human readable
279representation and notation.</p>
280
281<p>The LLVM representation aims to be light-weight and low-level
282while being expressive, typed, and extensible at the same time. It
283aims to be a "universal IR" of sorts, by being at a low enough level
284that high-level ideas may be cleanly mapped to it (similar to how
285microprocessors are "universal IR's", allowing many source languages to
286be mapped to them). By providing type information, LLVM can be used as
287the target of optimizations: for example, through pointer analysis, it
288can be proven that a C automatic variable is never accessed outside of
289the current function... allowing it to be promoted to a simple SSA
290value instead of a memory location.</p>
291
292</div>
293
294<!-- _______________________________________________________________________ -->
295<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
296
297<div class="doc_text">
298
299<p>It is important to note that this document describes 'well formed'
300LLVM assembly language. There is a difference between what the parser
301accepts and what is considered 'well formed'. For example, the
302following instruction is syntactically okay, but not well formed:</p>
303
304<div class="doc_code">
305<pre>
306%x = <a href="#i_add">add</a> i32 1, %x
307</pre>
308</div>
309
310<p>...because the definition of <tt>%x</tt> does not dominate all of
311its uses. The LLVM infrastructure provides a verification pass that may
312be used to verify that an LLVM module is well formed. This pass is
313automatically run by the parser after parsing input assembly and by
314the optimizer before it outputs bitcode. The violations pointed out
315by the verifier pass indicate bugs in transformation passes or input to
316the parser.</p>
317</div>
318
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]319<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]320
321<!-- *********************************************************************** -->
322<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
323<!-- *********************************************************************** -->
324
325<div class="doc_text">
326
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]327 <p>LLVM identifiers come in two basic types: global and local. Global
328 identifiers (functions, global variables) begin with the @ character. Local
329 identifiers (register names, types) begin with the % character. Additionally,
330 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]331
332<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]333 <li>Named values are represented as a string of characters with their prefix.
334 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
335 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]336 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]337 with quotes. In this way, anything except a <tt>&quot;</tt> character can
338 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]339
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]340 <li>Unnamed values are represented as an unsigned numeric value with their
341 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]342
343 <li>Constants, which are described in a <a href="#constants">section about
344 constants</a>, below.</li>
345</ol>
346
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]347<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]348don't need to worry about name clashes with reserved words, and the set of
349reserved words may be expanded in the future without penalty. Additionally,
350unnamed identifiers allow a compiler to quickly come up with a temporary
351variable without having to avoid symbol table conflicts.</p>
352
353<p>Reserved words in LLVM are very similar to reserved words in other
354languages. There are keywords for different opcodes
355('<tt><a href="#i_add">add</a></tt>',
356 '<tt><a href="#i_bitcast">bitcast</a></tt>',
357 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
358href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
359and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]360none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]361
362<p>Here is an example of LLVM code to multiply the integer variable
363'<tt>%X</tt>' by 8:</p>
364
365<p>The easy way:</p>
366
367<div class="doc_code">
368<pre>
369%result = <a href="#i_mul">mul</a> i32 %X, 8
370</pre>
371</div>
372
373<p>After strength reduction:</p>
374
375<div class="doc_code">
376<pre>
377%result = <a href="#i_shl">shl</a> i32 %X, i8 3
378</pre>
379</div>
380
381<p>And the hard way:</p>
382
383<div class="doc_code">
384<pre>
385<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
386<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
387%result = <a href="#i_add">add</a> i32 %1, %1
388</pre>
389</div>
390
391<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
392important lexical features of LLVM:</p>
393
394<ol>
395
396 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
397 line.</li>
398
399 <li>Unnamed temporaries are created when the result of a computation is not
400 assigned to a named value.</li>
401
402 <li>Unnamed temporaries are numbered sequentially</li>
403
404</ol>
405
406<p>...and it also shows a convention that we follow in this document. When
407demonstrating instructions, we will follow an instruction with a comment that
408defines the type and name of value produced. Comments are shown in italic
409text.</p>
410
411</div>
412
413<!-- *********************************************************************** -->
414<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
415<!-- *********************************************************************** -->
416
417<!-- ======================================================================= -->
418<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
419</div>
420
421<div class="doc_text">
422
423<p>LLVM programs are composed of "Module"s, each of which is a
424translation unit of the input programs. Each module consists of
425functions, global variables, and symbol table entries. Modules may be
426combined together with the LLVM linker, which merges function (and
427global variable) definitions, resolves forward declarations, and merges
428symbol table entries. Here is an example of the "hello world" module:</p>
429
430<div class="doc_code">
431<pre><i>; Declare the string constant as a global constant...</i>
432<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
433 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
434
435<i>; External declaration of the puts function</i>
436<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
437
438<i>; Definition of main function</i>
439define i32 @main() { <i>; i32()* </i>
440 <i>; Convert [13x i8 ]* to i8 *...</i>
441 %cast210 = <a
442 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
443
444 <i>; Call puts function to write out the string to stdout...</i>
445 <a
446 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
447 <a
448 href="#i_ret">ret</a> i32 0<br>}<br>
449</pre>
450</div>
451
452<p>This example is made up of a <a href="#globalvars">global variable</a>
453named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
454function, and a <a href="#functionstructure">function definition</a>
455for "<tt>main</tt>".</p>
456
457<p>In general, a module is made up of a list of global values,
458where both functions and global variables are global values. Global values are
459represented by a pointer to a memory location (in this case, a pointer to an
460array of char, and a pointer to a function), and have one of the following <a
461href="#linkage">linkage types</a>.</p>
462
463</div>
464
465<!-- ======================================================================= -->
466<div class="doc_subsection">
467 <a name="linkage">Linkage Types</a>
468</div>
469
470<div class="doc_text">
471
472<p>
473All Global Variables and Functions have one of the following types of linkage:
474</p>
475
476<dl>
477
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]478 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]479
480 <dd>Global values with internal linkage are only directly accessible by
481 objects in the current module. In particular, linking code into a module with
482 an internal global value may cause the internal to be renamed as necessary to
483 avoid collisions. Because the symbol is internal to the module, all
484 references can be updated. This corresponds to the notion of the
485 '<tt>static</tt>' keyword in C.
486 </dd>
487
488 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
489
490 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
491 the same name when linkage occurs. This is typically used to implement
492 inline functions, templates, or other code which must be generated in each
493 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
494 allowed to be discarded.
495 </dd>
496
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]497 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
498
499 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
500 linkage, except that unreferenced <tt>common</tt> globals may not be
501 discarded. This is used for globals that may be emitted in multiple
502 translation units, but that are not guaranteed to be emitted into every
503 translation unit that uses them. One example of this is tentative
504 definitions in C, such as "<tt>int X;</tt>" at global scope.
505 </dd>
506
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]507 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
508
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]509 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
510 that some targets may choose to emit different assembly sequences for them
511 for target-dependent reasons. This is used for globals that are declared
512 "weak" in C source code.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]513 </dd>
514
515 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
516
517 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
518 pointer to array type. When two global variables with appending linkage are
519 linked together, the two global arrays are appended together. This is the
520 LLVM, typesafe, equivalent of having the system linker append together
521 "sections" with identical names when .o files are linked.
522 </dd>
523
524 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]525 <dd>The semantics of this linkage follow the ELF object file model: the
526 symbol is weak until linked, if not linked, the symbol becomes null instead
527 of being an undefined reference.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]528 </dd>
529
530 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
531
532 <dd>If none of the above identifiers are used, the global is externally
533 visible, meaning that it participates in linkage and can be used to resolve
534 external symbol references.
535 </dd>
536</dl>
537
538 <p>
539 The next two types of linkage are targeted for Microsoft Windows platform
540 only. They are designed to support importing (exporting) symbols from (to)
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]541 DLLs (Dynamic Link Libraries).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]542 </p>
543
544 <dl>
545 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
546
547 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
548 or variable via a global pointer to a pointer that is set up by the DLL
549 exporting the symbol. On Microsoft Windows targets, the pointer name is
550 formed by combining <code>_imp__</code> and the function or variable name.
551 </dd>
552
553 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
554
555 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
556 pointer to a pointer in a DLL, so that it can be referenced with the
557 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
558 name is formed by combining <code>_imp__</code> and the function or variable
559 name.
560 </dd>
561
562</dl>
563
564<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
565variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
566variable and was linked with this one, one of the two would be renamed,
567preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
568external (i.e., lacking any linkage declarations), they are accessible
569outside of the current module.</p>
570<p>It is illegal for a function <i>declaration</i>
571to have any linkage type other than "externally visible", <tt>dllimport</tt>,
572or <tt>extern_weak</tt>.</p>
573<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
574linkages.
575</div>
576
577<!-- ======================================================================= -->
578<div class="doc_subsection">
579 <a name="callingconv">Calling Conventions</a>
580</div>
581
582<div class="doc_text">
583
584<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
585and <a href="#i_invoke">invokes</a> can all have an optional calling convention
586specified for the call. The calling convention of any pair of dynamic
587caller/callee must match, or the behavior of the program is undefined. The
588following calling conventions are supported by LLVM, and more may be added in
589the future:</p>
590
591<dl>
592 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
593
594 <dd>This calling convention (the default if no other calling convention is
595 specified) matches the target C calling conventions. This calling convention
596 supports varargs function calls and tolerates some mismatch in the declared
597 prototype and implemented declaration of the function (as does normal C).
598 </dd>
599
600 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
601
602 <dd>This calling convention attempts to make calls as fast as possible
603 (e.g. by passing things in registers). This calling convention allows the
604 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]605 without having to conform to an externally specified ABI (Application Binary
606 Interface). Implementations of this convention should allow arbitrary
Arnold Schwaighofer07444922008-05-14 09:17:12 +0000[diff] [blame]607 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
608 supported. This calling convention does not support varargs and requires the
609 prototype of all callees to exactly match the prototype of the function
610 definition.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]611 </dd>
612
613 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
614
615 <dd>This calling convention attempts to make code in the caller as efficient
616 as possible under the assumption that the call is not commonly executed. As
617 such, these calls often preserve all registers so that the call does not break
618 any live ranges in the caller side. This calling convention does not support
619 varargs and requires the prototype of all callees to exactly match the
620 prototype of the function definition.
621 </dd>
622
623 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
624
625 <dd>Any calling convention may be specified by number, allowing
626 target-specific calling conventions to be used. Target specific calling
627 conventions start at 64.
628 </dd>
629</dl>
630
631<p>More calling conventions can be added/defined on an as-needed basis, to
632support pascal conventions or any other well-known target-independent
633convention.</p>
634
635</div>
636
637<!-- ======================================================================= -->
638<div class="doc_subsection">
639 <a name="visibility">Visibility Styles</a>
640</div>
641
642<div class="doc_text">
643
644<p>
645All Global Variables and Functions have one of the following visibility styles:
646</p>
647
648<dl>
649 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
650
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]651 <dd>On targets that use the ELF object file format, default visibility means
652 that the declaration is visible to other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]653 modules and, in shared libraries, means that the declared entity may be
654 overridden. On Darwin, default visibility means that the declaration is
655 visible to other modules. Default visibility corresponds to "external
656 linkage" in the language.
657 </dd>
658
659 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
660
661 <dd>Two declarations of an object with hidden visibility refer to the same
662 object if they are in the same shared object. Usually, hidden visibility
663 indicates that the symbol will not be placed into the dynamic symbol table,
664 so no other module (executable or shared library) can reference it
665 directly.
666 </dd>
667
668 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
669
670 <dd>On ELF, protected visibility indicates that the symbol will be placed in
671 the dynamic symbol table, but that references within the defining module will
672 bind to the local symbol. That is, the symbol cannot be overridden by another
673 module.
674 </dd>
675</dl>
676
677</div>
678
679<!-- ======================================================================= -->
680<div class="doc_subsection">
681 <a name="globalvars">Global Variables</a>
682</div>
683
684<div class="doc_text">
685
686<p>Global variables define regions of memory allocated at compilation time
687instead of run-time. Global variables may optionally be initialized, may have
688an explicit section to be placed in, and may have an optional explicit alignment
689specified. A variable may be defined as "thread_local", which means that it
690will not be shared by threads (each thread will have a separated copy of the
691variable). A variable may be defined as a global "constant," which indicates
692that the contents of the variable will <b>never</b> be modified (enabling better
693optimization, allowing the global data to be placed in the read-only section of
694an executable, etc). Note that variables that need runtime initialization
695cannot be marked "constant" as there is a store to the variable.</p>
696
697<p>
698LLVM explicitly allows <em>declarations</em> of global variables to be marked
699constant, even if the final definition of the global is not. This capability
700can be used to enable slightly better optimization of the program, but requires
701the language definition to guarantee that optimizations based on the
702'constantness' are valid for the translation units that do not include the
703definition.
704</p>
705
706<p>As SSA values, global variables define pointer values that are in
707scope (i.e. they dominate) all basic blocks in the program. Global
708variables always define a pointer to their "content" type because they
709describe a region of memory, and all memory objects in LLVM are
710accessed through pointers.</p>
711
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]712<p>A global variable may be declared to reside in a target-specifc numbered
713address space. For targets that support them, address spaces may affect how
714optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]715the variable. The default address space is zero. The address space qualifier
716must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]717
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]718<p>LLVM allows an explicit section to be specified for globals. If the target
719supports it, it will emit globals to the section specified.</p>
720
721<p>An explicit alignment may be specified for a global. If not present, or if
722the alignment is set to zero, the alignment of the global is set by the target
723to whatever it feels convenient. If an explicit alignment is specified, the
724global is forced to have at least that much alignment. All alignments must be
725a power of 2.</p>
726
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]727<p>For example, the following defines a global in a numbered address space with
728an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]729
730<div class="doc_code">
731<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]732@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]733</pre>
734</div>
735
736</div>
737
738
739<!-- ======================================================================= -->
740<div class="doc_subsection">
741 <a name="functionstructure">Functions</a>
742</div>
743
744<div class="doc_text">
745
746<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
747an optional <a href="#linkage">linkage type</a>, an optional
748<a href="#visibility">visibility style</a>, an optional
749<a href="#callingconv">calling convention</a>, a return type, an optional
750<a href="#paramattrs">parameter attribute</a> for the return type, a function
751name, a (possibly empty) argument list (each with optional
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]752<a href="#paramattrs">parameter attributes</a>), optional
753<a href="#fnattrs">function attributes</a>, an optional section,
754an optional alignment, an optional <a href="#gc">garbage collector name</a>,
Chris Lattneree202542008-10-04 18:10:21 +0000[diff] [blame]755an opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]756
757LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
758optional <a href="#linkage">linkage type</a>, an optional
759<a href="#visibility">visibility style</a>, an optional
760<a href="#callingconv">calling convention</a>, a return type, an optional
761<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]762name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]763<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]764
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]765<p>A function definition contains a list of basic blocks, forming the CFG
766(Control Flow Graph) for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]767the function. Each basic block may optionally start with a label (giving the
768basic block a symbol table entry), contains a list of instructions, and ends
769with a <a href="#terminators">terminator</a> instruction (such as a branch or
770function return).</p>
771
772<p>The first basic block in a function is special in two ways: it is immediately
773executed on entrance to the function, and it is not allowed to have predecessor
774basic blocks (i.e. there can not be any branches to the entry block of a
775function). Because the block can have no predecessors, it also cannot have any
776<a href="#i_phi">PHI nodes</a>.</p>
777
778<p>LLVM allows an explicit section to be specified for functions. If the target
779supports it, it will emit functions to the section specified.</p>
780
781<p>An explicit alignment may be specified for a function. If not present, or if
782the alignment is set to zero, the alignment of the function is set by the target
783to whatever it feels convenient. If an explicit alignment is specified, the
784function is forced to have at least that much alignment. All alignments must be
785a power of 2.</p>
786
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]787 <h5>Syntax:</h5>
788
789<div class="doc_code">
790<pre>
791define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>] [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list]) [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N] [<a href="#gc">gc</a>] { ... }
792</pre>
793</div>
794
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]795</div>
796
797
798<!-- ======================================================================= -->
799<div class="doc_subsection">
800 <a name="aliasstructure">Aliases</a>
801</div>
802<div class="doc_text">
803 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]804 function, global variable, another alias or bitcast of global value). Aliases
805 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]806 optional <a href="#visibility">visibility style</a>.</p>
807
808 <h5>Syntax:</h5>
809
810<div class="doc_code">
811<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000[diff] [blame]812@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]813</pre>
814</div>
815
816</div>
817
818
819
820<!-- ======================================================================= -->
821<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
822<div class="doc_text">
823 <p>The return type and each parameter of a function type may have a set of
824 <i>parameter attributes</i> associated with them. Parameter attributes are
825 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]826 a function. Parameter attributes are considered to be part of the function,
827 not of the function type, so functions with different parameter attributes
828 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]829
830 <p>Parameter attributes are simple keywords that follow the type specified. If
831 multiple parameter attributes are needed, they are space separated. For
832 example:</p>
833
834<div class="doc_code">
835<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]836declare i32 @printf(i8* noalias , ...)
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]837declare i32 @atoi(i8 zeroext)
838declare signext i8 @returns_signed_char()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]839</pre>
840</div>
841
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]842 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
843 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]844
845 <p>Currently, only the following parameter attributes are defined:</p>
846 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]847 <dt><tt>zeroext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]848 <dd>This indicates to the code generator that the parameter or return value
849 should be zero-extended to a 32-bit value by the caller (for a parameter)
850 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]851
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]852 <dt><tt>signext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]853 <dd>This indicates to the code generator that the parameter or return value
854 should be sign-extended to a 32-bit value by the caller (for a parameter)
855 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]856
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]857 <dt><tt>inreg</tt></dt>
Dale Johannesenc08a0e22008-09-25 20:47:45 +0000[diff] [blame]858 <dd>This indicates that this parameter or return value should be treated
859 in a special target-dependent fashion during while emitting code for a
860 function call or return (usually, by putting it in a register as opposed
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]861 to memory, though some targets use it to distinguish between two different
862 kinds of registers). Use of this attribute is target-specific.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]863
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]864 <dt><tt><a name="byval">byval</a></tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]865 <dd>This indicates that the pointer parameter should really be passed by
866 value to the function. The attribute implies that a hidden copy of the
867 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]868 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]869 pointer arguments. It is generally used to pass structs and arrays by
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]870 value, but is also valid on pointers to scalars. The copy is considered to
871 belong to the caller not the callee (for example,
872 <tt><a href="#readonly">readonly</a></tt> functions should not write to
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]873 <tt>byval</tt> parameters). This is not a valid attribute for return
874 values. </dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]875
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]876 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]877 <dd>This indicates that the pointer parameter specifies the address of a
878 structure that is the return value of the function in the source program.
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]879 This pointer must be guaranteed by the caller to be valid: loads and stores
880 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]881 be applied to the first parameter. This is not a valid attribute for
882 return values. </dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]883
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]884 <dt><tt>noalias</tt></dt>
Owen Andersonc4fc4cd2008-02-18 04:09:01 +0000[diff] [blame]885 <dd>This indicates that the parameter does not alias any global or any other
886 parameter. The caller is responsible for ensuring that this is the case,
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]887 usually by placing the value in a stack allocation. This is not a valid
888 attribute for return values.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]889
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]890 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]891 <dd>This indicates that the pointer parameter can be excised using the
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]892 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
893 attribute for return values.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]894 </dl>
895
896</div>
897
898<!-- ======================================================================= -->
899<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]900 <a name="gc">Garbage Collector Names</a>
901</div>
902
903<div class="doc_text">
904<p>Each function may specify a garbage collector name, which is simply a
905string.</p>
906
907<div class="doc_code"><pre
908>define void @f() gc "name" { ...</pre></div>
909
910<p>The compiler declares the supported values of <i>name</i>. Specifying a
911collector which will cause the compiler to alter its output in order to support
912the named garbage collection algorithm.</p>
913</div>
914
915<!-- ======================================================================= -->
916<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]917 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]918</div>
919
920<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]921
922<p>Function attributes are set to communicate additional information about
923 a function. Function attributes are considered to be part of the function,
924 not of the function type, so functions with different parameter attributes
925 can have the same function type.</p>
926
927 <p>Function attributes are simple keywords that follow the type specified. If
928 multiple attributes are needed, they are space separated. For
929 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]930
931<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]932<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]933define void @f() noinline { ... }
934define void @f() alwaysinline { ... }
935define void @f() alwaysinline optsize { ... }
936define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]937</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]938</div>
939
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]940<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]941<dt><tt>alwaysinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]942<dd>This attribute indicates that the inliner should attempt to inline this
943function into callers whenever possible, ignoring any active inlining size
944threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]945
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]946<dt><tt>noinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]947<dd>This attribute indicates that the inliner should never inline this function
Chris Lattner377f3ae2008-10-05 17:14:59 +0000[diff] [blame]948in any situation. This attribute may not be used together with the
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]949<tt>alwaysinline</tt> attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]950
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]951<dt><tt>optsize</tt></dt>
Devang Patelc29099b2008-09-29 18:34:44 +0000[diff] [blame]952<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]953make choices that keep the code size of this function low, and otherwise do
954optimizations specifically to reduce code size.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]955
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]956<dt><tt>noreturn</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]957<dd>This function attribute indicates that the function never returns normally.
958This produces undefined behavior at runtime if the function ever does
959dynamically return.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]960
961<dt><tt>nounwind</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]962<dd>This function attribute indicates that the function never returns with an
963unwind or exceptional control flow. If the function does unwind, its runtime
964behavior is undefined.</dd>
965
966<dt><tt>readnone</tt></dt>
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]967<dd>This attribute indicates that the function computes its result (or the
968exception it throws) based strictly on its arguments, without dereferencing any
969pointer arguments or otherwise accessing any mutable state (e.g. memory, control
970registers, etc) visible to caller functions. It does not write through any
971pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
972never changes any state visible to callers.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]973
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]974<dt><tt><a name="readonly">readonly</a></tt></dt>
975<dd>This attribute indicates that the function does not write through any
976pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
977or otherwise modify any state (e.g. memory, control registers, etc) visible to
978caller functions. It may dereference pointer arguments and read state that may
979be set in the caller. A readonly function always returns the same value (or
980throws the same exception) when called with the same set of arguments and global
981state.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]982</dl>
983
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]984</div>
985
986<!-- ======================================================================= -->
987<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]988 <a name="moduleasm">Module-Level Inline Assembly</a>
989</div>
990
991<div class="doc_text">
992<p>
993Modules may contain "module-level inline asm" blocks, which corresponds to the
994GCC "file scope inline asm" blocks. These blocks are internally concatenated by
995LLVM and treated as a single unit, but may be separated in the .ll file if
996desired. The syntax is very simple:
997</p>
998
999<div class="doc_code">
1000<pre>
1001module asm "inline asm code goes here"
1002module asm "more can go here"
1003</pre>
1004</div>
1005
1006<p>The strings can contain any character by escaping non-printable characters.
1007 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1008 for the number.
1009</p>
1010
1011<p>
1012 The inline asm code is simply printed to the machine code .s file when
1013 assembly code is generated.
1014</p>
1015</div>
1016
1017<!-- ======================================================================= -->
1018<div class="doc_subsection">
1019 <a name="datalayout">Data Layout</a>
1020</div>
1021
1022<div class="doc_text">
1023<p>A module may specify a target specific data layout string that specifies how
1024data is to be laid out in memory. The syntax for the data layout is simply:</p>
1025<pre> target datalayout = "<i>layout specification</i>"</pre>
1026<p>The <i>layout specification</i> consists of a list of specifications
1027separated by the minus sign character ('-'). Each specification starts with a
1028letter and may include other information after the letter to define some
1029aspect of the data layout. The specifications accepted are as follows: </p>
1030<dl>
1031 <dt><tt>E</tt></dt>
1032 <dd>Specifies that the target lays out data in big-endian form. That is, the
1033 bits with the most significance have the lowest address location.</dd>
1034 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1035 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1036 the bits with the least significance have the lowest address location.</dd>
1037 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1038 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1039 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1040 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1041 too.</dd>
1042 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1043 <dd>This specifies the alignment for an integer type of a given bit
1044 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1045 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1046 <dd>This specifies the alignment for a vector type of a given bit
1047 <i>size</i>.</dd>
1048 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1049 <dd>This specifies the alignment for a floating point type of a given bit
1050 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1051 (double).</dd>
1052 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1053 <dd>This specifies the alignment for an aggregate type of a given bit
1054 <i>size</i>.</dd>
1055</dl>
1056<p>When constructing the data layout for a given target, LLVM starts with a
1057default set of specifications which are then (possibly) overriden by the
1058specifications in the <tt>datalayout</tt> keyword. The default specifications
1059are given in this list:</p>
1060<ul>
1061 <li><tt>E</tt> - big endian</li>
1062 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1063 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1064 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1065 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1066 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1067 <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]1068 alignment of 64-bits</li>
1069 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1070 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1071 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1072 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1073 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1074</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1075<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1076following rules:
1077<ol>
1078 <li>If the type sought is an exact match for one of the specifications, that
1079 specification is used.</li>
1080 <li>If no match is found, and the type sought is an integer type, then the
1081 smallest integer type that is larger than the bitwidth of the sought type is
1082 used. If none of the specifications are larger than the bitwidth then the the
1083 largest integer type is used. For example, given the default specifications
1084 above, the i7 type will use the alignment of i8 (next largest) while both
1085 i65 and i256 will use the alignment of i64 (largest specified).</li>
1086 <li>If no match is found, and the type sought is a vector type, then the
1087 largest vector type that is smaller than the sought vector type will be used
1088 as a fall back. This happens because <128 x double> can be implemented in
1089 terms of 64 <2 x double>, for example.</li>
1090</ol>
1091</div>
1092
1093<!-- *********************************************************************** -->
1094<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1095<!-- *********************************************************************** -->
1096
1097<div class="doc_text">
1098
1099<p>The LLVM type system is one of the most important features of the
1100intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1101optimizations to be performed on the intermediate representation directly,
1102without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1103extra analyses on the side before the transformation. A strong type
1104system makes it easier to read the generated code and enables novel
1105analyses and transformations that are not feasible to perform on normal
1106three address code representations.</p>
1107
1108</div>
1109
1110<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1111<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1112Classifications</a> </div>
1113<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1114<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1115classifications:</p>
1116
1117<table border="1" cellspacing="0" cellpadding="4">
1118 <tbody>
1119 <tr><th>Classification</th><th>Types</th></tr>
1120 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1121 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1122 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1123 </tr>
1124 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1125 <td><a href="#t_floating">floating point</a></td>
1126 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1127 </tr>
1128 <tr>
1129 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1130 <td><a href="#t_integer">integer</a>,
1131 <a href="#t_floating">floating point</a>,
1132 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1133 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1134 <a href="#t_struct">structure</a>,
1135 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1136 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1137 </td>
1138 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1139 <tr>
1140 <td><a href="#t_primitive">primitive</a></td>
1141 <td><a href="#t_label">label</a>,
1142 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1143 <a href="#t_floating">floating point</a>.</td>
1144 </tr>
1145 <tr>
1146 <td><a href="#t_derived">derived</a></td>
1147 <td><a href="#t_integer">integer</a>,
1148 <a href="#t_array">array</a>,
1149 <a href="#t_function">function</a>,
1150 <a href="#t_pointer">pointer</a>,
1151 <a href="#t_struct">structure</a>,
1152 <a href="#t_pstruct">packed structure</a>,
1153 <a href="#t_vector">vector</a>,
1154 <a href="#t_opaque">opaque</a>.
1155 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1156 </tbody>
1157</table>
1158
1159<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1160most important. Values of these types are the only ones which can be
1161produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1162instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1163</div>
1164
1165<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1166<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1167
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1168<div class="doc_text">
1169<p>The primitive types are the fundamental building blocks of the LLVM
1170system.</p>
1171
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1172</div>
1173
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1174<!-- _______________________________________________________________________ -->
1175<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1176
1177<div class="doc_text">
1178 <table>
1179 <tbody>
1180 <tr><th>Type</th><th>Description</th></tr>
1181 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1182 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1183 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1184 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1185 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1186 </tbody>
1187 </table>
1188</div>
1189
1190<!-- _______________________________________________________________________ -->
1191<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1192
1193<div class="doc_text">
1194<h5>Overview:</h5>
1195<p>The void type does not represent any value and has no size.</p>
1196
1197<h5>Syntax:</h5>
1198
1199<pre>
1200 void
1201</pre>
1202</div>
1203
1204<!-- _______________________________________________________________________ -->
1205<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1206
1207<div class="doc_text">
1208<h5>Overview:</h5>
1209<p>The label type represents code labels.</p>
1210
1211<h5>Syntax:</h5>
1212
1213<pre>
1214 label
1215</pre>
1216</div>
1217
1218
1219<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1220<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1221
1222<div class="doc_text">
1223
1224<p>The real power in LLVM comes from the derived types in the system.
1225This is what allows a programmer to represent arrays, functions,
1226pointers, and other useful types. Note that these derived types may be
1227recursive: For example, it is possible to have a two dimensional array.</p>
1228
1229</div>
1230
1231<!-- _______________________________________________________________________ -->
1232<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1233
1234<div class="doc_text">
1235
1236<h5>Overview:</h5>
1237<p>The integer type is a very simple derived type that simply specifies an
1238arbitrary bit width for the integer type desired. Any bit width from 1 bit to
12392^23-1 (about 8 million) can be specified.</p>
1240
1241<h5>Syntax:</h5>
1242
1243<pre>
1244 iN
1245</pre>
1246
1247<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1248value.</p>
1249
1250<h5>Examples:</h5>
1251<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1252 <tbody>
1253 <tr>
1254 <td><tt>i1</tt></td>
1255 <td>a single-bit integer.</td>
1256 </tr><tr>
1257 <td><tt>i32</tt></td>
1258 <td>a 32-bit integer.</td>
1259 </tr><tr>
1260 <td><tt>i1942652</tt></td>
1261 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1262 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1263 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1264</table>
1265</div>
1266
1267<!-- _______________________________________________________________________ -->
1268<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1269
1270<div class="doc_text">
1271
1272<h5>Overview:</h5>
1273
1274<p>The array type is a very simple derived type that arranges elements
1275sequentially in memory. The array type requires a size (number of
1276elements) and an underlying data type.</p>
1277
1278<h5>Syntax:</h5>
1279
1280<pre>
1281 [&lt;# elements&gt; x &lt;elementtype&gt;]
1282</pre>
1283
1284<p>The number of elements is a constant integer value; elementtype may
1285be any type with a size.</p>
1286
1287<h5>Examples:</h5>
1288<table class="layout">
1289 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1290 <td class="left"><tt>[40 x i32]</tt></td>
1291 <td class="left">Array of 40 32-bit integer values.</td>
1292 </tr>
1293 <tr class="layout">
1294 <td class="left"><tt>[41 x i32]</tt></td>
1295 <td class="left">Array of 41 32-bit integer values.</td>
1296 </tr>
1297 <tr class="layout">
1298 <td class="left"><tt>[4 x i8]</tt></td>
1299 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1300 </tr>
1301</table>
1302<p>Here are some examples of multidimensional arrays:</p>
1303<table class="layout">
1304 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1305 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1306 <td class="left">3x4 array of 32-bit integer values.</td>
1307 </tr>
1308 <tr class="layout">
1309 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1310 <td class="left">12x10 array of single precision floating point values.</td>
1311 </tr>
1312 <tr class="layout">
1313 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1314 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1315 </tr>
1316</table>
1317
1318<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1319length array. Normally, accesses past the end of an array are undefined in
1320LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1321As a special case, however, zero length arrays are recognized to be variable
1322length. This allows implementation of 'pascal style arrays' with the LLVM
1323type "{ i32, [0 x float]}", for example.</p>
1324
1325</div>
1326
1327<!-- _______________________________________________________________________ -->
1328<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1329<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1330
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1331<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1332
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1333<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1334consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1335return 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]1336If 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]1337class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1338
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1339<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1340
1341<pre>
1342 &lt;returntype list&gt; (&lt;parameter list&gt;)
1343</pre>
1344
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1345<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1346specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1347which indicates that the function takes a variable number of arguments.
1348Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1349 href="#int_varargs">variable argument handling intrinsic</a> functions.
1350'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1351<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1352
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1353<h5>Examples:</h5>
1354<table class="layout">
1355 <tr class="layout">
1356 <td class="left"><tt>i32 (i32)</tt></td>
1357 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1358 </td>
1359 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1360 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1361 </tt></td>
1362 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1363 an <tt>i16</tt> that should be sign extended and a
1364 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1365 <tt>float</tt>.
1366 </td>
1367 </tr><tr class="layout">
1368 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1369 <td class="left">A vararg function that takes at least one
1370 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1371 which returns an integer. This is the signature for <tt>printf</tt> in
1372 LLVM.
1373 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1374 </tr><tr class="layout">
1375 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Devang Patel6dddba22008-03-24 18:10:52 +0000[diff] [blame]1376 <td class="left">A function taking an <tt>i32></tt>, returning two
1377 <tt> i32 </tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1378 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1379 </tr>
1380</table>
1381
1382</div>
1383<!-- _______________________________________________________________________ -->
1384<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1385<div class="doc_text">
1386<h5>Overview:</h5>
1387<p>The structure type is used to represent a collection of data members
1388together in memory. The packing of the field types is defined to match
1389the ABI of the underlying processor. The elements of a structure may
1390be any type that has a size.</p>
1391<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1392and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1393field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1394instruction.</p>
1395<h5>Syntax:</h5>
1396<pre> { &lt;type list&gt; }<br></pre>
1397<h5>Examples:</h5>
1398<table class="layout">
1399 <tr class="layout">
1400 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1401 <td class="left">A triple of three <tt>i32</tt> values</td>
1402 </tr><tr class="layout">
1403 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1404 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1405 second element is a <a href="#t_pointer">pointer</a> to a
1406 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1407 an <tt>i32</tt>.</td>
1408 </tr>
1409</table>
1410</div>
1411
1412<!-- _______________________________________________________________________ -->
1413<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1414</div>
1415<div class="doc_text">
1416<h5>Overview:</h5>
1417<p>The packed structure type is used to represent a collection of data members
1418together in memory. There is no padding between fields. Further, the alignment
1419of a packed structure is 1 byte. The elements of a packed structure may
1420be any type that has a size.</p>
1421<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1422and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1423field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1424instruction.</p>
1425<h5>Syntax:</h5>
1426<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1427<h5>Examples:</h5>
1428<table class="layout">
1429 <tr class="layout">
1430 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1431 <td class="left">A triple of three <tt>i32</tt> values</td>
1432 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1433 <td class="left">
1434<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1435 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1436 second element is a <a href="#t_pointer">pointer</a> to a
1437 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1438 an <tt>i32</tt>.</td>
1439 </tr>
1440</table>
1441</div>
1442
1443<!-- _______________________________________________________________________ -->
1444<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1445<div class="doc_text">
1446<h5>Overview:</h5>
1447<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1448reference to another object, which must live in memory. Pointer types may have
1449an optional address space attribute defining the target-specific numbered
1450address space where the pointed-to object resides. The default address space is
1451zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1452<h5>Syntax:</h5>
1453<pre> &lt;type&gt; *<br></pre>
1454<h5>Examples:</h5>
1455<table class="layout">
1456 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1457 <td class="left"><tt>[4x i32]*</tt></td>
1458 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1459 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1460 </tr>
1461 <tr class="layout">
1462 <td class="left"><tt>i32 (i32 *) *</tt></td>
1463 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1464 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1465 <tt>i32</tt>.</td>
1466 </tr>
1467 <tr class="layout">
1468 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1469 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1470 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1471 </tr>
1472</table>
1473</div>
1474
1475<!-- _______________________________________________________________________ -->
1476<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1477<div class="doc_text">
1478
1479<h5>Overview:</h5>
1480
1481<p>A vector type is a simple derived type that represents a vector
1482of elements. Vector types are used when multiple primitive data
1483are operated in parallel using a single instruction (SIMD).
1484A vector type requires a size (number of
1485elements) and an underlying primitive data type. Vectors must have a power
1486of two length (1, 2, 4, 8, 16 ...). Vector types are
1487considered <a href="#t_firstclass">first class</a>.</p>
1488
1489<h5>Syntax:</h5>
1490
1491<pre>
1492 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1493</pre>
1494
1495<p>The number of elements is a constant integer value; elementtype may
1496be any integer or floating point type.</p>
1497
1498<h5>Examples:</h5>
1499
1500<table class="layout">
1501 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1502 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1503 <td class="left">Vector of 4 32-bit integer values.</td>
1504 </tr>
1505 <tr class="layout">
1506 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1507 <td class="left">Vector of 8 32-bit floating-point values.</td>
1508 </tr>
1509 <tr class="layout">
1510 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1511 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1512 </tr>
1513</table>
1514</div>
1515
1516<!-- _______________________________________________________________________ -->
1517<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1518<div class="doc_text">
1519
1520<h5>Overview:</h5>
1521
1522<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1523corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1524In LLVM, opaque types can eventually be resolved to any type (not just a
1525structure type).</p>
1526
1527<h5>Syntax:</h5>
1528
1529<pre>
1530 opaque
1531</pre>
1532
1533<h5>Examples:</h5>
1534
1535<table class="layout">
1536 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1537 <td class="left"><tt>opaque</tt></td>
1538 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1539 </tr>
1540</table>
1541</div>
1542
1543
1544<!-- *********************************************************************** -->
1545<div class="doc_section"> <a name="constants">Constants</a> </div>
1546<!-- *********************************************************************** -->
1547
1548<div class="doc_text">
1549
1550<p>LLVM has several different basic types of constants. This section describes
1551them all and their syntax.</p>
1552
1553</div>
1554
1555<!-- ======================================================================= -->
1556<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1557
1558<div class="doc_text">
1559
1560<dl>
1561 <dt><b>Boolean constants</b></dt>
1562
1563 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1564 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1565 </dd>
1566
1567 <dt><b>Integer constants</b></dt>
1568
1569 <dd>Standard integers (such as '4') are constants of the <a
1570 href="#t_integer">integer</a> type. Negative numbers may be used with
1571 integer types.
1572 </dd>
1573
1574 <dt><b>Floating point constants</b></dt>
1575
1576 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1577 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1578 notation (see below). The assembler requires the exact decimal value of
1579 a floating-point constant. For example, the assembler accepts 1.25 but
1580 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1581 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1582
1583 <dt><b>Null pointer constants</b></dt>
1584
1585 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1586 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1587
1588</dl>
1589
1590<p>The one non-intuitive notation for constants is the optional hexadecimal form
1591of floating point constants. For example, the form '<tt>double
15920x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
15934.5e+15</tt>'. The only time hexadecimal floating point constants are required
1594(and the only time that they are generated by the disassembler) is when a
1595floating point constant must be emitted but it cannot be represented as a
1596decimal floating point number. For example, NaN's, infinities, and other
1597special values are represented in their IEEE hexadecimal format so that
1598assembly and disassembly do not cause any bits to change in the constants.</p>
1599
1600</div>
1601
1602<!-- ======================================================================= -->
1603<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1604</div>
1605
1606<div class="doc_text">
1607<p>Aggregate constants arise from aggregation of simple constants
1608and smaller aggregate constants.</p>
1609
1610<dl>
1611 <dt><b>Structure constants</b></dt>
1612
1613 <dd>Structure constants are represented with notation similar to structure
1614 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1615 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1616 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]1617 must have <a href="#t_struct">structure type</a>, and the number and
1618 types of elements must match those specified by the type.
1619 </dd>
1620
1621 <dt><b>Array constants</b></dt>
1622
1623 <dd>Array constants are represented with notation similar to array type
1624 definitions (a comma separated list of elements, surrounded by square brackets
1625 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1626 constants must have <a href="#t_array">array type</a>, and the number and
1627 types of elements must match those specified by the type.
1628 </dd>
1629
1630 <dt><b>Vector constants</b></dt>
1631
1632 <dd>Vector constants are represented with notation similar to vector type
1633 definitions (a comma separated list of elements, surrounded by
1634 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1635 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1636 href="#t_vector">vector type</a>, and the number and types of elements must
1637 match those specified by the type.
1638 </dd>
1639
1640 <dt><b>Zero initialization</b></dt>
1641
1642 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1643 value to zero of <em>any</em> type, including scalar and aggregate types.
1644 This is often used to avoid having to print large zero initializers (e.g. for
1645 large arrays) and is always exactly equivalent to using explicit zero
1646 initializers.
1647 </dd>
1648</dl>
1649
1650</div>
1651
1652<!-- ======================================================================= -->
1653<div class="doc_subsection">
1654 <a name="globalconstants">Global Variable and Function Addresses</a>
1655</div>
1656
1657<div class="doc_text">
1658
1659<p>The addresses of <a href="#globalvars">global variables</a> and <a
1660href="#functionstructure">functions</a> are always implicitly valid (link-time)
1661constants. These constants are explicitly referenced when the <a
1662href="#identifiers">identifier for the global</a> is used and always have <a
1663href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1664file:</p>
1665
1666<div class="doc_code">
1667<pre>
1668@X = global i32 17
1669@Y = global i32 42
1670@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1671</pre>
1672</div>
1673
1674</div>
1675
1676<!-- ======================================================================= -->
1677<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1678<div class="doc_text">
1679 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1680 no specific value. Undefined values may be of any type and be used anywhere
1681 a constant is permitted.</p>
1682
1683 <p>Undefined values indicate to the compiler that the program is well defined
1684 no matter what value is used, giving the compiler more freedom to optimize.
1685 </p>
1686</div>
1687
1688<!-- ======================================================================= -->
1689<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1690</div>
1691
1692<div class="doc_text">
1693
1694<p>Constant expressions are used to allow expressions involving other constants
1695to be used as constants. Constant expressions may be of any <a
1696href="#t_firstclass">first class</a> type and may involve any LLVM operation
1697that does not have side effects (e.g. load and call are not supported). The
1698following is the syntax for constant expressions:</p>
1699
1700<dl>
1701 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1702 <dd>Truncate a constant to another type. The bit size of CST must be larger
1703 than the bit size of TYPE. Both types must be integers.</dd>
1704
1705 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1706 <dd>Zero extend a constant to another type. The bit size of CST must be
1707 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1708
1709 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1710 <dd>Sign extend a constant to another type. The bit size of CST must be
1711 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1712
1713 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1714 <dd>Truncate a floating point constant to another floating point type. The
1715 size of CST must be larger than the size of TYPE. Both types must be
1716 floating point.</dd>
1717
1718 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1719 <dd>Floating point extend a constant to another type. The size of CST must be
1720 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1721
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1722 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1723 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1724 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1725 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1726 of the same number of elements. If the value won't fit in the integer type,
1727 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1728
1729 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1730 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1731 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1732 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1733 of the same number of elements. If the value won't fit in the integer type,
1734 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1735
1736 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1737 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1738 constant. TYPE must be a scalar or vector floating point type. CST must be of
1739 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1740 of the same number of elements. If the value won't fit in the floating point
1741 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1742
1743 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1744 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1745 constant. TYPE must be a scalar or vector floating point type. CST must be of
1746 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1747 of the same number of elements. If the value won't fit in the floating point
1748 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1749
1750 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1751 <dd>Convert a pointer typed constant to the corresponding integer constant
1752 TYPE must be an integer type. CST must be of pointer type. The CST value is
1753 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1754
1755 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1756 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1757 pointer type. CST must be of integer type. The CST value is zero extended,
1758 truncated, or unchanged to make it fit in a pointer size. This one is
1759 <i>really</i> dangerous!</dd>
1760
1761 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1762 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1763 identical (same number of bits). The conversion is done as if the CST value
1764 was stored to memory and read back as TYPE. In other words, no bits change
1765 with this operator, just the type. This can be used for conversion of
1766 vector types to any other type, as long as they have the same bit width. For
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]1767 pointers it is only valid to cast to another pointer type. It is not valid
1768 to bitcast to or from an aggregate type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1769 </dd>
1770
1771 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1772
1773 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1774 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1775 instruction, the index list may have zero or more indexes, which are required
1776 to make sense for the type of "CSTPTR".</dd>
1777
1778 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1779
1780 <dd>Perform the <a href="#i_select">select operation</a> on
1781 constants.</dd>
1782
1783 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1784 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1785
1786 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1787 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1788
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1789 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1790 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1791
1792 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1793 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1794
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1795 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1796
1797 <dd>Perform the <a href="#i_extractelement">extractelement
1798 operation</a> on constants.
1799
1800 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1801
1802 <dd>Perform the <a href="#i_insertelement">insertelement
1803 operation</a> on constants.</dd>
1804
1805
1806 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1807
1808 <dd>Perform the <a href="#i_shufflevector">shufflevector
1809 operation</a> on constants.</dd>
1810
1811 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1812
1813 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1814 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1815 binary</a> operations. The constraints on operands are the same as those for
1816 the corresponding instruction (e.g. no bitwise operations on floating point
1817 values are allowed).</dd>
1818</dl>
1819</div>
1820
1821<!-- *********************************************************************** -->
1822<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1823<!-- *********************************************************************** -->
1824
1825<!-- ======================================================================= -->
1826<div class="doc_subsection">
1827<a name="inlineasm">Inline Assembler Expressions</a>
1828</div>
1829
1830<div class="doc_text">
1831
1832<p>
1833LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1834Module-Level Inline Assembly</a>) through the use of a special value. This
1835value represents the inline assembler as a string (containing the instructions
1836to emit), a list of operand constraints (stored as a string), and a flag that
1837indicates whether or not the inline asm expression has side effects. An example
1838inline assembler expression is:
1839</p>
1840
1841<div class="doc_code">
1842<pre>
1843i32 (i32) asm "bswap $0", "=r,r"
1844</pre>
1845</div>
1846
1847<p>
1848Inline assembler expressions may <b>only</b> be used as the callee operand of
1849a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1850</p>
1851
1852<div class="doc_code">
1853<pre>
1854%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1855</pre>
1856</div>
1857
1858<p>
1859Inline asms with side effects not visible in the constraint list must be marked
1860as having side effects. This is done through the use of the
1861'<tt>sideeffect</tt>' keyword, like so:
1862</p>
1863
1864<div class="doc_code">
1865<pre>
1866call void asm sideeffect "eieio", ""()
1867</pre>
1868</div>
1869
1870<p>TODO: The format of the asm and constraints string still need to be
1871documented here. Constraints on what can be done (e.g. duplication, moving, etc
Chris Lattner1cdd64e2008-10-04 18:36:02 +0000[diff] [blame]1872need to be documented). This is probably best done by reference to another
1873document that covers inline asm from a holistic perspective.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1874</p>
1875
1876</div>
1877
1878<!-- *********************************************************************** -->
1879<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1880<!-- *********************************************************************** -->
1881
1882<div class="doc_text">
1883
1884<p>The LLVM instruction set consists of several different
1885classifications of instructions: <a href="#terminators">terminator
1886instructions</a>, <a href="#binaryops">binary instructions</a>,
1887<a href="#bitwiseops">bitwise binary instructions</a>, <a
1888 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1889instructions</a>.</p>
1890
1891</div>
1892
1893<!-- ======================================================================= -->
1894<div class="doc_subsection"> <a name="terminators">Terminator
1895Instructions</a> </div>
1896
1897<div class="doc_text">
1898
1899<p>As mentioned <a href="#functionstructure">previously</a>, every
1900basic block in a program ends with a "Terminator" instruction, which
1901indicates which block should be executed after the current block is
1902finished. These terminator instructions typically yield a '<tt>void</tt>'
1903value: they produce control flow, not values (the one exception being
1904the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1905<p>There are six different terminator instructions: the '<a
1906 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1907instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1908the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1909 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1910 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1911
1912</div>
1913
1914<!-- _______________________________________________________________________ -->
1915<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1916Instruction</a> </div>
1917<div class="doc_text">
1918<h5>Syntax:</h5>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1919<pre>
1920 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]1921 ret void <i>; Return from void function</i>
1922</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1923
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1924<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1925
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1926<p>The '<tt>ret</tt>' instruction is used to return control flow (and
1927optionally a value) from a function back to the caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1928<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1929returns a value and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1930control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1931
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1932<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1933
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1934<p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
1935the return value. The type of the return value must be a
1936'<a href="#t_firstclass">first class</a>' type.</p>
1937
1938<p>A function is not <a href="#wellformed">well formed</a> if
1939it it has a non-void return type and contains a '<tt>ret</tt>'
1940instruction with no return value or a return value with a type that
1941does not match its type, or if it has a void return type and contains
1942a '<tt>ret</tt>' instruction with a return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1943
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1944<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1945
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1946<p>When the '<tt>ret</tt>' instruction is executed, control flow
1947returns back to the calling function's context. If the caller is a "<a
1948 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1949the instruction after the call. If the caller was an "<a
1950 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1951at the beginning of the "normal" destination block. If the instruction
1952returns a value, that value shall set the call or invoke instruction's
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1953return value.
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1954
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1955<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1956
1957<pre>
1958 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1959 ret void <i>; Return from a void function</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1960 ret { i32, i8 } { i32 4, i8 2 } <i>; Return an aggregate of values 4 and 2</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1961</pre>
1962</div>
1963<!-- _______________________________________________________________________ -->
1964<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1965<div class="doc_text">
1966<h5>Syntax:</h5>
1967<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1968</pre>
1969<h5>Overview:</h5>
1970<p>The '<tt>br</tt>' instruction is used to cause control flow to
1971transfer to a different basic block in the current function. There are
1972two forms of this instruction, corresponding to a conditional branch
1973and an unconditional branch.</p>
1974<h5>Arguments:</h5>
1975<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1976single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1977unconditional form of the '<tt>br</tt>' instruction takes a single
1978'<tt>label</tt>' value as a target.</p>
1979<h5>Semantics:</h5>
1980<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1981argument is evaluated. If the value is <tt>true</tt>, control flows
1982to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1983control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1984<h5>Example:</h5>
1985<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
1986 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1987</div>
1988<!-- _______________________________________________________________________ -->
1989<div class="doc_subsubsection">
1990 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1991</div>
1992
1993<div class="doc_text">
1994<h5>Syntax:</h5>
1995
1996<pre>
1997 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1998</pre>
1999
2000<h5>Overview:</h5>
2001
2002<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2003several different places. It is a generalization of the '<tt>br</tt>'
2004instruction, allowing a branch to occur to one of many possible
2005destinations.</p>
2006
2007
2008<h5>Arguments:</h5>
2009
2010<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2011comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2012an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2013table is not allowed to contain duplicate constant entries.</p>
2014
2015<h5>Semantics:</h5>
2016
2017<p>The <tt>switch</tt> instruction specifies a table of values and
2018destinations. When the '<tt>switch</tt>' instruction is executed, this
2019table is searched for the given value. If the value is found, control flow is
2020transfered to the corresponding destination; otherwise, control flow is
2021transfered to the default destination.</p>
2022
2023<h5>Implementation:</h5>
2024
2025<p>Depending on properties of the target machine and the particular
2026<tt>switch</tt> instruction, this instruction may be code generated in different
2027ways. For example, it could be generated as a series of chained conditional
2028branches or with a lookup table.</p>
2029
2030<h5>Example:</h5>
2031
2032<pre>
2033 <i>; Emulate a conditional br instruction</i>
2034 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2035 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
2036
2037 <i>; Emulate an unconditional br instruction</i>
2038 switch i32 0, label %dest [ ]
2039
2040 <i>; Implement a jump table:</i>
2041 switch i32 %val, label %otherwise [ i32 0, label %onzero
2042 i32 1, label %onone
2043 i32 2, label %ontwo ]
2044</pre>
2045</div>
2046
2047<!-- _______________________________________________________________________ -->
2048<div class="doc_subsubsection">
2049 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2050</div>
2051
2052<div class="doc_text">
2053
2054<h5>Syntax:</h5>
2055
2056<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2057 &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]2058 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2059</pre>
2060
2061<h5>Overview:</h5>
2062
2063<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2064function, with the possibility of control flow transfer to either the
2065'<tt>normal</tt>' label or the
2066'<tt>exception</tt>' label. If the callee function returns with the
2067"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2068"normal" label. If the callee (or any indirect callees) returns with the "<a
2069href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2070continued at the dynamically nearest "exception" label.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2071
2072<h5>Arguments:</h5>
2073
2074<p>This instruction requires several arguments:</p>
2075
2076<ol>
2077 <li>
2078 The optional "cconv" marker indicates which <a href="#callingconv">calling
2079 convention</a> the call should use. If none is specified, the call defaults
2080 to using C calling conventions.
2081 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2082
2083 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2084 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2085 and '<tt>inreg</tt>' attributes are valid here.</li>
2086
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2087 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2088 function value being invoked. In most cases, this is a direct function
2089 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2090 an arbitrary pointer to function value.
2091 </li>
2092
2093 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2094 function to be invoked. </li>
2095
2096 <li>'<tt>function args</tt>': argument list whose types match the function
2097 signature argument types. If the function signature indicates the function
2098 accepts a variable number of arguments, the extra arguments can be
2099 specified. </li>
2100
2101 <li>'<tt>normal label</tt>': the label reached when the called function
2102 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2103
2104 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2105 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2106
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2107 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2108 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2109 '<tt>readnone</tt>' attributes are valid here.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2110</ol>
2111
2112<h5>Semantics:</h5>
2113
2114<p>This instruction is designed to operate as a standard '<tt><a
2115href="#i_call">call</a></tt>' instruction in most regards. The primary
2116difference is that it establishes an association with a label, which is used by
2117the runtime library to unwind the stack.</p>
2118
2119<p>This instruction is used in languages with destructors to ensure that proper
2120cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2121exception. Additionally, this is important for implementation of
2122'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2123
2124<h5>Example:</h5>
2125<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2126 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2127 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2128 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2129 unwind label %TestCleanup <i>; {i32}:retval set</i>
2130</pre>
2131</div>
2132
2133
2134<!-- _______________________________________________________________________ -->
2135
2136<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2137Instruction</a> </div>
2138
2139<div class="doc_text">
2140
2141<h5>Syntax:</h5>
2142<pre>
2143 unwind
2144</pre>
2145
2146<h5>Overview:</h5>
2147
2148<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2149at the first callee in the dynamic call stack which used an <a
2150href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2151primarily used to implement exception handling.</p>
2152
2153<h5>Semantics:</h5>
2154
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2155<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2156immediately halt. The dynamic call stack is then searched for the first <a
2157href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2158execution continues at the "exceptional" destination block specified by the
2159<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2160dynamic call chain, undefined behavior results.</p>
2161</div>
2162
2163<!-- _______________________________________________________________________ -->
2164
2165<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2166Instruction</a> </div>
2167
2168<div class="doc_text">
2169
2170<h5>Syntax:</h5>
2171<pre>
2172 unreachable
2173</pre>
2174
2175<h5>Overview:</h5>
2176
2177<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2178instruction is used to inform the optimizer that a particular portion of the
2179code is not reachable. This can be used to indicate that the code after a
2180no-return function cannot be reached, and other facts.</p>
2181
2182<h5>Semantics:</h5>
2183
2184<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2185</div>
2186
2187
2188
2189<!-- ======================================================================= -->
2190<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2191<div class="doc_text">
2192<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2193program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2194produce a single value. The operands might represent
2195multiple 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]2196The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2197<p>There are several different binary operators:</p>
2198</div>
2199<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2200<div class="doc_subsubsection">
2201 <a name="i_add">'<tt>add</tt>' Instruction</a>
2202</div>
2203
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2204<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2205
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2206<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2207
2208<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2209 &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]2210</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2211
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2212<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2213
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2214<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2215
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2216<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2217
2218<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2219 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2220 <a href="#t_vector">vector</a> values. Both arguments must have identical
2221 types.</p>
2222
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2223<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2224
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2225<p>The value produced is the integer or floating point sum of the two
2226operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2227
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2228<p>If an integer sum has unsigned overflow, the result returned is the
2229mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2230the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2231
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2232<p>Because LLVM integers use a two's complement representation, this
2233instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2234
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2235<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2236
2237<pre>
2238 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2239</pre>
2240</div>
2241<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2242<div class="doc_subsubsection">
2243 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2244</div>
2245
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2246<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2247
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2248<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2249
2250<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2251 &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]2252</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2253
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2254<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2255
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2256<p>The '<tt>sub</tt>' instruction returns the difference of its two
2257operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2258
2259<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2260'<tt>neg</tt>' instruction present in most other intermediate
2261representations.</p>
2262
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2263<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2264
2265<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2266 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2267 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2268 types.</p>
2269
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2270<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2271
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2272<p>The value produced is the integer or floating point difference of
2273the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2274
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2275<p>If an integer difference has unsigned overflow, the result returned is the
2276mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2277the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2278
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2279<p>Because LLVM integers use a two's complement representation, this
2280instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2281
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2282<h5>Example:</h5>
2283<pre>
2284 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2285 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2286</pre>
2287</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2288
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2289<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2290<div class="doc_subsubsection">
2291 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2292</div>
2293
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2294<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2295
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2296<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2297<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]2298</pre>
2299<h5>Overview:</h5>
2300<p>The '<tt>mul</tt>' instruction returns the product of its two
2301operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2302
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2303<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2304
2305<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2306href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2307or <a href="#t_vector">vector</a> values. Both arguments must have identical
2308types.</p>
2309
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2310<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2311
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2312<p>The value produced is the integer or floating point product of the
2313two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2314
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2315<p>If the result of an integer multiplication has unsigned overflow,
2316the result returned is the mathematical result modulo
23172<sup>n</sup>, where n is the bit width of the result.</p>
2318<p>Because LLVM integers use a two's complement representation, and the
2319result is the same width as the operands, this instruction returns the
2320correct result for both signed and unsigned integers. If a full product
2321(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2322should be sign-extended or zero-extended as appropriate to the
2323width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2324<h5>Example:</h5>
2325<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2326</pre>
2327</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2328
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2329<!-- _______________________________________________________________________ -->
2330<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2331</a></div>
2332<div class="doc_text">
2333<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2334<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]2335</pre>
2336<h5>Overview:</h5>
2337<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2338operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2339
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2340<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2341
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2342<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2343<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2344values. Both arguments must have identical types.</p>
2345
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2346<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2347
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2348<p>The value produced is the unsigned integer quotient of the two operands.</p>
2349<p>Note that unsigned integer division and signed integer division are distinct
2350operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2351<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2352<h5>Example:</h5>
2353<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2354</pre>
2355</div>
2356<!-- _______________________________________________________________________ -->
2357<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2358</a> </div>
2359<div class="doc_text">
2360<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2361<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2362 &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]2363</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2364
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2365<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2366
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2367<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2368operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2369
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2370<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2371
2372<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2373<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2374values. Both arguments must have identical types.</p>
2375
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2376<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2377<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]2378<p>Note that signed integer division and unsigned integer division are distinct
2379operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2380<p>Division by zero leads to undefined behavior. Overflow also leads to
2381undefined behavior; this is a rare case, but can occur, for example,
2382by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2383<h5>Example:</h5>
2384<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2385</pre>
2386</div>
2387<!-- _______________________________________________________________________ -->
2388<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2389Instruction</a> </div>
2390<div class="doc_text">
2391<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2392<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2393 &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]2394</pre>
2395<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2396
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2397<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2398operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2399
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2400<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2401
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2402<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2403<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2404of floating point values. Both arguments must have identical types.</p>
2405
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2406<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2407
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2408<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2409
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2410<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2411
2412<pre>
2413 &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]2414</pre>
2415</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2416
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2417<!-- _______________________________________________________________________ -->
2418<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2419</div>
2420<div class="doc_text">
2421<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2422<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]2423</pre>
2424<h5>Overview:</h5>
2425<p>The '<tt>urem</tt>' instruction returns the remainder from the
2426unsigned division of its two arguments.</p>
2427<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2428<p>The two arguments to the '<tt>urem</tt>' instruction must be
2429<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2430values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2431<h5>Semantics:</h5>
2432<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2433This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2434<p>Note that unsigned integer remainder and signed integer remainder are
2435distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2436<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2437<h5>Example:</h5>
2438<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2439</pre>
2440
2441</div>
2442<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2443<div class="doc_subsubsection">
2444 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2445</div>
2446
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2447<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2448
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2449<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2450
2451<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2452 &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]2453</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2454
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2455<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2456
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2457<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2458signed division of its two operands. This instruction can also take
2459<a href="#t_vector">vector</a> versions of the values in which case
2460the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2461
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2462<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2463
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2464<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2465<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2466values. Both arguments must have identical types.</p>
2467
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2468<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2469
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2470<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2471has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2472operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2473a value. For more information about the difference, see <a
2474 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2475Math Forum</a>. For a table of how this is implemented in various languages,
2476please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2477Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2478<p>Note that signed integer remainder and unsigned integer remainder are
2479distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2480<p>Taking the remainder of a division by zero leads to undefined behavior.
2481Overflow also leads to undefined behavior; this is a rare case, but can occur,
2482for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2483(The remainder doesn't actually overflow, but this rule lets srem be
2484implemented using instructions that return both the result of the division
2485and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2486<h5>Example:</h5>
2487<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2488</pre>
2489
2490</div>
2491<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2492<div class="doc_subsubsection">
2493 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2494
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2495<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2496
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2497<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2498<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]2499</pre>
2500<h5>Overview:</h5>
2501<p>The '<tt>frem</tt>' instruction returns the remainder from the
2502division of its two operands.</p>
2503<h5>Arguments:</h5>
2504<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2505<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2506of floating point values. Both arguments must have identical types.</p>
2507
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2508<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2509
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2510<p>This instruction returns the <i>remainder</i> of a division.
2511The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2512
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2513<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2514
2515<pre>
2516 &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]2517</pre>
2518</div>
2519
2520<!-- ======================================================================= -->
2521<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2522Operations</a> </div>
2523<div class="doc_text">
2524<p>Bitwise binary operators are used to do various forms of
2525bit-twiddling in a program. They are generally very efficient
2526instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2527instructions. They require two operands of the same type, execute an operation on them,
2528and 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]2529</div>
2530
2531<!-- _______________________________________________________________________ -->
2532<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2533Instruction</a> </div>
2534<div class="doc_text">
2535<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2536<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]2537</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2538
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2539<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2540
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2541<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2542the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2543
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2544<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2545
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2546<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2547 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2548type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2549
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2550<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2551
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2552<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2553where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2554equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2555
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2556<h5>Example:</h5><pre>
2557 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2558 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2559 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2560 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2561</pre>
2562</div>
2563<!-- _______________________________________________________________________ -->
2564<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2565Instruction</a> </div>
2566<div class="doc_text">
2567<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2568<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]2569</pre>
2570
2571<h5>Overview:</h5>
2572<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2573operand shifted to the right a specified number of bits with zero fill.</p>
2574
2575<h5>Arguments:</h5>
2576<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2577<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]2578type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2579
2580<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2581
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2582<p>This instruction always performs a logical shift right operation. The most
2583significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2584shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2585the number of bits in <tt>op1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2586
2587<h5>Example:</h5>
2588<pre>
2589 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2590 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2591 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2592 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2593 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2594</pre>
2595</div>
2596
2597<!-- _______________________________________________________________________ -->
2598<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2599Instruction</a> </div>
2600<div class="doc_text">
2601
2602<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2603<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]2604</pre>
2605
2606<h5>Overview:</h5>
2607<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2608operand shifted to the right a specified number of bits with sign extension.</p>
2609
2610<h5>Arguments:</h5>
2611<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2612<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]2613type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2614
2615<h5>Semantics:</h5>
2616<p>This instruction always performs an arithmetic shift right operation,
2617The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2618of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2619larger than the number of bits in <tt>op1</tt>, the result is undefined.
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2620</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2621
2622<h5>Example:</h5>
2623<pre>
2624 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2625 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2626 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2627 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2628 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2629</pre>
2630</div>
2631
2632<!-- _______________________________________________________________________ -->
2633<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2634Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2635
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2636<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2637
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2638<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2639
2640<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2641 &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]2642</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2643
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2644<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2645
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2646<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2647its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2648
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2649<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2650
2651<p>The two arguments to the '<tt>and</tt>' instruction must be
2652<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2653values. Both arguments must have identical types.</p>
2654
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2655<h5>Semantics:</h5>
2656<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2657<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2658<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2659<table border="1" cellspacing="0" cellpadding="4">
2660 <tbody>
2661 <tr>
2662 <td>In0</td>
2663 <td>In1</td>
2664 <td>Out</td>
2665 </tr>
2666 <tr>
2667 <td>0</td>
2668 <td>0</td>
2669 <td>0</td>
2670 </tr>
2671 <tr>
2672 <td>0</td>
2673 <td>1</td>
2674 <td>0</td>
2675 </tr>
2676 <tr>
2677 <td>1</td>
2678 <td>0</td>
2679 <td>0</td>
2680 </tr>
2681 <tr>
2682 <td>1</td>
2683 <td>1</td>
2684 <td>1</td>
2685 </tr>
2686 </tbody>
2687</table>
2688</div>
2689<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2690<pre>
2691 &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]2692 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2693 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2694</pre>
2695</div>
2696<!-- _______________________________________________________________________ -->
2697<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2698<div class="doc_text">
2699<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2700<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]2701</pre>
2702<h5>Overview:</h5>
2703<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2704or of its two operands.</p>
2705<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2706
2707<p>The two arguments to the '<tt>or</tt>' instruction must be
2708<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2709values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2710<h5>Semantics:</h5>
2711<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2712<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2713<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2714<table border="1" cellspacing="0" cellpadding="4">
2715 <tbody>
2716 <tr>
2717 <td>In0</td>
2718 <td>In1</td>
2719 <td>Out</td>
2720 </tr>
2721 <tr>
2722 <td>0</td>
2723 <td>0</td>
2724 <td>0</td>
2725 </tr>
2726 <tr>
2727 <td>0</td>
2728 <td>1</td>
2729 <td>1</td>
2730 </tr>
2731 <tr>
2732 <td>1</td>
2733 <td>0</td>
2734 <td>1</td>
2735 </tr>
2736 <tr>
2737 <td>1</td>
2738 <td>1</td>
2739 <td>1</td>
2740 </tr>
2741 </tbody>
2742</table>
2743</div>
2744<h5>Example:</h5>
2745<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2746 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2747 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2748</pre>
2749</div>
2750<!-- _______________________________________________________________________ -->
2751<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2752Instruction</a> </div>
2753<div class="doc_text">
2754<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2755<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]2756</pre>
2757<h5>Overview:</h5>
2758<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2759or of its two operands. The <tt>xor</tt> is used to implement the
2760"one's complement" operation, which is the "~" operator in C.</p>
2761<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2762<p>The two arguments to the '<tt>xor</tt>' instruction must be
2763<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2764values. Both arguments must have identical types.</p>
2765
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2766<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2767
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2768<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2769<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2770<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2771<table border="1" cellspacing="0" cellpadding="4">
2772 <tbody>
2773 <tr>
2774 <td>In0</td>
2775 <td>In1</td>
2776 <td>Out</td>
2777 </tr>
2778 <tr>
2779 <td>0</td>
2780 <td>0</td>
2781 <td>0</td>
2782 </tr>
2783 <tr>
2784 <td>0</td>
2785 <td>1</td>
2786 <td>1</td>
2787 </tr>
2788 <tr>
2789 <td>1</td>
2790 <td>0</td>
2791 <td>1</td>
2792 </tr>
2793 <tr>
2794 <td>1</td>
2795 <td>1</td>
2796 <td>0</td>
2797 </tr>
2798 </tbody>
2799</table>
2800</div>
2801<p> </p>
2802<h5>Example:</h5>
2803<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2804 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2805 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2806 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2807</pre>
2808</div>
2809
2810<!-- ======================================================================= -->
2811<div class="doc_subsection">
2812 <a name="vectorops">Vector Operations</a>
2813</div>
2814
2815<div class="doc_text">
2816
2817<p>LLVM supports several instructions to represent vector operations in a
2818target-independent manner. These instructions cover the element-access and
2819vector-specific operations needed to process vectors effectively. While LLVM
2820does directly support these vector operations, many sophisticated algorithms
2821will want to use target-specific intrinsics to take full advantage of a specific
2822target.</p>
2823
2824</div>
2825
2826<!-- _______________________________________________________________________ -->
2827<div class="doc_subsubsection">
2828 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2829</div>
2830
2831<div class="doc_text">
2832
2833<h5>Syntax:</h5>
2834
2835<pre>
2836 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2837</pre>
2838
2839<h5>Overview:</h5>
2840
2841<p>
2842The '<tt>extractelement</tt>' instruction extracts a single scalar
2843element from a vector at a specified index.
2844</p>
2845
2846
2847<h5>Arguments:</h5>
2848
2849<p>
2850The first operand of an '<tt>extractelement</tt>' instruction is a
2851value of <a href="#t_vector">vector</a> type. The second operand is
2852an index indicating the position from which to extract the element.
2853The index may be a variable.</p>
2854
2855<h5>Semantics:</h5>
2856
2857<p>
2858The result is a scalar of the same type as the element type of
2859<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2860<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2861results are undefined.
2862</p>
2863
2864<h5>Example:</h5>
2865
2866<pre>
2867 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2868</pre>
2869</div>
2870
2871
2872<!-- _______________________________________________________________________ -->
2873<div class="doc_subsubsection">
2874 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2875</div>
2876
2877<div class="doc_text">
2878
2879<h5>Syntax:</h5>
2880
2881<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]2882 &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]2883</pre>
2884
2885<h5>Overview:</h5>
2886
2887<p>
2888The '<tt>insertelement</tt>' instruction inserts a scalar
2889element into a vector at a specified index.
2890</p>
2891
2892
2893<h5>Arguments:</h5>
2894
2895<p>
2896The first operand of an '<tt>insertelement</tt>' instruction is a
2897value of <a href="#t_vector">vector</a> type. The second operand is a
2898scalar value whose type must equal the element type of the first
2899operand. The third operand is an index indicating the position at
2900which to insert the value. The index may be a variable.</p>
2901
2902<h5>Semantics:</h5>
2903
2904<p>
2905The result is a vector of the same type as <tt>val</tt>. Its
2906element values are those of <tt>val</tt> except at position
2907<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2908exceeds the length of <tt>val</tt>, the results are undefined.
2909</p>
2910
2911<h5>Example:</h5>
2912
2913<pre>
2914 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2915</pre>
2916</div>
2917
2918<!-- _______________________________________________________________________ -->
2919<div class="doc_subsubsection">
2920 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2921</div>
2922
2923<div class="doc_text">
2924
2925<h5>Syntax:</h5>
2926
2927<pre>
2928 &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;n x i32&gt; &lt;mask&gt; <i>; yields &lt;n x &lt;ty&gt;&gt;</i>
2929</pre>
2930
2931<h5>Overview:</h5>
2932
2933<p>
2934The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2935from two input vectors, returning a vector of the same type.
2936</p>
2937
2938<h5>Arguments:</h5>
2939
2940<p>
2941The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2942with types that match each other and types that match the result of the
2943instruction. The third argument is a shuffle mask, which has the same number
2944of elements as the other vector type, but whose element type is always 'i32'.
2945</p>
2946
2947<p>
2948The shuffle mask operand is required to be a constant vector with either
2949constant integer or undef values.
2950</p>
2951
2952<h5>Semantics:</h5>
2953
2954<p>
2955The elements of the two input vectors are numbered from left to right across
2956both of the vectors. The shuffle mask operand specifies, for each element of
2957the result vector, which element of the two input registers the result element
2958gets. The element selector may be undef (meaning "don't care") and the second
2959operand may be undef if performing a shuffle from only one vector.
2960</p>
2961
2962<h5>Example:</h5>
2963
2964<pre>
2965 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2966 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2967 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2968 &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.
2969</pre>
2970</div>
2971
2972
2973<!-- ======================================================================= -->
2974<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2975 <a name="aggregateops">Aggregate Operations</a>
2976</div>
2977
2978<div class="doc_text">
2979
2980<p>LLVM supports several instructions for working with aggregate values.
2981</p>
2982
2983</div>
2984
2985<!-- _______________________________________________________________________ -->
2986<div class="doc_subsubsection">
2987 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
2988</div>
2989
2990<div class="doc_text">
2991
2992<h5>Syntax:</h5>
2993
2994<pre>
2995 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
2996</pre>
2997
2998<h5>Overview:</h5>
2999
3000<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3001The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3002or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3003</p>
3004
3005
3006<h5>Arguments:</h5>
3007
3008<p>
3009The first operand of an '<tt>extractvalue</tt>' instruction is a
3010value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3011type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3012in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3013'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3014</p>
3015
3016<h5>Semantics:</h5>
3017
3018<p>
3019The result is the value at the position in the aggregate specified by
3020the index operands.
3021</p>
3022
3023<h5>Example:</h5>
3024
3025<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3026 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3027</pre>
3028</div>
3029
3030
3031<!-- _______________________________________________________________________ -->
3032<div class="doc_subsubsection">
3033 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3034</div>
3035
3036<div class="doc_text">
3037
3038<h5>Syntax:</h5>
3039
3040<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3041 &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]3042</pre>
3043
3044<h5>Overview:</h5>
3045
3046<p>
3047The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3048into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3049</p>
3050
3051
3052<h5>Arguments:</h5>
3053
3054<p>
3055The first operand of an '<tt>insertvalue</tt>' instruction is a
3056value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3057The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3058The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3059indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3060indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3061'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3062The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3063by the indices.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3064
3065<h5>Semantics:</h5>
3066
3067<p>
3068The result is an aggregate of the same type as <tt>val</tt>. Its
3069value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3070specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3071</p>
3072
3073<h5>Example:</h5>
3074
3075<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3076 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3077</pre>
3078</div>
3079
3080
3081<!-- ======================================================================= -->
3082<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3083 <a name="memoryops">Memory Access and Addressing Operations</a>
3084</div>
3085
3086<div class="doc_text">
3087
3088<p>A key design point of an SSA-based representation is how it
3089represents memory. In LLVM, no memory locations are in SSA form, which
3090makes things very simple. This section describes how to read, write,
3091allocate, and free memory in LLVM.</p>
3092
3093</div>
3094
3095<!-- _______________________________________________________________________ -->
3096<div class="doc_subsubsection">
3097 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3098</div>
3099
3100<div class="doc_text">
3101
3102<h5>Syntax:</h5>
3103
3104<pre>
3105 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3106</pre>
3107
3108<h5>Overview:</h5>
3109
3110<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3111heap and returns a pointer to it. The object is always allocated in the generic
3112address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3113
3114<h5>Arguments:</h5>
3115
3116<p>The '<tt>malloc</tt>' instruction allocates
3117<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3118bytes of memory from the operating system and returns a pointer of the
3119appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3120number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3121If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3122be aligned to at least that boundary. If not specified, or if zero, the target can
3123choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3124
3125<p>'<tt>type</tt>' must be a sized type.</p>
3126
3127<h5>Semantics:</h5>
3128
3129<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3130a pointer is returned. The result of a zero byte allocattion is undefined. The
3131result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3132
3133<h5>Example:</h5>
3134
3135<pre>
3136 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
3137
3138 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3139 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3140 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3141 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3142 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3143</pre>
3144</div>
3145
3146<!-- _______________________________________________________________________ -->
3147<div class="doc_subsubsection">
3148 <a name="i_free">'<tt>free</tt>' Instruction</a>
3149</div>
3150
3151<div class="doc_text">
3152
3153<h5>Syntax:</h5>
3154
3155<pre>
3156 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
3157</pre>
3158
3159<h5>Overview:</h5>
3160
3161<p>The '<tt>free</tt>' instruction returns memory back to the unused
3162memory heap to be reallocated in the future.</p>
3163
3164<h5>Arguments:</h5>
3165
3166<p>'<tt>value</tt>' shall be a pointer value that points to a value
3167that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3168instruction.</p>
3169
3170<h5>Semantics:</h5>
3171
3172<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3173after this instruction executes. If the pointer is null, the operation
3174is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3175
3176<h5>Example:</h5>
3177
3178<pre>
3179 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3180 free [4 x i8]* %array
3181</pre>
3182</div>
3183
3184<!-- _______________________________________________________________________ -->
3185<div class="doc_subsubsection">
3186 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3187</div>
3188
3189<div class="doc_text">
3190
3191<h5>Syntax:</h5>
3192
3193<pre>
3194 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3195</pre>
3196
3197<h5>Overview:</h5>
3198
3199<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3200currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3201returns to its caller. The object is always allocated in the generic address
3202space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3203
3204<h5>Arguments:</h5>
3205
3206<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3207bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3208appropriate type to the program. If "NumElements" is specified, it is the
3209number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3210If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3211to be aligned to at least that boundary. If not specified, or if zero, the target
3212can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3213
3214<p>'<tt>type</tt>' may be any sized type.</p>
3215
3216<h5>Semantics:</h5>
3217
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3218<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3219there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3220memory is automatically released when the function returns. The '<tt>alloca</tt>'
3221instruction is commonly used to represent automatic variables that must
3222have an address available. When the function returns (either with the <tt><a
3223 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]3224instructions), the memory is reclaimed. Allocating zero bytes
3225is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3226
3227<h5>Example:</h5>
3228
3229<pre>
3230 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3231 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3232 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3233 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3234</pre>
3235</div>
3236
3237<!-- _______________________________________________________________________ -->
3238<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3239Instruction</a> </div>
3240<div class="doc_text">
3241<h5>Syntax:</h5>
3242<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>
3243<h5>Overview:</h5>
3244<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3245<h5>Arguments:</h5>
3246<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3247address from which to load. The pointer must point to a <a
3248 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3249marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3250the number or order of execution of this <tt>load</tt> with other
3251volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3252instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3253<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3254The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3255(that is, the alignment of the memory address). A value of 0 or an
3256omitted "align" argument means that the operation has the preferential
3257alignment for the target. It is the responsibility of the code emitter
3258to ensure that the alignment information is correct. Overestimating
3259the alignment results in an undefined behavior. Underestimating the
3260alignment may produce less efficient code. An alignment of 1 is always
3261safe.
3262</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3263<h5>Semantics:</h5>
3264<p>The location of memory pointed to is loaded.</p>
3265<h5>Examples:</h5>
3266<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3267 <a
3268 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3269 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3270</pre>
3271</div>
3272<!-- _______________________________________________________________________ -->
3273<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3274Instruction</a> </div>
3275<div class="doc_text">
3276<h5>Syntax:</h5>
3277<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3278 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3279</pre>
3280<h5>Overview:</h5>
3281<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3282<h5>Arguments:</h5>
3283<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3284to 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]3285operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3286of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3287operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3288optimizer is not allowed to modify the number or order of execution of
3289this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3290 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3291<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3292The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3293(that is, the alignment of the memory address). A value of 0 or an
3294omitted "align" argument means that the operation has the preferential
3295alignment for the target. It is the responsibility of the code emitter
3296to ensure that the alignment information is correct. Overestimating
3297the alignment results in an undefined behavior. Underestimating the
3298alignment may produce less efficient code. An alignment of 1 is always
3299safe.
3300</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3301<h5>Semantics:</h5>
3302<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3303at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3304<h5>Example:</h5>
3305<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3306 store i32 3, i32* %ptr <i>; yields {void}</i>
3307 %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]3308</pre>
3309</div>
3310
3311<!-- _______________________________________________________________________ -->
3312<div class="doc_subsubsection">
3313 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3314</div>
3315
3316<div class="doc_text">
3317<h5>Syntax:</h5>
3318<pre>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame^]3319 &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]3320</pre>
3321
3322<h5>Overview:</h5>
3323
3324<p>
3325The '<tt>getelementptr</tt>' instruction is used to get the address of a
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame^]3326subelement of an aggregate data structure. It performs address calculation only
3327and does not access memory.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3328
3329<h5>Arguments:</h5>
3330
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame^]3331<p>The first argument is always a pointer, and forms the basis of the
3332calculation. The remaining arguments are indices, that indicate which of the
3333elements of the aggregate object are indexed. The interpretation of each index
3334is dependent on the type being indexed into. The first index always indexes the
3335pointer value given as the first argument, the second index indexes a value of
3336the type pointed to (not necessarily the value directly pointed to, since the
3337first index can be non-zero), etc. The first type indexed into must be a pointer
3338value, subsequent types can be arrays, vectors and structs. Note that subsequent
3339types being indexed into can never be pointers, since that would require loading
3340the pointer before continuing calculation.</p>
3341
3342<p>The type of each index argument depends on the type it is indexing into.
3343When indexing into a (packed) structure, only <tt>i32</tt> integer
3344<b>constants</b> are allowed. When indexing into an array, pointer or vector,
3345only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3346will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3347
3348<p>For example, let's consider a C code fragment and how it gets
3349compiled to LLVM:</p>
3350
3351<div class="doc_code">
3352<pre>
3353struct RT {
3354 char A;
3355 int B[10][20];
3356 char C;
3357};
3358struct ST {
3359 int X;
3360 double Y;
3361 struct RT Z;
3362};
3363
3364int *foo(struct ST *s) {
3365 return &amp;s[1].Z.B[5][13];
3366}
3367</pre>
3368</div>
3369
3370<p>The LLVM code generated by the GCC frontend is:</p>
3371
3372<div class="doc_code">
3373<pre>
3374%RT = type { i8 , [10 x [20 x i32]], i8 }
3375%ST = type { i32, double, %RT }
3376
3377define i32* %foo(%ST* %s) {
3378entry:
3379 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3380 ret i32* %reg
3381}
3382</pre>
3383</div>
3384
3385<h5>Semantics:</h5>
3386
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3387<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3388type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3389}</tt>' type, a structure. The second index indexes into the third element of
3390the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3391i8 }</tt>' type, another structure. The third index indexes into the second
3392element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3393array. The two dimensions of the array are subscripted into, yielding an
3394'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3395to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3396
3397<p>Note that it is perfectly legal to index partially through a
3398structure, returning a pointer to an inner element. Because of this,
3399the LLVM code for the given testcase is equivalent to:</p>
3400
3401<pre>
3402 define i32* %foo(%ST* %s) {
3403 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3404 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3405 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3406 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3407 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3408 ret i32* %t5
3409 }
3410</pre>
3411
3412<p>Note that it is undefined to access an array out of bounds: array and
3413pointer indexes must always be within the defined bounds of the array type.
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3414The one exception for this rule is zero length arrays. These arrays are
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3415defined to be accessible as variable length arrays, which requires access
3416beyond the zero'th element.</p>
3417
3418<p>The getelementptr instruction is often confusing. For some more insight
3419into how it works, see <a href="GetElementPtr.html">the getelementptr
3420FAQ</a>.</p>
3421
3422<h5>Example:</h5>
3423
3424<pre>
3425 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame^]3426 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3427 <i>; yields i8*:vptr</i>
3428 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
3429 <i>; yields i8*:eptr</i>
3430 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3431</pre>
3432</div>
3433
3434<!-- ======================================================================= -->
3435<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3436</div>
3437<div class="doc_text">
3438<p>The instructions in this category are the conversion instructions (casting)
3439which all take a single operand and a type. They perform various bit conversions
3440on the operand.</p>
3441</div>
3442
3443<!-- _______________________________________________________________________ -->
3444<div class="doc_subsubsection">
3445 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3446</div>
3447<div class="doc_text">
3448
3449<h5>Syntax:</h5>
3450<pre>
3451 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3452</pre>
3453
3454<h5>Overview:</h5>
3455<p>
3456The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3457</p>
3458
3459<h5>Arguments:</h5>
3460<p>
3461The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3462be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3463and type of the result, which must be an <a href="#t_integer">integer</a>
3464type. The bit size of <tt>value</tt> must be larger than the bit size of
3465<tt>ty2</tt>. Equal sized types are not allowed.</p>
3466
3467<h5>Semantics:</h5>
3468<p>
3469The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3470and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3471larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3472It will always truncate bits.</p>
3473
3474<h5>Example:</h5>
3475<pre>
3476 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3477 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3478 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3479</pre>
3480</div>
3481
3482<!-- _______________________________________________________________________ -->
3483<div class="doc_subsubsection">
3484 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3485</div>
3486<div class="doc_text">
3487
3488<h5>Syntax:</h5>
3489<pre>
3490 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3491</pre>
3492
3493<h5>Overview:</h5>
3494<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3495<tt>ty2</tt>.</p>
3496
3497
3498<h5>Arguments:</h5>
3499<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3500<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3501also be of <a href="#t_integer">integer</a> type. The bit size of the
3502<tt>value</tt> must be smaller than the bit size of the destination type,
3503<tt>ty2</tt>.</p>
3504
3505<h5>Semantics:</h5>
3506<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3507bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3508
3509<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3510
3511<h5>Example:</h5>
3512<pre>
3513 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3514 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3515</pre>
3516</div>
3517
3518<!-- _______________________________________________________________________ -->
3519<div class="doc_subsubsection">
3520 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3521</div>
3522<div class="doc_text">
3523
3524<h5>Syntax:</h5>
3525<pre>
3526 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3527</pre>
3528
3529<h5>Overview:</h5>
3530<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3531
3532<h5>Arguments:</h5>
3533<p>
3534The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3535<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3536also be of <a href="#t_integer">integer</a> type. The bit size of the
3537<tt>value</tt> must be smaller than the bit size of the destination type,
3538<tt>ty2</tt>.</p>
3539
3540<h5>Semantics:</h5>
3541<p>
3542The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3543bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3544the type <tt>ty2</tt>.</p>
3545
3546<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3547
3548<h5>Example:</h5>
3549<pre>
3550 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3551 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3552</pre>
3553</div>
3554
3555<!-- _______________________________________________________________________ -->
3556<div class="doc_subsubsection">
3557 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3558</div>
3559
3560<div class="doc_text">
3561
3562<h5>Syntax:</h5>
3563
3564<pre>
3565 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3566</pre>
3567
3568<h5>Overview:</h5>
3569<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3570<tt>ty2</tt>.</p>
3571
3572
3573<h5>Arguments:</h5>
3574<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3575 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3576cast it to. The size of <tt>value</tt> must be larger than the size of
3577<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3578<i>no-op cast</i>.</p>
3579
3580<h5>Semantics:</h5>
3581<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3582<a href="#t_floating">floating point</a> type to a smaller
3583<a href="#t_floating">floating point</a> type. If the value cannot fit within
3584the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3585
3586<h5>Example:</h5>
3587<pre>
3588 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3589 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3590</pre>
3591</div>
3592
3593<!-- _______________________________________________________________________ -->
3594<div class="doc_subsubsection">
3595 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3596</div>
3597<div class="doc_text">
3598
3599<h5>Syntax:</h5>
3600<pre>
3601 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3602</pre>
3603
3604<h5>Overview:</h5>
3605<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3606floating point value.</p>
3607
3608<h5>Arguments:</h5>
3609<p>The '<tt>fpext</tt>' instruction takes a
3610<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3611and a <a href="#t_floating">floating point</a> type to cast it to. The source
3612type must be smaller than the destination type.</p>
3613
3614<h5>Semantics:</h5>
3615<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3616<a href="#t_floating">floating point</a> type to a larger
3617<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3618used to make a <i>no-op cast</i> because it always changes bits. Use
3619<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3620
3621<h5>Example:</h5>
3622<pre>
3623 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3624 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3625</pre>
3626</div>
3627
3628<!-- _______________________________________________________________________ -->
3629<div class="doc_subsubsection">
3630 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3631</div>
3632<div class="doc_text">
3633
3634<h5>Syntax:</h5>
3635<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3636 &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]3637</pre>
3638
3639<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3640<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3641unsigned integer equivalent of type <tt>ty2</tt>.
3642</p>
3643
3644<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3645<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]3646scalar or vector <a href="#t_floating">floating point</a> value, and a type
3647to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3648type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3649vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3650
3651<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3652<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3653<a href="#t_floating">floating point</a> operand into the nearest (rounding
3654towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3655the results are undefined.</p>
3656
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3657<h5>Example:</h5>
3658<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3659 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3660 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3661 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3662</pre>
3663</div>
3664
3665<!-- _______________________________________________________________________ -->
3666<div class="doc_subsubsection">
3667 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3668</div>
3669<div class="doc_text">
3670
3671<h5>Syntax:</h5>
3672<pre>
3673 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3674</pre>
3675
3676<h5>Overview:</h5>
3677<p>The '<tt>fptosi</tt>' instruction converts
3678<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3679</p>
3680
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3681<h5>Arguments:</h5>
3682<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]3683scalar or vector <a href="#t_floating">floating point</a> value, and a type
3684to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3685type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3686vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3687
3688<h5>Semantics:</h5>
3689<p>The '<tt>fptosi</tt>' instruction converts its
3690<a href="#t_floating">floating point</a> operand into the nearest (rounding
3691towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3692the results are undefined.</p>
3693
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3694<h5>Example:</h5>
3695<pre>
3696 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3697 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3698 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3699</pre>
3700</div>
3701
3702<!-- _______________________________________________________________________ -->
3703<div class="doc_subsubsection">
3704 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3705</div>
3706<div class="doc_text">
3707
3708<h5>Syntax:</h5>
3709<pre>
3710 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3711</pre>
3712
3713<h5>Overview:</h5>
3714<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3715integer and converts that value to the <tt>ty2</tt> type.</p>
3716
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3717<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3718<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3719scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3720to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3721type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3722floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3723
3724<h5>Semantics:</h5>
3725<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3726integer quantity and converts it to the corresponding floating point value. If
3727the value cannot fit in the floating point value, the results are undefined.</p>
3728
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3729<h5>Example:</h5>
3730<pre>
3731 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3732 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3733</pre>
3734</div>
3735
3736<!-- _______________________________________________________________________ -->
3737<div class="doc_subsubsection">
3738 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3739</div>
3740<div class="doc_text">
3741
3742<h5>Syntax:</h5>
3743<pre>
3744 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3745</pre>
3746
3747<h5>Overview:</h5>
3748<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3749integer and converts that value to the <tt>ty2</tt> type.</p>
3750
3751<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3752<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3753scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3754to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3755type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3756floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3757
3758<h5>Semantics:</h5>
3759<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3760integer quantity and converts it to the corresponding floating point value. If
3761the value cannot fit in the floating point value, the results are undefined.</p>
3762
3763<h5>Example:</h5>
3764<pre>
3765 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3766 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3767</pre>
3768</div>
3769
3770<!-- _______________________________________________________________________ -->
3771<div class="doc_subsubsection">
3772 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3773</div>
3774<div class="doc_text">
3775
3776<h5>Syntax:</h5>
3777<pre>
3778 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3779</pre>
3780
3781<h5>Overview:</h5>
3782<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3783the integer type <tt>ty2</tt>.</p>
3784
3785<h5>Arguments:</h5>
3786<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3787must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3788<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3789
3790<h5>Semantics:</h5>
3791<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3792<tt>ty2</tt> by interpreting the pointer value as an integer and either
3793truncating or zero extending that value to the size of the integer type. If
3794<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3795<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3796are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3797change.</p>
3798
3799<h5>Example:</h5>
3800<pre>
3801 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3802 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3803</pre>
3804</div>
3805
3806<!-- _______________________________________________________________________ -->
3807<div class="doc_subsubsection">
3808 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3809</div>
3810<div class="doc_text">
3811
3812<h5>Syntax:</h5>
3813<pre>
3814 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3815</pre>
3816
3817<h5>Overview:</h5>
3818<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3819a pointer type, <tt>ty2</tt>.</p>
3820
3821<h5>Arguments:</h5>
3822<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3823value to cast, and a type to cast it to, which must be a
3824<a href="#t_pointer">pointer</a> type.
3825
3826<h5>Semantics:</h5>
3827<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3828<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3829the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3830size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3831the size of a pointer then a zero extension is done. If they are the same size,
3832nothing is done (<i>no-op cast</i>).</p>
3833
3834<h5>Example:</h5>
3835<pre>
3836 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3837 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3838 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3839</pre>
3840</div>
3841
3842<!-- _______________________________________________________________________ -->
3843<div class="doc_subsubsection">
3844 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3845</div>
3846<div class="doc_text">
3847
3848<h5>Syntax:</h5>
3849<pre>
3850 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3851</pre>
3852
3853<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3854
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3855<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3856<tt>ty2</tt> without changing any bits.</p>
3857
3858<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3859
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3860<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]3861a non-aggregate first class value, and a type to cast it to, which must also be
3862a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
3863<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3864and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3865type is a pointer, the destination type must also be a pointer. This
3866instruction supports bitwise conversion of vectors to integers and to vectors
3867of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3868
3869<h5>Semantics:</h5>
3870<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3871<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3872this conversion. The conversion is done as if the <tt>value</tt> had been
3873stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3874converted to other pointer types with this instruction. To convert pointers to
3875other types, use the <a href="#i_inttoptr">inttoptr</a> or
3876<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3877
3878<h5>Example:</h5>
3879<pre>
3880 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3881 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3882 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3883</pre>
3884</div>
3885
3886<!-- ======================================================================= -->
3887<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3888<div class="doc_text">
3889<p>The instructions in this category are the "miscellaneous"
3890instructions, which defy better classification.</p>
3891</div>
3892
3893<!-- _______________________________________________________________________ -->
3894<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3895</div>
3896<div class="doc_text">
3897<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3898<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]3899</pre>
3900<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3901<p>The '<tt>icmp</tt>' instruction returns a boolean value or
3902a vector of boolean values based on comparison
3903of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3904<h5>Arguments:</h5>
3905<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3906the condition code indicating the kind of comparison to perform. It is not
3907a value, just a keyword. The possible condition code are:
3908<ol>
3909 <li><tt>eq</tt>: equal</li>
3910 <li><tt>ne</tt>: not equal </li>
3911 <li><tt>ugt</tt>: unsigned greater than</li>
3912 <li><tt>uge</tt>: unsigned greater or equal</li>
3913 <li><tt>ult</tt>: unsigned less than</li>
3914 <li><tt>ule</tt>: unsigned less or equal</li>
3915 <li><tt>sgt</tt>: signed greater than</li>
3916 <li><tt>sge</tt>: signed greater or equal</li>
3917 <li><tt>slt</tt>: signed less than</li>
3918 <li><tt>sle</tt>: signed less or equal</li>
3919</ol>
3920<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3921<a href="#t_pointer">pointer</a>
3922or integer <a href="#t_vector">vector</a> typed.
3923They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3924<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3925<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]3926the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3927yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3928<ol>
3929 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3930 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3931 </li>
3932 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3933 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3934 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3935 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3936 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3937 <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]3938 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3939 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3940 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3941 <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]3942 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3943 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3944 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3945 <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]3946 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3947 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3948 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3949 <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]3950</ol>
3951<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3952values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3953<p>If the operands are integer vectors, then they are compared
3954element by element. The result is an <tt>i1</tt> vector with
3955the same number of elements as the values being compared.
3956Otherwise, the result is an <tt>i1</tt>.
3957</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3958
3959<h5>Example:</h5>
3960<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3961 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3962 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3963 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3964 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3965 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3966</pre>
3967</div>
3968
3969<!-- _______________________________________________________________________ -->
3970<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3971</div>
3972<div class="doc_text">
3973<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3974<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]3975</pre>
3976<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3977<p>The '<tt>fcmp</tt>' instruction returns a boolean value
3978or vector of boolean values based on comparison
3979of its operands.
3980<p>
3981If the operands are floating point scalars, then the result
3982type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
3983</p>
3984<p>If the operands are floating point vectors, then the result type
3985is a vector of boolean with the same number of elements as the
3986operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3987<h5>Arguments:</h5>
3988<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3989the condition code indicating the kind of comparison to perform. It is not
3990a value, just a keyword. The possible condition code are:
3991<ol>
3992 <li><tt>false</tt>: no comparison, always returns false</li>
3993 <li><tt>oeq</tt>: ordered and equal</li>
3994 <li><tt>ogt</tt>: ordered and greater than </li>
3995 <li><tt>oge</tt>: ordered and greater than or equal</li>
3996 <li><tt>olt</tt>: ordered and less than </li>
3997 <li><tt>ole</tt>: ordered and less than or equal</li>
3998 <li><tt>one</tt>: ordered and not equal</li>
3999 <li><tt>ord</tt>: ordered (no nans)</li>
4000 <li><tt>ueq</tt>: unordered or equal</li>
4001 <li><tt>ugt</tt>: unordered or greater than </li>
4002 <li><tt>uge</tt>: unordered or greater than or equal</li>
4003 <li><tt>ult</tt>: unordered or less than </li>
4004 <li><tt>ule</tt>: unordered or less than or equal</li>
4005 <li><tt>une</tt>: unordered or not equal</li>
4006 <li><tt>uno</tt>: unordered (either nans)</li>
4007 <li><tt>true</tt>: no comparison, always returns true</li>
4008</ol>
4009<p><i>Ordered</i> means that neither operand is a QNAN while
4010<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4011<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4012either a <a href="#t_floating">floating point</a> type
4013or a <a href="#t_vector">vector</a> of floating point type.
4014They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4015<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4016<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4017according to the condition code given as <tt>cond</tt>.
4018If the operands are vectors, then the vectors are compared
4019element by element.
4020Each comparison performed
4021always yields an <a href="#t_primitive">i1</a> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4022<ol>
4023 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4024 <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]4025 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4026 <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]4027 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4028 <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]4029 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4030 <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]4031 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4032 <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]4033 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4034 <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]4035 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4036 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4037 <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]4038 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4039 <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]4040 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4041 <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]4042 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4043 <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]4044 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4045 <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]4046 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4047 <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]4048 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4049 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4050 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4051</ol>
4052
4053<h5>Example:</h5>
4054<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]4055 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4056 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4057 &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]4058</pre>
4059</div>
4060
4061<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4062<div class="doc_subsubsection">
4063 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4064</div>
4065<div class="doc_text">
4066<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4067<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]4068</pre>
4069<h5>Overview:</h5>
4070<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4071element-wise comparison of its two integer vector operands.</p>
4072<h5>Arguments:</h5>
4073<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4074the condition code indicating the kind of comparison to perform. It is not
4075a value, just a keyword. The possible condition code are:
4076<ol>
4077 <li><tt>eq</tt>: equal</li>
4078 <li><tt>ne</tt>: not equal </li>
4079 <li><tt>ugt</tt>: unsigned greater than</li>
4080 <li><tt>uge</tt>: unsigned greater or equal</li>
4081 <li><tt>ult</tt>: unsigned less than</li>
4082 <li><tt>ule</tt>: unsigned less or equal</li>
4083 <li><tt>sgt</tt>: signed greater than</li>
4084 <li><tt>sge</tt>: signed greater or equal</li>
4085 <li><tt>slt</tt>: signed less than</li>
4086 <li><tt>sle</tt>: signed less or equal</li>
4087</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4088<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4089<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4090<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4091<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4092according to the condition code given as <tt>cond</tt>. The comparison yields a
4093<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4094identical type as the values being compared. The most significant bit in each
4095element is 1 if the element-wise comparison evaluates to true, and is 0
4096otherwise. All other bits of the result are undefined. The condition codes
4097are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4098instruction</a>.
4099
4100<h5>Example:</h5>
4101<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4102 &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>
4103 &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]4104</pre>
4105</div>
4106
4107<!-- _______________________________________________________________________ -->
4108<div class="doc_subsubsection">
4109 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4110</div>
4111<div class="doc_text">
4112<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4113<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]4114<h5>Overview:</h5>
4115<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4116element-wise comparison of its two floating point vector operands. The output
4117elements have the same width as the input elements.</p>
4118<h5>Arguments:</h5>
4119<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4120the condition code indicating the kind of comparison to perform. It is not
4121a value, just a keyword. The possible condition code are:
4122<ol>
4123 <li><tt>false</tt>: no comparison, always returns false</li>
4124 <li><tt>oeq</tt>: ordered and equal</li>
4125 <li><tt>ogt</tt>: ordered and greater than </li>
4126 <li><tt>oge</tt>: ordered and greater than or equal</li>
4127 <li><tt>olt</tt>: ordered and less than </li>
4128 <li><tt>ole</tt>: ordered and less than or equal</li>
4129 <li><tt>one</tt>: ordered and not equal</li>
4130 <li><tt>ord</tt>: ordered (no nans)</li>
4131 <li><tt>ueq</tt>: unordered or equal</li>
4132 <li><tt>ugt</tt>: unordered or greater than </li>
4133 <li><tt>uge</tt>: unordered or greater than or equal</li>
4134 <li><tt>ult</tt>: unordered or less than </li>
4135 <li><tt>ule</tt>: unordered or less than or equal</li>
4136 <li><tt>une</tt>: unordered or not equal</li>
4137 <li><tt>uno</tt>: unordered (either nans)</li>
4138 <li><tt>true</tt>: no comparison, always returns true</li>
4139</ol>
4140<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4141<a href="#t_floating">floating point</a> typed. They must also be identical
4142types.</p>
4143<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4144<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4145according to the condition code given as <tt>cond</tt>. The comparison yields a
4146<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4147an identical number of elements as the values being compared, and each element
4148having identical with to the width of the floating point elements. The most
4149significant bit in each element is 1 if the element-wise comparison evaluates to
4150true, and is 0 otherwise. All other bits of the result are undefined. The
4151condition codes are evaluated identically to the
4152<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.
4153
4154<h5>Example:</h5>
4155<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4156 &lt;result&gt; = vfcmp oeq &lt;2 x float&gt; &lt; float 4, float 0 &gt;, &lt; float 5, float 0 &gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4157 &lt;result&gt; = vfcmp ult &lt;2 x double&gt; &lt; double 1, double 2 &gt;, &lt; double 2, double 2&gt; <i>; yields: result=&lt;2 x i64&gt; &lt; i64 -1, i64 0 &gt;</i>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4158</pre>
4159</div>
4160
4161<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4162<div class="doc_subsubsection">
4163 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4164</div>
4165
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4166<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4167
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4168<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4169
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4170<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4171<h5>Overview:</h5>
4172<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4173the SSA graph representing the function.</p>
4174<h5>Arguments:</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 type of the incoming values is specified with the first type
4177field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4178as arguments, with one pair for each predecessor basic block of the
4179current block. Only values of <a href="#t_firstclass">first class</a>
4180type may be used as the value arguments to the PHI node. Only labels
4181may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4182
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4183<p>There must be no non-phi instructions between the start of a basic
4184block and the PHI instructions: i.e. PHI instructions must be first in
4185a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4186
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4187<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4188
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4189<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4190specified by the pair corresponding to the predecessor basic block that executed
4191just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4192
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4193<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4194<pre>
4195Loop: ; Infinite loop that counts from 0 on up...
4196 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4197 %nextindvar = add i32 %indvar, 1
4198 br label %Loop
4199</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4200</div>
4201
4202<!-- _______________________________________________________________________ -->
4203<div class="doc_subsubsection">
4204 <a name="i_select">'<tt>select</tt>' Instruction</a>
4205</div>
4206
4207<div class="doc_text">
4208
4209<h5>Syntax:</h5>
4210
4211<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4212 &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>
4213
4214 <i>selty</i> is either i1 or {&lt;N x i1&gt}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4215</pre>
4216
4217<h5>Overview:</h5>
4218
4219<p>
4220The '<tt>select</tt>' instruction is used to choose one value based on a
4221condition, without branching.
4222</p>
4223
4224
4225<h5>Arguments:</h5>
4226
4227<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4228The '<tt>select</tt>' instruction requires an 'i1' value or
4229a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4230condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4231type. If the val1/val2 are vectors and
4232the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4233individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4234</p>
4235
4236<h5>Semantics:</h5>
4237
4238<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4239If 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]4240value argument; otherwise, it returns the second value argument.
4241</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4242<p>
4243If the condition is a vector of i1, then the value arguments must
4244be vectors of the same size, and the selection is done element
4245by element.
4246</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4247
4248<h5>Example:</h5>
4249
4250<pre>
4251 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4252</pre>
4253</div>
4254
4255
4256<!-- _______________________________________________________________________ -->
4257<div class="doc_subsubsection">
4258 <a name="i_call">'<tt>call</tt>' Instruction</a>
4259</div>
4260
4261<div class="doc_text">
4262
4263<h5>Syntax:</h5>
4264<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4265 &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]4266</pre>
4267
4268<h5>Overview:</h5>
4269
4270<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4271
4272<h5>Arguments:</h5>
4273
4274<p>This instruction requires several arguments:</p>
4275
4276<ol>
4277 <li>
4278 <p>The optional "tail" marker indicates whether the callee function accesses
4279 any allocas or varargs in the caller. If the "tail" marker is present, the
4280 function call is eligible for tail call optimization. Note that calls may
4281 be marked "tail" even if they do not occur before a <a
4282 href="#i_ret"><tt>ret</tt></a> instruction.
4283 </li>
4284 <li>
4285 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4286 convention</a> the call should use. If none is specified, the call defaults
4287 to using C calling conventions.
4288 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4289
4290 <li>
4291 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4292 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4293 and '<tt>inreg</tt>' attributes are valid here.</p>
4294 </li>
4295
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4296 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4297 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4298 the type of the return value. Functions that return no value are marked
4299 <tt><a href="#t_void">void</a></tt>.</p>
4300 </li>
4301 <li>
4302 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4303 value being invoked. The argument types must match the types implied by
4304 this signature. This type can be omitted if the function is not varargs
4305 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4306 </li>
4307 <li>
4308 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4309 be invoked. In most cases, this is a direct function invocation, but
4310 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4311 to function value.</p>
4312 </li>
4313 <li>
4314 <p>'<tt>function args</tt>': argument list whose types match the
4315 function signature argument types. All arguments must be of
4316 <a href="#t_firstclass">first class</a> type. If the function signature
4317 indicates the function accepts a variable number of arguments, the extra
4318 arguments can be specified.</p>
4319 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4320 <li>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4321 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4322 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4323 '<tt>readnone</tt>' attributes are valid here.</p>
4324 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4325</ol>
4326
4327<h5>Semantics:</h5>
4328
4329<p>The '<tt>call</tt>' instruction is used to cause control flow to
4330transfer to a specified function, with its incoming arguments bound to
4331the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4332instruction in the called function, control flow continues with the
4333instruction after the function call, and the return value of the
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4334function is bound to the result argument.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4335
4336<h5>Example:</h5>
4337
4338<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4339 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4340 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4341 %X = tail call i32 @foo() <i>; yields i32</i>
4342 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4343 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4344
4345 %struct.A = type { i32, i8 }
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4346 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4347 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4348 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattnerac454b32008-10-08 06:26:11 +0000[diff] [blame]4349 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijman2c4e05a2008-10-07 10:03:45 +0000[diff] [blame]4350 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4351</pre>
4352
4353</div>
4354
4355<!-- _______________________________________________________________________ -->
4356<div class="doc_subsubsection">
4357 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4358</div>
4359
4360<div class="doc_text">
4361
4362<h5>Syntax:</h5>
4363
4364<pre>
4365 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4366</pre>
4367
4368<h5>Overview:</h5>
4369
4370<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4371the "variable argument" area of a function call. It is used to implement the
4372<tt>va_arg</tt> macro in C.</p>
4373
4374<h5>Arguments:</h5>
4375
4376<p>This instruction takes a <tt>va_list*</tt> value and the type of
4377the argument. It returns a value of the specified argument type and
4378increments the <tt>va_list</tt> to point to the next argument. The
4379actual type of <tt>va_list</tt> is target specific.</p>
4380
4381<h5>Semantics:</h5>
4382
4383<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4384type from the specified <tt>va_list</tt> and causes the
4385<tt>va_list</tt> to point to the next argument. For more information,
4386see the variable argument handling <a href="#int_varargs">Intrinsic
4387Functions</a>.</p>
4388
4389<p>It is legal for this instruction to be called in a function which does not
4390take a variable number of arguments, for example, the <tt>vfprintf</tt>
4391function.</p>
4392
4393<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4394href="#intrinsics">intrinsic function</a> because it takes a type as an
4395argument.</p>
4396
4397<h5>Example:</h5>
4398
4399<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4400
4401</div>
4402
4403<!-- *********************************************************************** -->
4404<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4405<!-- *********************************************************************** -->
4406
4407<div class="doc_text">
4408
4409<p>LLVM supports the notion of an "intrinsic function". These functions have
4410well known names and semantics and are required to follow certain restrictions.
4411Overall, these intrinsics represent an extension mechanism for the LLVM
4412language that does not require changing all of the transformations in LLVM when
4413adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4414
4415<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4416prefix is reserved in LLVM for intrinsic names; thus, function names may not
4417begin with this prefix. Intrinsic functions must always be external functions:
4418you cannot define the body of intrinsic functions. Intrinsic functions may
4419only be used in call or invoke instructions: it is illegal to take the address
4420of an intrinsic function. Additionally, because intrinsic functions are part
4421of the LLVM language, it is required if any are added that they be documented
4422here.</p>
4423
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4424<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4425a family of functions that perform the same operation but on different data
4426types. Because LLVM can represent over 8 million different integer types,
4427overloading is used commonly to allow an intrinsic function to operate on any
4428integer type. One or more of the argument types or the result type can be
4429overloaded to accept any integer type. Argument types may also be defined as
4430exactly matching a previous argument's type or the result type. This allows an
4431intrinsic function which accepts multiple arguments, but needs all of them to
4432be of the same type, to only be overloaded with respect to a single argument or
4433the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4434
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4435<p>Overloaded intrinsics will have the names of its overloaded argument types
4436encoded into its function name, each preceded by a period. Only those types
4437which are overloaded result in a name suffix. Arguments whose type is matched
4438against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4439take an integer of any width and returns an integer of exactly the same integer
4440width. This leads to a family of functions such as
4441<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4442Only one type, the return type, is overloaded, and only one type suffix is
4443required. Because the argument's type is matched against the return type, it
4444does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4445
4446<p>To learn how to add an intrinsic function, please see the
4447<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4448</p>
4449
4450</div>
4451
4452<!-- ======================================================================= -->
4453<div class="doc_subsection">
4454 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4455</div>
4456
4457<div class="doc_text">
4458
4459<p>Variable argument support is defined in LLVM with the <a
4460 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4461intrinsic functions. These functions are related to the similarly
4462named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4463
4464<p>All of these functions operate on arguments that use a
4465target-specific value type "<tt>va_list</tt>". The LLVM assembly
4466language reference manual does not define what this type is, so all
4467transformations should be prepared to handle these functions regardless of
4468the type used.</p>
4469
4470<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4471instruction and the variable argument handling intrinsic functions are
4472used.</p>
4473
4474<div class="doc_code">
4475<pre>
4476define i32 @test(i32 %X, ...) {
4477 ; Initialize variable argument processing
4478 %ap = alloca i8*
4479 %ap2 = bitcast i8** %ap to i8*
4480 call void @llvm.va_start(i8* %ap2)
4481
4482 ; Read a single integer argument
4483 %tmp = va_arg i8** %ap, i32
4484
4485 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4486 %aq = alloca i8*
4487 %aq2 = bitcast i8** %aq to i8*
4488 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4489 call void @llvm.va_end(i8* %aq2)
4490
4491 ; Stop processing of arguments.
4492 call void @llvm.va_end(i8* %ap2)
4493 ret i32 %tmp
4494}
4495
4496declare void @llvm.va_start(i8*)
4497declare void @llvm.va_copy(i8*, i8*)
4498declare void @llvm.va_end(i8*)
4499</pre>
4500</div>
4501
4502</div>
4503
4504<!-- _______________________________________________________________________ -->
4505<div class="doc_subsubsection">
4506 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4507</div>
4508
4509
4510<div class="doc_text">
4511<h5>Syntax:</h5>
4512<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4513<h5>Overview:</h5>
4514<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
4515<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4516href="#i_va_arg">va_arg</a></tt>.</p>
4517
4518<h5>Arguments:</h5>
4519
4520<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4521
4522<h5>Semantics:</h5>
4523
4524<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4525macro available in C. In a target-dependent way, it initializes the
4526<tt>va_list</tt> element to which the argument points, so that the next call to
4527<tt>va_arg</tt> will produce the first variable argument passed to the function.
4528Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4529last argument of the function as the compiler can figure that out.</p>
4530
4531</div>
4532
4533<!-- _______________________________________________________________________ -->
4534<div class="doc_subsubsection">
4535 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4536</div>
4537
4538<div class="doc_text">
4539<h5>Syntax:</h5>
4540<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4541<h5>Overview:</h5>
4542
4543<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4544which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4545or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4546
4547<h5>Arguments:</h5>
4548
4549<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4550
4551<h5>Semantics:</h5>
4552
4553<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4554macro available in C. In a target-dependent way, it destroys the
4555<tt>va_list</tt> element to which the argument points. Calls to <a
4556href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4557<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4558<tt>llvm.va_end</tt>.</p>
4559
4560</div>
4561
4562<!-- _______________________________________________________________________ -->
4563<div class="doc_subsubsection">
4564 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4565</div>
4566
4567<div class="doc_text">
4568
4569<h5>Syntax:</h5>
4570
4571<pre>
4572 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4573</pre>
4574
4575<h5>Overview:</h5>
4576
4577<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4578from the source argument list to the destination argument list.</p>
4579
4580<h5>Arguments:</h5>
4581
4582<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4583The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4584
4585
4586<h5>Semantics:</h5>
4587
4588<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4589macro available in C. In a target-dependent way, it copies the source
4590<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4591intrinsic is necessary because the <tt><a href="#int_va_start">
4592llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4593example, memory allocation.</p>
4594
4595</div>
4596
4597<!-- ======================================================================= -->
4598<div class="doc_subsection">
4599 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4600</div>
4601
4602<div class="doc_text">
4603
4604<p>
4605LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4606Collection</a> (GC) requires the implementation and generation of these
4607intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4608These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4609stack</a>, as well as garbage collector implementations that require <a
4610href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4611Front-ends for type-safe garbage collected languages should generate these
4612intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4613href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4614</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4615
4616<p>The garbage collection intrinsics only operate on objects in the generic
4617 address space (address space zero).</p>
4618
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4619</div>
4620
4621<!-- _______________________________________________________________________ -->
4622<div class="doc_subsubsection">
4623 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4624</div>
4625
4626<div class="doc_text">
4627
4628<h5>Syntax:</h5>
4629
4630<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4631 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4632</pre>
4633
4634<h5>Overview:</h5>
4635
4636<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4637the code generator, and allows some metadata to be associated with it.</p>
4638
4639<h5>Arguments:</h5>
4640
4641<p>The first argument specifies the address of a stack object that contains the
4642root pointer. The second pointer (which must be either a constant or a global
4643value address) contains the meta-data to be associated with the root.</p>
4644
4645<h5>Semantics:</h5>
4646
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4647<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]4648location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4649the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4650intrinsic may only be used in a function which <a href="#gc">specifies a GC
4651algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4652
4653</div>
4654
4655
4656<!-- _______________________________________________________________________ -->
4657<div class="doc_subsubsection">
4658 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4659</div>
4660
4661<div class="doc_text">
4662
4663<h5>Syntax:</h5>
4664
4665<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4666 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4667</pre>
4668
4669<h5>Overview:</h5>
4670
4671<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4672locations, allowing garbage collector implementations that require read
4673barriers.</p>
4674
4675<h5>Arguments:</h5>
4676
4677<p>The second argument is the address to read from, which should be an address
4678allocated from the garbage collector. The first object is a pointer to the
4679start of the referenced object, if needed by the language runtime (otherwise
4680null).</p>
4681
4682<h5>Semantics:</h5>
4683
4684<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4685instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4686garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4687may only be used in a function which <a href="#gc">specifies a GC
4688algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4689
4690</div>
4691
4692
4693<!-- _______________________________________________________________________ -->
4694<div class="doc_subsubsection">
4695 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4696</div>
4697
4698<div class="doc_text">
4699
4700<h5>Syntax:</h5>
4701
4702<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4703 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4704</pre>
4705
4706<h5>Overview:</h5>
4707
4708<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4709locations, allowing garbage collector implementations that require write
4710barriers (such as generational or reference counting collectors).</p>
4711
4712<h5>Arguments:</h5>
4713
4714<p>The first argument is the reference to store, the second is the start of the
4715object to store it to, and the third is the address of the field of Obj to
4716store to. If the runtime does not require a pointer to the object, Obj may be
4717null.</p>
4718
4719<h5>Semantics:</h5>
4720
4721<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4722instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4723garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4724may only be used in a function which <a href="#gc">specifies a GC
4725algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4726
4727</div>
4728
4729
4730
4731<!-- ======================================================================= -->
4732<div class="doc_subsection">
4733 <a name="int_codegen">Code Generator Intrinsics</a>
4734</div>
4735
4736<div class="doc_text">
4737<p>
4738These intrinsics are provided by LLVM to expose special features that may only
4739be implemented with code generator support.
4740</p>
4741
4742</div>
4743
4744<!-- _______________________________________________________________________ -->
4745<div class="doc_subsubsection">
4746 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4747</div>
4748
4749<div class="doc_text">
4750
4751<h5>Syntax:</h5>
4752<pre>
4753 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4754</pre>
4755
4756<h5>Overview:</h5>
4757
4758<p>
4759The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4760target-specific value indicating the return address of the current function
4761or one of its callers.
4762</p>
4763
4764<h5>Arguments:</h5>
4765
4766<p>
4767The argument to this intrinsic indicates which function to return the address
4768for. Zero indicates the calling function, one indicates its caller, etc. The
4769argument is <b>required</b> to be a constant integer value.
4770</p>
4771
4772<h5>Semantics:</h5>
4773
4774<p>
4775The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4776the return address of the specified call frame, or zero if it cannot be
4777identified. The value returned by this intrinsic is likely to be incorrect or 0
4778for arguments other than zero, so it should only be used for debugging purposes.
4779</p>
4780
4781<p>
4782Note that calling this intrinsic does not prevent function inlining or other
4783aggressive transformations, so the value returned may not be that of the obvious
4784source-language caller.
4785</p>
4786</div>
4787
4788
4789<!-- _______________________________________________________________________ -->
4790<div class="doc_subsubsection">
4791 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4792</div>
4793
4794<div class="doc_text">
4795
4796<h5>Syntax:</h5>
4797<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4798 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4799</pre>
4800
4801<h5>Overview:</h5>
4802
4803<p>
4804The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4805target-specific frame pointer value for the specified stack frame.
4806</p>
4807
4808<h5>Arguments:</h5>
4809
4810<p>
4811The argument to this intrinsic indicates which function to return the frame
4812pointer for. Zero indicates the calling function, one indicates its caller,
4813etc. The argument is <b>required</b> to be a constant integer value.
4814</p>
4815
4816<h5>Semantics:</h5>
4817
4818<p>
4819The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4820the frame address of the specified call frame, or zero if it cannot be
4821identified. The value returned by this intrinsic is likely to be incorrect or 0
4822for arguments other than zero, so it should only be used for debugging purposes.
4823</p>
4824
4825<p>
4826Note that calling this intrinsic does not prevent function inlining or other
4827aggressive transformations, so the value returned may not be that of the obvious
4828source-language caller.
4829</p>
4830</div>
4831
4832<!-- _______________________________________________________________________ -->
4833<div class="doc_subsubsection">
4834 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4835</div>
4836
4837<div class="doc_text">
4838
4839<h5>Syntax:</h5>
4840<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4841 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4842</pre>
4843
4844<h5>Overview:</h5>
4845
4846<p>
4847The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4848the function stack, for use with <a href="#int_stackrestore">
4849<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4850features like scoped automatic variable sized arrays in C99.
4851</p>
4852
4853<h5>Semantics:</h5>
4854
4855<p>
4856This intrinsic returns a opaque pointer value that can be passed to <a
4857href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4858<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4859<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4860state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4861practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4862that were allocated after the <tt>llvm.stacksave</tt> was executed.
4863</p>
4864
4865</div>
4866
4867<!-- _______________________________________________________________________ -->
4868<div class="doc_subsubsection">
4869 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4870</div>
4871
4872<div class="doc_text">
4873
4874<h5>Syntax:</h5>
4875<pre>
4876 declare void @llvm.stackrestore(i8 * %ptr)
4877</pre>
4878
4879<h5>Overview:</h5>
4880
4881<p>
4882The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4883the function stack to the state it was in when the corresponding <a
4884href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4885useful for implementing language features like scoped automatic variable sized
4886arrays in C99.
4887</p>
4888
4889<h5>Semantics:</h5>
4890
4891<p>
4892See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4893</p>
4894
4895</div>
4896
4897
4898<!-- _______________________________________________________________________ -->
4899<div class="doc_subsubsection">
4900 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4901</div>
4902
4903<div class="doc_text">
4904
4905<h5>Syntax:</h5>
4906<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4907 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]4908</pre>
4909
4910<h5>Overview:</h5>
4911
4912
4913<p>
4914The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4915a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4916no
4917effect on the behavior of the program but can change its performance
4918characteristics.
4919</p>
4920
4921<h5>Arguments:</h5>
4922
4923<p>
4924<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4925determining if the fetch should be for a read (0) or write (1), and
4926<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4927locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4928<tt>locality</tt> arguments must be constant integers.
4929</p>
4930
4931<h5>Semantics:</h5>
4932
4933<p>
4934This intrinsic does not modify the behavior of the program. In particular,
4935prefetches cannot trap and do not produce a value. On targets that support this
4936intrinsic, the prefetch can provide hints to the processor cache for better
4937performance.
4938</p>
4939
4940</div>
4941
4942<!-- _______________________________________________________________________ -->
4943<div class="doc_subsubsection">
4944 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4945</div>
4946
4947<div class="doc_text">
4948
4949<h5>Syntax:</h5>
4950<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4951 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4952</pre>
4953
4954<h5>Overview:</h5>
4955
4956
4957<p>
4958The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4959(PC) in a region of
4960code to simulators and other tools. The method is target specific, but it is
4961expected that the marker will use exported symbols to transmit the PC of the
4962marker.
4963The marker makes no guarantees that it will remain with any specific instruction
4964after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4965optimizations. The intended use is to be inserted after optimizations to allow
4966correlations of simulation runs.
4967</p>
4968
4969<h5>Arguments:</h5>
4970
4971<p>
4972<tt>id</tt> is a numerical id identifying the marker.
4973</p>
4974
4975<h5>Semantics:</h5>
4976
4977<p>
4978This intrinsic does not modify the behavior of the program. Backends that do not
4979support this intrinisic may ignore it.
4980</p>
4981
4982</div>
4983
4984<!-- _______________________________________________________________________ -->
4985<div class="doc_subsubsection">
4986 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4987</div>
4988
4989<div class="doc_text">
4990
4991<h5>Syntax:</h5>
4992<pre>
4993 declare i64 @llvm.readcyclecounter( )
4994</pre>
4995
4996<h5>Overview:</h5>
4997
4998
4999<p>
5000The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5001counter register (or similar low latency, high accuracy clocks) on those targets
5002that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5003As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5004should only be used for small timings.
5005</p>
5006
5007<h5>Semantics:</h5>
5008
5009<p>
5010When directly supported, reading the cycle counter should not modify any memory.
5011Implementations are allowed to either return a application specific value or a
5012system wide value. On backends without support, this is lowered to a constant 0.
5013</p>
5014
5015</div>
5016
5017<!-- ======================================================================= -->
5018<div class="doc_subsection">
5019 <a name="int_libc">Standard C Library Intrinsics</a>
5020</div>
5021
5022<div class="doc_text">
5023<p>
5024LLVM provides intrinsics for a few important standard C library functions.
5025These intrinsics allow source-language front-ends to pass information about the
5026alignment of the pointer arguments to the code generator, providing opportunity
5027for more efficient code generation.
5028</p>
5029
5030</div>
5031
5032<!-- _______________________________________________________________________ -->
5033<div class="doc_subsubsection">
5034 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5035</div>
5036
5037<div class="doc_text">
5038
5039<h5>Syntax:</h5>
5040<pre>
5041 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5042 i32 &lt;len&gt;, i32 &lt;align&gt;)
5043 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5044 i64 &lt;len&gt;, i32 &lt;align&gt;)
5045</pre>
5046
5047<h5>Overview:</h5>
5048
5049<p>
5050The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5051location to the destination location.
5052</p>
5053
5054<p>
5055Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5056intrinsics do not return a value, and takes an extra alignment argument.
5057</p>
5058
5059<h5>Arguments:</h5>
5060
5061<p>
5062The first argument is a pointer to the destination, the second is a pointer to
5063the source. The third argument is an integer argument
5064specifying the number of bytes to copy, and the fourth argument is the alignment
5065of the source and destination locations.
5066</p>
5067
5068<p>
5069If the call to this intrinisic has an alignment value that is not 0 or 1, then
5070the caller guarantees that both the source and destination pointers are aligned
5071to that boundary.
5072</p>
5073
5074<h5>Semantics:</h5>
5075
5076<p>
5077The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5078location to the destination location, which are not allowed to overlap. It
5079copies "len" bytes of memory over. If the argument is known to be aligned to
5080some boundary, this can be specified as the fourth argument, otherwise it should
5081be set to 0 or 1.
5082</p>
5083</div>
5084
5085
5086<!-- _______________________________________________________________________ -->
5087<div class="doc_subsubsection">
5088 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5089</div>
5090
5091<div class="doc_text">
5092
5093<h5>Syntax:</h5>
5094<pre>
5095 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5096 i32 &lt;len&gt;, i32 &lt;align&gt;)
5097 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5098 i64 &lt;len&gt;, i32 &lt;align&gt;)
5099</pre>
5100
5101<h5>Overview:</h5>
5102
5103<p>
5104The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5105location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5106'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5107</p>
5108
5109<p>
5110Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5111intrinsics do not return a value, and takes an extra alignment argument.
5112</p>
5113
5114<h5>Arguments:</h5>
5115
5116<p>
5117The first argument is a pointer to the destination, the second is a pointer to
5118the source. The third argument is an integer argument
5119specifying the number of bytes to copy, and the fourth argument is the alignment
5120of the source and destination locations.
5121</p>
5122
5123<p>
5124If the call to this intrinisic has an alignment value that is not 0 or 1, then
5125the caller guarantees that the source and destination pointers are aligned to
5126that boundary.
5127</p>
5128
5129<h5>Semantics:</h5>
5130
5131<p>
5132The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5133location to the destination location, which may overlap. It
5134copies "len" bytes of memory over. If the argument is known to be aligned to
5135some boundary, this can be specified as the fourth argument, otherwise it should
5136be set to 0 or 1.
5137</p>
5138</div>
5139
5140
5141<!-- _______________________________________________________________________ -->
5142<div class="doc_subsubsection">
5143 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5144</div>
5145
5146<div class="doc_text">
5147
5148<h5>Syntax:</h5>
5149<pre>
5150 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5151 i32 &lt;len&gt;, i32 &lt;align&gt;)
5152 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5153 i64 &lt;len&gt;, i32 &lt;align&gt;)
5154</pre>
5155
5156<h5>Overview:</h5>
5157
5158<p>
5159The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5160byte value.
5161</p>
5162
5163<p>
5164Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5165does not return a value, and takes an extra alignment argument.
5166</p>
5167
5168<h5>Arguments:</h5>
5169
5170<p>
5171The first argument is a pointer to the destination to fill, the second is the
5172byte value to fill it with, the third argument is an integer
5173argument specifying the number of bytes to fill, and the fourth argument is the
5174known alignment of destination location.
5175</p>
5176
5177<p>
5178If the call to this intrinisic has an alignment value that is not 0 or 1, then
5179the caller guarantees that the destination pointer is aligned to that boundary.
5180</p>
5181
5182<h5>Semantics:</h5>
5183
5184<p>
5185The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5186the
5187destination location. If the argument is known to be aligned to some boundary,
5188this can be specified as the fourth argument, otherwise it should be set to 0 or
51891.
5190</p>
5191</div>
5192
5193
5194<!-- _______________________________________________________________________ -->
5195<div class="doc_subsubsection">
5196 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5197</div>
5198
5199<div class="doc_text">
5200
5201<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5202<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]5203floating point or vector of floating point type. Not all targets support all
5204types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5205<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5206 declare float @llvm.sqrt.f32(float %Val)
5207 declare double @llvm.sqrt.f64(double %Val)
5208 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5209 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5210 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5211</pre>
5212
5213<h5>Overview:</h5>
5214
5215<p>
5216The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5217returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5218<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5219negative numbers other than -0.0 (which allows for better optimization, because
5220there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5221defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5222</p>
5223
5224<h5>Arguments:</h5>
5225
5226<p>
5227The argument and return value are floating point numbers of the same type.
5228</p>
5229
5230<h5>Semantics:</h5>
5231
5232<p>
5233This function returns the sqrt of the specified operand if it is a nonnegative
5234floating point number.
5235</p>
5236</div>
5237
5238<!-- _______________________________________________________________________ -->
5239<div class="doc_subsubsection">
5240 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5241</div>
5242
5243<div class="doc_text">
5244
5245<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5246<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]5247floating point or vector of floating point type. Not all targets support all
5248types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5249<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5250 declare float @llvm.powi.f32(float %Val, i32 %power)
5251 declare double @llvm.powi.f64(double %Val, i32 %power)
5252 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5253 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5254 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5255</pre>
5256
5257<h5>Overview:</h5>
5258
5259<p>
5260The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5261specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5262multiplications is not defined. When a vector of floating point type is
5263used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5264</p>
5265
5266<h5>Arguments:</h5>
5267
5268<p>
5269The second argument is an integer power, and the first is a value to raise to
5270that power.
5271</p>
5272
5273<h5>Semantics:</h5>
5274
5275<p>
5276This function returns the first value raised to the second power with an
5277unspecified sequence of rounding operations.</p>
5278</div>
5279
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5280<!-- _______________________________________________________________________ -->
5281<div class="doc_subsubsection">
5282 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5283</div>
5284
5285<div class="doc_text">
5286
5287<h5>Syntax:</h5>
5288<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5289floating point or vector of floating point type. Not all targets support all
5290types however.
5291<pre>
5292 declare float @llvm.sin.f32(float %Val)
5293 declare double @llvm.sin.f64(double %Val)
5294 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5295 declare fp128 @llvm.sin.f128(fp128 %Val)
5296 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5297</pre>
5298
5299<h5>Overview:</h5>
5300
5301<p>
5302The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5303</p>
5304
5305<h5>Arguments:</h5>
5306
5307<p>
5308The argument and return value are floating point numbers of the same type.
5309</p>
5310
5311<h5>Semantics:</h5>
5312
5313<p>
5314This function returns the sine of the specified operand, returning the
5315same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5316conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5317</div>
5318
5319<!-- _______________________________________________________________________ -->
5320<div class="doc_subsubsection">
5321 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5322</div>
5323
5324<div class="doc_text">
5325
5326<h5>Syntax:</h5>
5327<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5328floating point or vector of floating point type. Not all targets support all
5329types however.
5330<pre>
5331 declare float @llvm.cos.f32(float %Val)
5332 declare double @llvm.cos.f64(double %Val)
5333 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5334 declare fp128 @llvm.cos.f128(fp128 %Val)
5335 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5336</pre>
5337
5338<h5>Overview:</h5>
5339
5340<p>
5341The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5342</p>
5343
5344<h5>Arguments:</h5>
5345
5346<p>
5347The argument and return value are floating point numbers of the same type.
5348</p>
5349
5350<h5>Semantics:</h5>
5351
5352<p>
5353This function returns the cosine of the specified operand, returning the
5354same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5355conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5356</div>
5357
5358<!-- _______________________________________________________________________ -->
5359<div class="doc_subsubsection">
5360 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5361</div>
5362
5363<div class="doc_text">
5364
5365<h5>Syntax:</h5>
5366<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5367floating point or vector of floating point type. Not all targets support all
5368types however.
5369<pre>
5370 declare float @llvm.pow.f32(float %Val, float %Power)
5371 declare double @llvm.pow.f64(double %Val, double %Power)
5372 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5373 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5374 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5375</pre>
5376
5377<h5>Overview:</h5>
5378
5379<p>
5380The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5381specified (positive or negative) power.
5382</p>
5383
5384<h5>Arguments:</h5>
5385
5386<p>
5387The second argument is a floating point power, and the first is a value to
5388raise to that power.
5389</p>
5390
5391<h5>Semantics:</h5>
5392
5393<p>
5394This function returns the first value raised to the second power,
5395returning the
5396same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5397conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5398</div>
5399
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5400
5401<!-- ======================================================================= -->
5402<div class="doc_subsection">
5403 <a name="int_manip">Bit Manipulation Intrinsics</a>
5404</div>
5405
5406<div class="doc_text">
5407<p>
5408LLVM provides intrinsics for a few important bit manipulation operations.
5409These allow efficient code generation for some algorithms.
5410</p>
5411
5412</div>
5413
5414<!-- _______________________________________________________________________ -->
5415<div class="doc_subsubsection">
5416 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5417</div>
5418
5419<div class="doc_text">
5420
5421<h5>Syntax:</h5>
5422<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5423type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5424<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5425 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5426 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5427 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5428</pre>
5429
5430<h5>Overview:</h5>
5431
5432<p>
5433The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5434values with an even number of bytes (positive multiple of 16 bits). These are
5435useful for performing operations on data that is not in the target's native
5436byte order.
5437</p>
5438
5439<h5>Semantics:</h5>
5440
5441<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5442The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5443and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5444intrinsic returns an i32 value that has the four bytes of the input i32
5445swapped, 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]5446i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5447<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5448additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5449</p>
5450
5451</div>
5452
5453<!-- _______________________________________________________________________ -->
5454<div class="doc_subsubsection">
5455 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5456</div>
5457
5458<div class="doc_text">
5459
5460<h5>Syntax:</h5>
5461<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5462width. Not all targets support all bit widths however.
5463<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5464 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
5465 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5466 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5467 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5468 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5469</pre>
5470
5471<h5>Overview:</h5>
5472
5473<p>
5474The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5475value.
5476</p>
5477
5478<h5>Arguments:</h5>
5479
5480<p>
5481The only argument is the value to be counted. The argument may be of any
5482integer type. The return type must match the argument type.
5483</p>
5484
5485<h5>Semantics:</h5>
5486
5487<p>
5488The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5489</p>
5490</div>
5491
5492<!-- _______________________________________________________________________ -->
5493<div class="doc_subsubsection">
5494 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5495</div>
5496
5497<div class="doc_text">
5498
5499<h5>Syntax:</h5>
5500<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5501integer bit width. Not all targets support all bit widths however.
5502<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5503 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5504 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5505 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5506 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5507 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5508</pre>
5509
5510<h5>Overview:</h5>
5511
5512<p>
5513The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5514leading zeros in a variable.
5515</p>
5516
5517<h5>Arguments:</h5>
5518
5519<p>
5520The only argument is the value to be counted. The argument may be of any
5521integer type. The return type must match the argument type.
5522</p>
5523
5524<h5>Semantics:</h5>
5525
5526<p>
5527The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5528in a variable. If the src == 0 then the result is the size in bits of the type
5529of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5530</p>
5531</div>
5532
5533
5534
5535<!-- _______________________________________________________________________ -->
5536<div class="doc_subsubsection">
5537 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5538</div>
5539
5540<div class="doc_text">
5541
5542<h5>Syntax:</h5>
5543<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5544integer bit width. Not all targets support all bit widths however.
5545<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5546 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5547 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5548 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5549 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5550 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5551</pre>
5552
5553<h5>Overview:</h5>
5554
5555<p>
5556The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5557trailing zeros.
5558</p>
5559
5560<h5>Arguments:</h5>
5561
5562<p>
5563The only argument is the value to be counted. The argument may be of any
5564integer type. The return type must match the argument type.
5565</p>
5566
5567<h5>Semantics:</h5>
5568
5569<p>
5570The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5571in a variable. If the src == 0 then the result is the size in bits of the type
5572of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5573</p>
5574</div>
5575
5576<!-- _______________________________________________________________________ -->
5577<div class="doc_subsubsection">
5578 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5579</div>
5580
5581<div class="doc_text">
5582
5583<h5>Syntax:</h5>
5584<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5585on any integer bit width.
5586<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5587 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5588 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5589</pre>
5590
5591<h5>Overview:</h5>
5592<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5593range of bits from an integer value and returns them in the same bit width as
5594the original value.</p>
5595
5596<h5>Arguments:</h5>
5597<p>The first argument, <tt>%val</tt> and the result may be integer types of
5598any bit width but they must have the same bit width. The second and third
5599arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5600
5601<h5>Semantics:</h5>
5602<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5603of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5604<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5605operates in forward mode.</p>
5606<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5607right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5608only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5609<ol>
5610 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5611 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5612 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5613 to determine the number of bits to retain.</li>
5614 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5615 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5616</ol>
5617<p>In reverse mode, a similar computation is made except that the bits are
5618returned in the reverse order. So, for example, if <tt>X</tt> has the value
5619<tt>i16 0x0ACF (101011001111)</tt> and we apply
5620<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5621<tt>i16 0x0026 (000000100110)</tt>.</p>
5622</div>
5623
5624<div class="doc_subsubsection">
5625 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5626</div>
5627
5628<div class="doc_text">
5629
5630<h5>Syntax:</h5>
5631<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5632on any integer bit width.
5633<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5634 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5635 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]5636</pre>
5637
5638<h5>Overview:</h5>
5639<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5640of bits in an integer value with another integer value. It returns the integer
5641with the replaced bits.</p>
5642
5643<h5>Arguments:</h5>
5644<p>The first argument, <tt>%val</tt> and the result may be integer types of
5645any bit width but they must have the same bit width. <tt>%val</tt> is the value
5646whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5647integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5648type since they specify only a bit index.</p>
5649
5650<h5>Semantics:</h5>
5651<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5652of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5653<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5654operates in forward mode.</p>
5655<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5656truncating it down to the size of the replacement area or zero extending it
5657up to that size.</p>
5658<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5659are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5660in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5661to the <tt>%hi</tt>th bit.
5662<p>In reverse mode, a similar computation is made except that the bits are
5663reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5664<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5665<h5>Examples:</h5>
5666<pre>
5667 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5668 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5669 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5670 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5671 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5672</pre>
5673</div>
5674
5675<!-- ======================================================================= -->
5676<div class="doc_subsection">
5677 <a name="int_debugger">Debugger Intrinsics</a>
5678</div>
5679
5680<div class="doc_text">
5681<p>
5682The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5683are described in the <a
5684href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5685Debugging</a> document.
5686</p>
5687</div>
5688
5689
5690<!-- ======================================================================= -->
5691<div class="doc_subsection">
5692 <a name="int_eh">Exception Handling Intrinsics</a>
5693</div>
5694
5695<div class="doc_text">
5696<p> The LLVM exception handling intrinsics (which all start with
5697<tt>llvm.eh.</tt> prefix), are described in the <a
5698href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5699Handling</a> document. </p>
5700</div>
5701
5702<!-- ======================================================================= -->
5703<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5704 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5705</div>
5706
5707<div class="doc_text">
5708<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5709 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5710 the <tt>nest</tt> attribute, from a function. The result is a callable
5711 function pointer lacking the nest parameter - the caller does not need
5712 to provide a value for it. Instead, the value to use is stored in
5713 advance in a "trampoline", a block of memory usually allocated
5714 on the stack, which also contains code to splice the nest value into the
5715 argument list. This is used to implement the GCC nested function address
5716 extension.
5717</p>
5718<p>
5719 For example, if the function is
5720 <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]5721 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]5722<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5723 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5724 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5725 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5726 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5727</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5728 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5729 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5730</div>
5731
5732<!-- _______________________________________________________________________ -->
5733<div class="doc_subsubsection">
5734 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5735</div>
5736<div class="doc_text">
5737<h5>Syntax:</h5>
5738<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5739declare 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]5740</pre>
5741<h5>Overview:</h5>
5742<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5743 This fills the memory pointed to by <tt>tramp</tt> with code
5744 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5745</p>
5746<h5>Arguments:</h5>
5747<p>
5748 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5749 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5750 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5751 intrinsic. Note that the size and the alignment are target-specific - LLVM
5752 currently provides no portable way of determining them, so a front-end that
5753 generates this intrinsic needs to have some target-specific knowledge.
5754 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]5755</p>
5756<h5>Semantics:</h5>
5757<p>
5758 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5759 dependent code, turning it into a function. A pointer to this function is
5760 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5761 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5762 before being called. The new function's signature is the same as that of
5763 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5764 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5765 of pointer type. Calling the new function is equivalent to calling
5766 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5767 missing <tt>nest</tt> argument. If, after calling
5768 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5769 modified, then the effect of any later call to the returned function pointer is
5770 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5771</p>
5772</div>
5773
5774<!-- ======================================================================= -->
5775<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5776 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5777</div>
5778
5779<div class="doc_text">
5780<p>
5781 These intrinsic functions expand the "universal IR" of LLVM to represent
5782 hardware constructs for atomic operations and memory synchronization. This
5783 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5784 is aimed at a low enough level to allow any programming models or APIs
5785 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5786 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5787 hardware behavior. Just as hardware provides a "universal IR" for source
5788 languages, it also provides a starting point for developing a "universal"
5789 atomic operation and synchronization IR.
5790</p>
5791<p>
5792 These do <em>not</em> form an API such as high-level threading libraries,
5793 software transaction memory systems, atomic primitives, and intrinsic
5794 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5795 application libraries. The hardware interface provided by LLVM should allow
5796 a clean implementation of all of these APIs and parallel programming models.
5797 No one model or paradigm should be selected above others unless the hardware
5798 itself ubiquitously does so.
5799
5800</p>
5801</div>
5802
5803<!-- _______________________________________________________________________ -->
5804<div class="doc_subsubsection">
5805 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5806</div>
5807<div class="doc_text">
5808<h5>Syntax:</h5>
5809<pre>
5810declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5811i1 &lt;device&gt; )
5812
5813</pre>
5814<h5>Overview:</h5>
5815<p>
5816 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5817 specific pairs of memory access types.
5818</p>
5819<h5>Arguments:</h5>
5820<p>
5821 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5822 The first four arguments enables a specific barrier as listed below. The fith
5823 argument specifies that the barrier applies to io or device or uncached memory.
5824
5825</p>
5826 <ul>
5827 <li><tt>ll</tt>: load-load barrier</li>
5828 <li><tt>ls</tt>: load-store barrier</li>
5829 <li><tt>sl</tt>: store-load barrier</li>
5830 <li><tt>ss</tt>: store-store barrier</li>
5831 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5832 </ul>
5833<h5>Semantics:</h5>
5834<p>
5835 This intrinsic causes the system to enforce some ordering constraints upon
5836 the loads and stores of the program. This barrier does not indicate
5837 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5838 which they occur. For any of the specified pairs of load and store operations
5839 (f.ex. load-load, or store-load), all of the first operations preceding the
5840 barrier will complete before any of the second operations succeeding the
5841 barrier begin. Specifically the semantics for each pairing is as follows:
5842</p>
5843 <ul>
5844 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5845 after the barrier begins.</li>
5846
5847 <li><tt>ls</tt>: All loads before the barrier must complete before any
5848 store after the barrier begins.</li>
5849 <li><tt>ss</tt>: All stores before the barrier must complete before any
5850 store after the barrier begins.</li>
5851 <li><tt>sl</tt>: All stores before the barrier must complete before any
5852 load after the barrier begins.</li>
5853 </ul>
5854<p>
5855 These semantics are applied with a logical "and" behavior when more than one
5856 is enabled in a single memory barrier intrinsic.
5857</p>
5858<p>
5859 Backends may implement stronger barriers than those requested when they do not
5860 support as fine grained a barrier as requested. Some architectures do not
5861 need all types of barriers and on such architectures, these become noops.
5862</p>
5863<h5>Example:</h5>
5864<pre>
5865%ptr = malloc i32
5866 store i32 4, %ptr
5867
5868%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5869 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5870 <i>; guarantee the above finishes</i>
5871 store i32 8, %ptr <i>; before this begins</i>
5872</pre>
5873</div>
5874
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5875<!-- _______________________________________________________________________ -->
5876<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5877 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5878</div>
5879<div class="doc_text">
5880<h5>Syntax:</h5>
5881<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5882 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
5883 any integer bit width and for different address spaces. Not all targets
5884 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5885
5886<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5887declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
5888declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
5889declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
5890declare 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]5891
5892</pre>
5893<h5>Overview:</h5>
5894<p>
5895 This loads a value in memory and compares it to a given value. If they are
5896 equal, it stores a new value into the memory.
5897</p>
5898<h5>Arguments:</h5>
5899<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5900 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5901 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5902 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5903 this integer type. While any bit width integer may be used, targets may only
5904 lower representations they support in hardware.
5905
5906</p>
5907<h5>Semantics:</h5>
5908<p>
5909 This entire intrinsic must be executed atomically. It first loads the value
5910 in memory pointed to by <tt>ptr</tt> and compares it with the value
5911 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5912 loaded value is yielded in all cases. This provides the equivalent of an
5913 atomic compare-and-swap operation within the SSA framework.
5914</p>
5915<h5>Examples:</h5>
5916
5917<pre>
5918%ptr = malloc i32
5919 store i32 4, %ptr
5920
5921%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5922%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5923 <i>; yields {i32}:result1 = 4</i>
5924%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5925%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5926
5927%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5928%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5929 <i>; yields {i32}:result2 = 8</i>
5930%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5931
5932%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5933</pre>
5934</div>
5935
5936<!-- _______________________________________________________________________ -->
5937<div class="doc_subsubsection">
5938 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5939</div>
5940<div class="doc_text">
5941<h5>Syntax:</h5>
5942
5943<p>
5944 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5945 integer bit width. Not all targets support all bit widths however.</p>
5946<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5947declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
5948declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
5949declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
5950declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5951
5952</pre>
5953<h5>Overview:</h5>
5954<p>
5955 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
5956 the value from memory. It then stores the value in <tt>val</tt> in the memory
5957 at <tt>ptr</tt>.
5958</p>
5959<h5>Arguments:</h5>
5960
5961<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5962 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5963 <tt>val</tt> argument and the result must be integers of the same bit width.
5964 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
5965 integer type. The targets may only lower integer representations they
5966 support.
5967</p>
5968<h5>Semantics:</h5>
5969<p>
5970 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
5971 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
5972 equivalent of an atomic swap operation within the SSA framework.
5973
5974</p>
5975<h5>Examples:</h5>
5976<pre>
5977%ptr = malloc i32
5978 store i32 4, %ptr
5979
5980%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5981%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5982 <i>; yields {i32}:result1 = 4</i>
5983%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5984%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5985
5986%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5987%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5988 <i>; yields {i32}:result2 = 8</i>
5989
5990%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
5991%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
5992</pre>
5993</div>
5994
5995<!-- _______________________________________________________________________ -->
5996<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5997 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5998
5999</div>
6000<div class="doc_text">
6001<h5>Syntax:</h5>
6002<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6003 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]6004 integer bit width. Not all targets support all bit widths however.</p>
6005<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6006declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6007declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6008declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6009declare 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]6010
6011</pre>
6012<h5>Overview:</h5>
6013<p>
6014 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6015 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6016</p>
6017<h5>Arguments:</h5>
6018<p>
6019
6020 The intrinsic takes two arguments, the first a pointer to an integer value
6021 and the second an integer value. The result is also an integer value. These
6022 integer types can have any bit width, but they must all have the same bit
6023 width. The targets may only lower integer representations they support.
6024</p>
6025<h5>Semantics:</h5>
6026<p>
6027 This intrinsic does a series of operations atomically. It first loads the
6028 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6029 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6030</p>
6031
6032<h5>Examples:</h5>
6033<pre>
6034%ptr = malloc i32
6035 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6036%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6037 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6038%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6039 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6040%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6041 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6042%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6043</pre>
6044</div>
6045
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6046<!-- _______________________________________________________________________ -->
6047<div class="doc_subsubsection">
6048 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6049
6050</div>
6051<div class="doc_text">
6052<h5>Syntax:</h5>
6053<p>
6054 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]6055 any integer bit width and for different address spaces. Not all targets
6056 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6057<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6058declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6059declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6060declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6061declare 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]6062
6063</pre>
6064<h5>Overview:</h5>
6065<p>
6066 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6067 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6068</p>
6069<h5>Arguments:</h5>
6070<p>
6071
6072 The intrinsic takes two arguments, the first a pointer to an integer value
6073 and the second an integer value. The result is also an integer value. These
6074 integer types can have any bit width, but they must all have the same bit
6075 width. The targets may only lower integer representations they support.
6076</p>
6077<h5>Semantics:</h5>
6078<p>
6079 This intrinsic does a series of operations atomically. It first loads the
6080 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6081 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6082</p>
6083
6084<h5>Examples:</h5>
6085<pre>
6086%ptr = malloc i32
6087 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6088%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6089 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6090%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6091 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6092%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6093 <i>; yields {i32}:result3 = 2</i>
6094%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6095</pre>
6096</div>
6097
6098<!-- _______________________________________________________________________ -->
6099<div class="doc_subsubsection">
6100 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6101 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6102 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6103 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6104
6105</div>
6106<div class="doc_text">
6107<h5>Syntax:</h5>
6108<p>
6109 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6110 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6111 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6112 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6113<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6114declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6115declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6116declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6117declare 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]6118
6119</pre>
6120
6121<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6122declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6123declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6124declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6125declare 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]6126
6127</pre>
6128
6129<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6130declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6131declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6132declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6133declare 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]6134
6135</pre>
6136
6137<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6138declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6139declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6140declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6141declare 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]6142
6143</pre>
6144<h5>Overview:</h5>
6145<p>
6146 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6147 the value stored in memory at <tt>ptr</tt>. It yields the original value
6148 at <tt>ptr</tt>.
6149</p>
6150<h5>Arguments:</h5>
6151<p>
6152
6153 These intrinsics take two arguments, the first a pointer to an integer value
6154 and the second an integer value. The result is also an integer value. These
6155 integer types can have any bit width, but they must all have the same bit
6156 width. The targets may only lower integer representations they support.
6157</p>
6158<h5>Semantics:</h5>
6159<p>
6160 These intrinsics does a series of operations atomically. They first load the
6161 value stored at <tt>ptr</tt>. They then do the bitwise operation
6162 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6163 value stored at <tt>ptr</tt>.
6164</p>
6165
6166<h5>Examples:</h5>
6167<pre>
6168%ptr = malloc i32
6169 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6170%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6171 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6172%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6173 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6174%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6175 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6176%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6177 <i>; yields {i32}:result3 = FF</i>
6178%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6179</pre>
6180</div>
6181
6182
6183<!-- _______________________________________________________________________ -->
6184<div class="doc_subsubsection">
6185 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6186 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6187 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6188 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6189
6190</div>
6191<div class="doc_text">
6192<h5>Syntax:</h5>
6193<p>
6194 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6195 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6196 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6197 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6198 support all bit widths however.</p>
6199<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6200declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6201declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6202declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6203declare 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]6204
6205</pre>
6206
6207<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6208declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6209declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6210declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6211declare 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]6212
6213</pre>
6214
6215<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6216declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6217declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6218declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6219declare 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]6220
6221</pre>
6222
6223<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6224declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6225declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6226declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6227declare 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]6228
6229</pre>
6230<h5>Overview:</h5>
6231<p>
6232 These intrinsics takes the signed or unsigned minimum or maximum of
6233 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6234 original value at <tt>ptr</tt>.
6235</p>
6236<h5>Arguments:</h5>
6237<p>
6238
6239 These intrinsics take two arguments, the first a pointer to an integer value
6240 and the second an integer value. The result is also an integer value. These
6241 integer types can have any bit width, but they must all have the same bit
6242 width. The targets may only lower integer representations they support.
6243</p>
6244<h5>Semantics:</h5>
6245<p>
6246 These intrinsics does a series of operations atomically. They first load the
6247 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6248 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6249 the original value stored at <tt>ptr</tt>.
6250</p>
6251
6252<h5>Examples:</h5>
6253<pre>
6254%ptr = malloc i32
6255 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6256%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6257 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6258%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6259 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6260%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6261 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6262%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6263 <i>; yields {i32}:result3 = 8</i>
6264%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6265</pre>
6266</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6267
6268<!-- ======================================================================= -->
6269<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6270 <a name="int_general">General Intrinsics</a>
6271</div>
6272
6273<div class="doc_text">
6274<p> This class of intrinsics is designed to be generic and has
6275no specific purpose. </p>
6276</div>
6277
6278<!-- _______________________________________________________________________ -->
6279<div class="doc_subsubsection">
6280 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6281</div>
6282
6283<div class="doc_text">
6284
6285<h5>Syntax:</h5>
6286<pre>
6287 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6288</pre>
6289
6290<h5>Overview:</h5>
6291
6292<p>
6293The '<tt>llvm.var.annotation</tt>' intrinsic
6294</p>
6295
6296<h5>Arguments:</h5>
6297
6298<p>
6299The first argument is a pointer to a value, the second is a pointer to a
6300global string, the third is a pointer to a global string which is the source
6301file name, and the last argument is the line number.
6302</p>
6303
6304<h5>Semantics:</h5>
6305
6306<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6307This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6308This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6309annotations. These have no other defined use, they are ignored by code
6310generation and optimization.
6311</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6312</div>
6313
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6314<!-- _______________________________________________________________________ -->
6315<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6316 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6317</div>
6318
6319<div class="doc_text">
6320
6321<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6322<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6323any integer bit width.
6324</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6325<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6326 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6327 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6328 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6329 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6330 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]6331</pre>
6332
6333<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6334
6335<p>
6336The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6337</p>
6338
6339<h5>Arguments:</h5>
6340
6341<p>
6342The first argument is an integer value (result of some expression),
6343the second is a pointer to a global string, the third is a pointer to a global
6344string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6345It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6346</p>
6347
6348<h5>Semantics:</h5>
6349
6350<p>
6351This intrinsic allows annotations to be put on arbitrary expressions
6352with arbitrary strings. This can be useful for special purpose optimizations
6353that want to look for these annotations. These have no other defined use, they
6354are ignored by code generation and optimization.
6355</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6356
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6357<!-- _______________________________________________________________________ -->
6358<div class="doc_subsubsection">
6359 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6360</div>
6361
6362<div class="doc_text">
6363
6364<h5>Syntax:</h5>
6365<pre>
6366 declare void @llvm.trap()
6367</pre>
6368
6369<h5>Overview:</h5>
6370
6371<p>
6372The '<tt>llvm.trap</tt>' intrinsic
6373</p>
6374
6375<h5>Arguments:</h5>
6376
6377<p>
6378None
6379</p>
6380
6381<h5>Semantics:</h5>
6382
6383<p>
6384This intrinsics is lowered to the target dependent trap instruction. If the
6385target does not have a trap instruction, this intrinsic will be lowered to the
6386call of the abort() function.
6387</p>
6388</div>
6389
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6390<!-- *********************************************************************** -->
6391<hr>
6392<address>
6393 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6394 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
6395 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6396 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6397
6398 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6399 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6400 Last modified: $Date$
6401</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6402
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6403</body>
6404</html>