Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / llvm / 377f3ae16d62abbfb4a887fffeb9848a0f68b852 / . / docs / LangRef.html
blob: 5a77020187dcf3783c22127350d13398913977bb [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
752<a href="#paramattrs">parameter attributes</a>), an optional section, an
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]753optional alignment, an optional <a href="#gc">garbage collector name</a>,
Chris Lattneree202542008-10-04 18:10:21 +0000[diff] [blame]754an opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]755
756LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
757optional <a href="#linkage">linkage type</a>, an optional
758<a href="#visibility">visibility style</a>, an optional
759<a href="#callingconv">calling convention</a>, a return type, an optional
760<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]761name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]762<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]763
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]764<p>A function definition contains a list of basic blocks, forming the CFG
765(Control Flow Graph) for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]766the function. Each basic block may optionally start with a label (giving the
767basic block a symbol table entry), contains a list of instructions, and ends
768with a <a href="#terminators">terminator</a> instruction (such as a branch or
769function return).</p>
770
771<p>The first basic block in a function is special in two ways: it is immediately
772executed on entrance to the function, and it is not allowed to have predecessor
773basic blocks (i.e. there can not be any branches to the entry block of a
774function). Because the block can have no predecessors, it also cannot have any
775<a href="#i_phi">PHI nodes</a>.</p>
776
777<p>LLVM allows an explicit section to be specified for functions. If the target
778supports it, it will emit functions to the section specified.</p>
779
780<p>An explicit alignment may be specified for a function. If not present, or if
781the alignment is set to zero, the alignment of the function is set by the target
782to whatever it feels convenient. If an explicit alignment is specified, the
783function is forced to have at least that much alignment. All alignments must be
784a power of 2.</p>
785
786</div>
787
788
789<!-- ======================================================================= -->
790<div class="doc_subsection">
791 <a name="aliasstructure">Aliases</a>
792</div>
793<div class="doc_text">
794 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]795 function, global variable, another alias or bitcast of global value). Aliases
796 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]797 optional <a href="#visibility">visibility style</a>.</p>
798
799 <h5>Syntax:</h5>
800
801<div class="doc_code">
802<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000[diff] [blame]803@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]804</pre>
805</div>
806
807</div>
808
809
810
811<!-- ======================================================================= -->
812<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
813<div class="doc_text">
814 <p>The return type and each parameter of a function type may have a set of
815 <i>parameter attributes</i> associated with them. Parameter attributes are
816 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]817 a function. Parameter attributes are considered to be part of the function,
818 not of the function type, so functions with different parameter attributes
819 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]820
821 <p>Parameter attributes are simple keywords that follow the type specified. If
822 multiple parameter attributes are needed, they are space separated. For
823 example:</p>
824
825<div class="doc_code">
826<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]827declare i32 @printf(i8* noalias , ...)
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]828declare i32 @atoi(i8 zeroext)
829declare signext i8 @returns_signed_char()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]830</pre>
831</div>
832
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]833 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
834 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]835
836 <p>Currently, only the following parameter attributes are defined:</p>
837 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]838 <dt><tt>zeroext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]839 <dd>This indicates to the code generator that the parameter or return value
840 should be zero-extended to a 32-bit value by the caller (for a parameter)
841 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]842
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]843 <dt><tt>signext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]844 <dd>This indicates to the code generator that the parameter or return value
845 should be sign-extended to a 32-bit value by the caller (for a parameter)
846 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]847
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]848 <dt><tt>inreg</tt></dt>
Dale Johannesenc08a0e22008-09-25 20:47:45 +0000[diff] [blame]849 <dd>This indicates that this parameter or return value should be treated
850 in a special target-dependent fashion during while emitting code for a
851 function call or return (usually, by putting it in a register as opposed
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]852 to memory, though some targets use it to distinguish between two different
853 kinds of registers). Use of this attribute is target-specific.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]854
855 <dt><tt>byval</tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]856 <dd>This indicates that the pointer parameter should really be passed by
857 value to the function. The attribute implies that a hidden copy of the
858 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]859 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]860 pointer arguments. It is generally used to pass structs and arrays by
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]861 value, but is also valid on pointers to scalars.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]862
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]863 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]864 <dd>This indicates that the pointer parameter specifies the address of a
865 structure that is the return value of the function in the source program.
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]866 This pointer must be guaranteed by the caller to be valid: loads and stores
867 to the structure may be assumed by the callee to not to trap. This may only
868 be applied to the first parameter.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]869
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]870 <dt><tt>noalias</tt></dt>
Owen Andersonc4fc4cd2008-02-18 04:09:01 +0000[diff] [blame]871 <dd>This indicates that the parameter does not alias any global or any other
872 parameter. The caller is responsible for ensuring that this is the case,
873 usually by placing the value in a stack allocation.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]874
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]875 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]876 <dd>This indicates that the pointer parameter can be excised using the
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]877 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]878 </dl>
879
880</div>
881
882<!-- ======================================================================= -->
883<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]884 <a name="gc">Garbage Collector Names</a>
885</div>
886
887<div class="doc_text">
888<p>Each function may specify a garbage collector name, which is simply a
889string.</p>
890
891<div class="doc_code"><pre
892>define void @f() gc "name" { ...</pre></div>
893
894<p>The compiler declares the supported values of <i>name</i>. Specifying a
895collector which will cause the compiler to alter its output in order to support
896the named garbage collection algorithm.</p>
897</div>
898
899<!-- ======================================================================= -->
900<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]901 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]902</div>
903
904<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]905
906<p>Function attributes are set to communicate additional information about
907 a function. Function attributes are considered to be part of the function,
908 not of the function type, so functions with different parameter attributes
909 can have the same function type.</p>
910
911 <p>Function attributes are simple keywords that follow the type specified. If
912 multiple attributes are needed, they are space separated. For
913 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]914
915<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]916<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]917define void @f() noinline { ... }
918define void @f() alwaysinline { ... }
919define void @f() alwaysinline optsize { ... }
920define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]921</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]922</div>
923
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]924<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]925<dt><tt>alwaysinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]926<dd>This attribute indicates that the inliner should attempt to inline this
927function into callers whenever possible, ignoring any active inlining size
928threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]929
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]930<dt><tt>noinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]931<dd>This attribute indicates that the inliner should never inline this function
Chris Lattner377f3ae2008-10-05 17:14:59 +0000[diff] [blame^]932in any situation. This attribute may not be used together with the
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]933<tt>alwaysinline</tt> attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]934
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]935<dt><tt>optsize</tt></dt>
Devang Patelc29099b2008-09-29 18:34:44 +0000[diff] [blame]936<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]937make choices that keep the code size of this function low, and otherwise do
938optimizations specifically to reduce code size.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]939
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]940<dt><tt>noreturn</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]941<dd>This function attribute indicates that the function never returns normally.
942This produces undefined behavior at runtime if the function ever does
943dynamically return.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]944
945<dt><tt>nounwind</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]946<dd>This function attribute indicates that the function never returns with an
947unwind or exceptional control flow. If the function does unwind, its runtime
948behavior is undefined.</dd>
949
950<dt><tt>readnone</tt></dt>
951<dd>This attribute indicates that the function computes its result (or its
952thrown exception) based strictly on its arguments. It does not read any global
953mutable state (e.g. memory, control registers, etc) visible to caller functions.
954Furthermore, <tt>readnone</tt> functions never change any state visible to their
955caller.
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]956
957<dt><tt>readonly</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]958<dd>This function attribute indicates that the function has no side-effects on
959the calling function, but that it depends on state (memory state, control
960register state, etc) that may be set in the caller. A readonly function always
961returns the same value (or throws the same exception) whenever it is called with
962a particular set of arguments and global state.</dd>
963
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]964</dl>
965
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]966</div>
967
968<!-- ======================================================================= -->
969<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]970 <a name="moduleasm">Module-Level Inline Assembly</a>
971</div>
972
973<div class="doc_text">
974<p>
975Modules may contain "module-level inline asm" blocks, which corresponds to the
976GCC "file scope inline asm" blocks. These blocks are internally concatenated by
977LLVM and treated as a single unit, but may be separated in the .ll file if
978desired. The syntax is very simple:
979</p>
980
981<div class="doc_code">
982<pre>
983module asm "inline asm code goes here"
984module asm "more can go here"
985</pre>
986</div>
987
988<p>The strings can contain any character by escaping non-printable characters.
989 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
990 for the number.
991</p>
992
993<p>
994 The inline asm code is simply printed to the machine code .s file when
995 assembly code is generated.
996</p>
997</div>
998
999<!-- ======================================================================= -->
1000<div class="doc_subsection">
1001 <a name="datalayout">Data Layout</a>
1002</div>
1003
1004<div class="doc_text">
1005<p>A module may specify a target specific data layout string that specifies how
1006data is to be laid out in memory. The syntax for the data layout is simply:</p>
1007<pre> target datalayout = "<i>layout specification</i>"</pre>
1008<p>The <i>layout specification</i> consists of a list of specifications
1009separated by the minus sign character ('-'). Each specification starts with a
1010letter and may include other information after the letter to define some
1011aspect of the data layout. The specifications accepted are as follows: </p>
1012<dl>
1013 <dt><tt>E</tt></dt>
1014 <dd>Specifies that the target lays out data in big-endian form. That is, the
1015 bits with the most significance have the lowest address location.</dd>
1016 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1017 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1018 the bits with the least significance have the lowest address location.</dd>
1019 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1020 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1021 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1022 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1023 too.</dd>
1024 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1025 <dd>This specifies the alignment for an integer type of a given bit
1026 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1027 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1028 <dd>This specifies the alignment for a vector type of a given bit
1029 <i>size</i>.</dd>
1030 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1031 <dd>This specifies the alignment for a floating point type of a given bit
1032 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1033 (double).</dd>
1034 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1035 <dd>This specifies the alignment for an aggregate type of a given bit
1036 <i>size</i>.</dd>
1037</dl>
1038<p>When constructing the data layout for a given target, LLVM starts with a
1039default set of specifications which are then (possibly) overriden by the
1040specifications in the <tt>datalayout</tt> keyword. The default specifications
1041are given in this list:</p>
1042<ul>
1043 <li><tt>E</tt> - big endian</li>
1044 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1045 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1046 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1047 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1048 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1049 <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]1050 alignment of 64-bits</li>
1051 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1052 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1053 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1054 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1055 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1056</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1057<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1058following rules:
1059<ol>
1060 <li>If the type sought is an exact match for one of the specifications, that
1061 specification is used.</li>
1062 <li>If no match is found, and the type sought is an integer type, then the
1063 smallest integer type that is larger than the bitwidth of the sought type is
1064 used. If none of the specifications are larger than the bitwidth then the the
1065 largest integer type is used. For example, given the default specifications
1066 above, the i7 type will use the alignment of i8 (next largest) while both
1067 i65 and i256 will use the alignment of i64 (largest specified).</li>
1068 <li>If no match is found, and the type sought is a vector type, then the
1069 largest vector type that is smaller than the sought vector type will be used
1070 as a fall back. This happens because <128 x double> can be implemented in
1071 terms of 64 <2 x double>, for example.</li>
1072</ol>
1073</div>
1074
1075<!-- *********************************************************************** -->
1076<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1077<!-- *********************************************************************** -->
1078
1079<div class="doc_text">
1080
1081<p>The LLVM type system is one of the most important features of the
1082intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1083optimizations to be performed on the intermediate representation directly,
1084without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1085extra analyses on the side before the transformation. A strong type
1086system makes it easier to read the generated code and enables novel
1087analyses and transformations that are not feasible to perform on normal
1088three address code representations.</p>
1089
1090</div>
1091
1092<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1093<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1094Classifications</a> </div>
1095<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1096<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1097classifications:</p>
1098
1099<table border="1" cellspacing="0" cellpadding="4">
1100 <tbody>
1101 <tr><th>Classification</th><th>Types</th></tr>
1102 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1103 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1104 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1105 </tr>
1106 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1107 <td><a href="#t_floating">floating point</a></td>
1108 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1109 </tr>
1110 <tr>
1111 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1112 <td><a href="#t_integer">integer</a>,
1113 <a href="#t_floating">floating point</a>,
1114 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1115 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1116 <a href="#t_struct">structure</a>,
1117 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1118 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1119 </td>
1120 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1121 <tr>
1122 <td><a href="#t_primitive">primitive</a></td>
1123 <td><a href="#t_label">label</a>,
1124 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1125 <a href="#t_floating">floating point</a>.</td>
1126 </tr>
1127 <tr>
1128 <td><a href="#t_derived">derived</a></td>
1129 <td><a href="#t_integer">integer</a>,
1130 <a href="#t_array">array</a>,
1131 <a href="#t_function">function</a>,
1132 <a href="#t_pointer">pointer</a>,
1133 <a href="#t_struct">structure</a>,
1134 <a href="#t_pstruct">packed structure</a>,
1135 <a href="#t_vector">vector</a>,
1136 <a href="#t_opaque">opaque</a>.
1137 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1138 </tbody>
1139</table>
1140
1141<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1142most important. Values of these types are the only ones which can be
1143produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1144instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1145</div>
1146
1147<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1148<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1149
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1150<div class="doc_text">
1151<p>The primitive types are the fundamental building blocks of the LLVM
1152system.</p>
1153
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1154</div>
1155
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1156<!-- _______________________________________________________________________ -->
1157<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1158
1159<div class="doc_text">
1160 <table>
1161 <tbody>
1162 <tr><th>Type</th><th>Description</th></tr>
1163 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1164 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1165 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1166 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1167 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1168 </tbody>
1169 </table>
1170</div>
1171
1172<!-- _______________________________________________________________________ -->
1173<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1174
1175<div class="doc_text">
1176<h5>Overview:</h5>
1177<p>The void type does not represent any value and has no size.</p>
1178
1179<h5>Syntax:</h5>
1180
1181<pre>
1182 void
1183</pre>
1184</div>
1185
1186<!-- _______________________________________________________________________ -->
1187<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1188
1189<div class="doc_text">
1190<h5>Overview:</h5>
1191<p>The label type represents code labels.</p>
1192
1193<h5>Syntax:</h5>
1194
1195<pre>
1196 label
1197</pre>
1198</div>
1199
1200
1201<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1202<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1203
1204<div class="doc_text">
1205
1206<p>The real power in LLVM comes from the derived types in the system.
1207This is what allows a programmer to represent arrays, functions,
1208pointers, and other useful types. Note that these derived types may be
1209recursive: For example, it is possible to have a two dimensional array.</p>
1210
1211</div>
1212
1213<!-- _______________________________________________________________________ -->
1214<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1215
1216<div class="doc_text">
1217
1218<h5>Overview:</h5>
1219<p>The integer type is a very simple derived type that simply specifies an
1220arbitrary bit width for the integer type desired. Any bit width from 1 bit to
12212^23-1 (about 8 million) can be specified.</p>
1222
1223<h5>Syntax:</h5>
1224
1225<pre>
1226 iN
1227</pre>
1228
1229<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1230value.</p>
1231
1232<h5>Examples:</h5>
1233<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1234 <tbody>
1235 <tr>
1236 <td><tt>i1</tt></td>
1237 <td>a single-bit integer.</td>
1238 </tr><tr>
1239 <td><tt>i32</tt></td>
1240 <td>a 32-bit integer.</td>
1241 </tr><tr>
1242 <td><tt>i1942652</tt></td>
1243 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1244 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1245 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1246</table>
1247</div>
1248
1249<!-- _______________________________________________________________________ -->
1250<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1251
1252<div class="doc_text">
1253
1254<h5>Overview:</h5>
1255
1256<p>The array type is a very simple derived type that arranges elements
1257sequentially in memory. The array type requires a size (number of
1258elements) and an underlying data type.</p>
1259
1260<h5>Syntax:</h5>
1261
1262<pre>
1263 [&lt;# elements&gt; x &lt;elementtype&gt;]
1264</pre>
1265
1266<p>The number of elements is a constant integer value; elementtype may
1267be any type with a size.</p>
1268
1269<h5>Examples:</h5>
1270<table class="layout">
1271 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1272 <td class="left"><tt>[40 x i32]</tt></td>
1273 <td class="left">Array of 40 32-bit integer values.</td>
1274 </tr>
1275 <tr class="layout">
1276 <td class="left"><tt>[41 x i32]</tt></td>
1277 <td class="left">Array of 41 32-bit integer values.</td>
1278 </tr>
1279 <tr class="layout">
1280 <td class="left"><tt>[4 x i8]</tt></td>
1281 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1282 </tr>
1283</table>
1284<p>Here are some examples of multidimensional arrays:</p>
1285<table class="layout">
1286 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1287 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1288 <td class="left">3x4 array of 32-bit integer values.</td>
1289 </tr>
1290 <tr class="layout">
1291 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1292 <td class="left">12x10 array of single precision floating point values.</td>
1293 </tr>
1294 <tr class="layout">
1295 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1296 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1297 </tr>
1298</table>
1299
1300<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1301length array. Normally, accesses past the end of an array are undefined in
1302LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1303As a special case, however, zero length arrays are recognized to be variable
1304length. This allows implementation of 'pascal style arrays' with the LLVM
1305type "{ i32, [0 x float]}", for example.</p>
1306
1307</div>
1308
1309<!-- _______________________________________________________________________ -->
1310<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1311<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1312
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1313<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1314
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1315<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1316consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1317return 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]1318If 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]1319class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1320
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1321<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1322
1323<pre>
1324 &lt;returntype list&gt; (&lt;parameter list&gt;)
1325</pre>
1326
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1327<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1328specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1329which indicates that the function takes a variable number of arguments.
1330Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1331 href="#int_varargs">variable argument handling intrinsic</a> functions.
1332'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1333<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1334
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1335<h5>Examples:</h5>
1336<table class="layout">
1337 <tr class="layout">
1338 <td class="left"><tt>i32 (i32)</tt></td>
1339 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1340 </td>
1341 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1342 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1343 </tt></td>
1344 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1345 an <tt>i16</tt> that should be sign extended and a
1346 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1347 <tt>float</tt>.
1348 </td>
1349 </tr><tr class="layout">
1350 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1351 <td class="left">A vararg function that takes at least one
1352 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1353 which returns an integer. This is the signature for <tt>printf</tt> in
1354 LLVM.
1355 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1356 </tr><tr class="layout">
1357 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Devang Patel6dddba22008-03-24 18:10:52 +0000[diff] [blame]1358 <td class="left">A function taking an <tt>i32></tt>, returning two
1359 <tt> i32 </tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1360 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1361 </tr>
1362</table>
1363
1364</div>
1365<!-- _______________________________________________________________________ -->
1366<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1367<div class="doc_text">
1368<h5>Overview:</h5>
1369<p>The structure type is used to represent a collection of data members
1370together in memory. The packing of the field types is defined to match
1371the ABI of the underlying processor. The elements of a structure may
1372be any type that has a size.</p>
1373<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1374and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1375field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1376instruction.</p>
1377<h5>Syntax:</h5>
1378<pre> { &lt;type list&gt; }<br></pre>
1379<h5>Examples:</h5>
1380<table class="layout">
1381 <tr class="layout">
1382 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1383 <td class="left">A triple of three <tt>i32</tt> values</td>
1384 </tr><tr class="layout">
1385 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1386 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1387 second element is a <a href="#t_pointer">pointer</a> to a
1388 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1389 an <tt>i32</tt>.</td>
1390 </tr>
1391</table>
1392</div>
1393
1394<!-- _______________________________________________________________________ -->
1395<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1396</div>
1397<div class="doc_text">
1398<h5>Overview:</h5>
1399<p>The packed structure type is used to represent a collection of data members
1400together in memory. There is no padding between fields. Further, the alignment
1401of a packed structure is 1 byte. The elements of a packed structure may
1402be any type that has a size.</p>
1403<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1404and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1405field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1406instruction.</p>
1407<h5>Syntax:</h5>
1408<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1409<h5>Examples:</h5>
1410<table class="layout">
1411 <tr class="layout">
1412 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1413 <td class="left">A triple of three <tt>i32</tt> values</td>
1414 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1415 <td class="left">
1416<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1417 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1418 second element is a <a href="#t_pointer">pointer</a> to a
1419 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1420 an <tt>i32</tt>.</td>
1421 </tr>
1422</table>
1423</div>
1424
1425<!-- _______________________________________________________________________ -->
1426<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1427<div class="doc_text">
1428<h5>Overview:</h5>
1429<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1430reference to another object, which must live in memory. Pointer types may have
1431an optional address space attribute defining the target-specific numbered
1432address space where the pointed-to object resides. The default address space is
1433zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1434<h5>Syntax:</h5>
1435<pre> &lt;type&gt; *<br></pre>
1436<h5>Examples:</h5>
1437<table class="layout">
1438 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1439 <td class="left"><tt>[4x i32]*</tt></td>
1440 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1441 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1442 </tr>
1443 <tr class="layout">
1444 <td class="left"><tt>i32 (i32 *) *</tt></td>
1445 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1446 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1447 <tt>i32</tt>.</td>
1448 </tr>
1449 <tr class="layout">
1450 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1451 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1452 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1453 </tr>
1454</table>
1455</div>
1456
1457<!-- _______________________________________________________________________ -->
1458<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1459<div class="doc_text">
1460
1461<h5>Overview:</h5>
1462
1463<p>A vector type is a simple derived type that represents a vector
1464of elements. Vector types are used when multiple primitive data
1465are operated in parallel using a single instruction (SIMD).
1466A vector type requires a size (number of
1467elements) and an underlying primitive data type. Vectors must have a power
1468of two length (1, 2, 4, 8, 16 ...). Vector types are
1469considered <a href="#t_firstclass">first class</a>.</p>
1470
1471<h5>Syntax:</h5>
1472
1473<pre>
1474 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1475</pre>
1476
1477<p>The number of elements is a constant integer value; elementtype may
1478be any integer or floating point type.</p>
1479
1480<h5>Examples:</h5>
1481
1482<table class="layout">
1483 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1484 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1485 <td class="left">Vector of 4 32-bit integer values.</td>
1486 </tr>
1487 <tr class="layout">
1488 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1489 <td class="left">Vector of 8 32-bit floating-point values.</td>
1490 </tr>
1491 <tr class="layout">
1492 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1493 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1494 </tr>
1495</table>
1496</div>
1497
1498<!-- _______________________________________________________________________ -->
1499<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1500<div class="doc_text">
1501
1502<h5>Overview:</h5>
1503
1504<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1505corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1506In LLVM, opaque types can eventually be resolved to any type (not just a
1507structure type).</p>
1508
1509<h5>Syntax:</h5>
1510
1511<pre>
1512 opaque
1513</pre>
1514
1515<h5>Examples:</h5>
1516
1517<table class="layout">
1518 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1519 <td class="left"><tt>opaque</tt></td>
1520 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1521 </tr>
1522</table>
1523</div>
1524
1525
1526<!-- *********************************************************************** -->
1527<div class="doc_section"> <a name="constants">Constants</a> </div>
1528<!-- *********************************************************************** -->
1529
1530<div class="doc_text">
1531
1532<p>LLVM has several different basic types of constants. This section describes
1533them all and their syntax.</p>
1534
1535</div>
1536
1537<!-- ======================================================================= -->
1538<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1539
1540<div class="doc_text">
1541
1542<dl>
1543 <dt><b>Boolean constants</b></dt>
1544
1545 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1546 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1547 </dd>
1548
1549 <dt><b>Integer constants</b></dt>
1550
1551 <dd>Standard integers (such as '4') are constants of the <a
1552 href="#t_integer">integer</a> type. Negative numbers may be used with
1553 integer types.
1554 </dd>
1555
1556 <dt><b>Floating point constants</b></dt>
1557
1558 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1559 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1560 notation (see below). The assembler requires the exact decimal value of
1561 a floating-point constant. For example, the assembler accepts 1.25 but
1562 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1563 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1564
1565 <dt><b>Null pointer constants</b></dt>
1566
1567 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1568 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1569
1570</dl>
1571
1572<p>The one non-intuitive notation for constants is the optional hexadecimal form
1573of floating point constants. For example, the form '<tt>double
15740x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
15754.5e+15</tt>'. The only time hexadecimal floating point constants are required
1576(and the only time that they are generated by the disassembler) is when a
1577floating point constant must be emitted but it cannot be represented as a
1578decimal floating point number. For example, NaN's, infinities, and other
1579special values are represented in their IEEE hexadecimal format so that
1580assembly and disassembly do not cause any bits to change in the constants.</p>
1581
1582</div>
1583
1584<!-- ======================================================================= -->
1585<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1586</div>
1587
1588<div class="doc_text">
1589<p>Aggregate constants arise from aggregation of simple constants
1590and smaller aggregate constants.</p>
1591
1592<dl>
1593 <dt><b>Structure constants</b></dt>
1594
1595 <dd>Structure constants are represented with notation similar to structure
1596 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1597 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1598 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]1599 must have <a href="#t_struct">structure type</a>, and the number and
1600 types of elements must match those specified by the type.
1601 </dd>
1602
1603 <dt><b>Array constants</b></dt>
1604
1605 <dd>Array constants are represented with notation similar to array type
1606 definitions (a comma separated list of elements, surrounded by square brackets
1607 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1608 constants must have <a href="#t_array">array type</a>, and the number and
1609 types of elements must match those specified by the type.
1610 </dd>
1611
1612 <dt><b>Vector constants</b></dt>
1613
1614 <dd>Vector constants are represented with notation similar to vector type
1615 definitions (a comma separated list of elements, surrounded by
1616 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1617 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1618 href="#t_vector">vector type</a>, and the number and types of elements must
1619 match those specified by the type.
1620 </dd>
1621
1622 <dt><b>Zero initialization</b></dt>
1623
1624 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1625 value to zero of <em>any</em> type, including scalar and aggregate types.
1626 This is often used to avoid having to print large zero initializers (e.g. for
1627 large arrays) and is always exactly equivalent to using explicit zero
1628 initializers.
1629 </dd>
1630</dl>
1631
1632</div>
1633
1634<!-- ======================================================================= -->
1635<div class="doc_subsection">
1636 <a name="globalconstants">Global Variable and Function Addresses</a>
1637</div>
1638
1639<div class="doc_text">
1640
1641<p>The addresses of <a href="#globalvars">global variables</a> and <a
1642href="#functionstructure">functions</a> are always implicitly valid (link-time)
1643constants. These constants are explicitly referenced when the <a
1644href="#identifiers">identifier for the global</a> is used and always have <a
1645href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1646file:</p>
1647
1648<div class="doc_code">
1649<pre>
1650@X = global i32 17
1651@Y = global i32 42
1652@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1653</pre>
1654</div>
1655
1656</div>
1657
1658<!-- ======================================================================= -->
1659<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1660<div class="doc_text">
1661 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1662 no specific value. Undefined values may be of any type and be used anywhere
1663 a constant is permitted.</p>
1664
1665 <p>Undefined values indicate to the compiler that the program is well defined
1666 no matter what value is used, giving the compiler more freedom to optimize.
1667 </p>
1668</div>
1669
1670<!-- ======================================================================= -->
1671<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1672</div>
1673
1674<div class="doc_text">
1675
1676<p>Constant expressions are used to allow expressions involving other constants
1677to be used as constants. Constant expressions may be of any <a
1678href="#t_firstclass">first class</a> type and may involve any LLVM operation
1679that does not have side effects (e.g. load and call are not supported). The
1680following is the syntax for constant expressions:</p>
1681
1682<dl>
1683 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1684 <dd>Truncate a constant to another type. The bit size of CST must be larger
1685 than the bit size of TYPE. Both types must be integers.</dd>
1686
1687 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1688 <dd>Zero extend a constant to another type. The bit size of CST must be
1689 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1690
1691 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1692 <dd>Sign extend a constant to another type. The bit size of CST must be
1693 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1694
1695 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1696 <dd>Truncate a floating point constant to another floating point type. The
1697 size of CST must be larger than the size of TYPE. Both types must be
1698 floating point.</dd>
1699
1700 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1701 <dd>Floating point extend a constant to another type. The size of CST must be
1702 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1703
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1704 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1705 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1706 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1707 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1708 of the same number of elements. If the value won't fit in the integer type,
1709 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1710
1711 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1712 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1713 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1714 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1715 of the same number of elements. If the value won't fit in the integer type,
1716 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1717
1718 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1719 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1720 constant. TYPE must be a scalar or vector floating point type. CST must be of
1721 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1722 of the same number of elements. If the value won't fit in the floating point
1723 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1724
1725 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1726 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1727 constant. TYPE must be a scalar or vector floating point type. CST must be of
1728 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1729 of the same number of elements. If the value won't fit in the floating point
1730 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1731
1732 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1733 <dd>Convert a pointer typed constant to the corresponding integer constant
1734 TYPE must be an integer type. CST must be of pointer type. The CST value is
1735 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1736
1737 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1738 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1739 pointer type. CST must be of integer type. The CST value is zero extended,
1740 truncated, or unchanged to make it fit in a pointer size. This one is
1741 <i>really</i> dangerous!</dd>
1742
1743 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1744 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1745 identical (same number of bits). The conversion is done as if the CST value
1746 was stored to memory and read back as TYPE. In other words, no bits change
1747 with this operator, just the type. This can be used for conversion of
1748 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]1749 pointers it is only valid to cast to another pointer type. It is not valid
1750 to bitcast to or from an aggregate type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1751 </dd>
1752
1753 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1754
1755 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1756 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1757 instruction, the index list may have zero or more indexes, which are required
1758 to make sense for the type of "CSTPTR".</dd>
1759
1760 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1761
1762 <dd>Perform the <a href="#i_select">select operation</a> on
1763 constants.</dd>
1764
1765 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1766 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1767
1768 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1769 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1770
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1771 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1772 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1773
1774 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1775 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1776
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1777 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1778
1779 <dd>Perform the <a href="#i_extractelement">extractelement
1780 operation</a> on constants.
1781
1782 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1783
1784 <dd>Perform the <a href="#i_insertelement">insertelement
1785 operation</a> on constants.</dd>
1786
1787
1788 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1789
1790 <dd>Perform the <a href="#i_shufflevector">shufflevector
1791 operation</a> on constants.</dd>
1792
1793 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1794
1795 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1796 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1797 binary</a> operations. The constraints on operands are the same as those for
1798 the corresponding instruction (e.g. no bitwise operations on floating point
1799 values are allowed).</dd>
1800</dl>
1801</div>
1802
1803<!-- *********************************************************************** -->
1804<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1805<!-- *********************************************************************** -->
1806
1807<!-- ======================================================================= -->
1808<div class="doc_subsection">
1809<a name="inlineasm">Inline Assembler Expressions</a>
1810</div>
1811
1812<div class="doc_text">
1813
1814<p>
1815LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1816Module-Level Inline Assembly</a>) through the use of a special value. This
1817value represents the inline assembler as a string (containing the instructions
1818to emit), a list of operand constraints (stored as a string), and a flag that
1819indicates whether or not the inline asm expression has side effects. An example
1820inline assembler expression is:
1821</p>
1822
1823<div class="doc_code">
1824<pre>
1825i32 (i32) asm "bswap $0", "=r,r"
1826</pre>
1827</div>
1828
1829<p>
1830Inline assembler expressions may <b>only</b> be used as the callee operand of
1831a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1832</p>
1833
1834<div class="doc_code">
1835<pre>
1836%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1837</pre>
1838</div>
1839
1840<p>
1841Inline asms with side effects not visible in the constraint list must be marked
1842as having side effects. This is done through the use of the
1843'<tt>sideeffect</tt>' keyword, like so:
1844</p>
1845
1846<div class="doc_code">
1847<pre>
1848call void asm sideeffect "eieio", ""()
1849</pre>
1850</div>
1851
1852<p>TODO: The format of the asm and constraints string still need to be
1853documented here. Constraints on what can be done (e.g. duplication, moving, etc
Chris Lattner1cdd64e2008-10-04 18:36:02 +0000[diff] [blame]1854need to be documented). This is probably best done by reference to another
1855document that covers inline asm from a holistic perspective.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1856</p>
1857
1858</div>
1859
1860<!-- *********************************************************************** -->
1861<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1862<!-- *********************************************************************** -->
1863
1864<div class="doc_text">
1865
1866<p>The LLVM instruction set consists of several different
1867classifications of instructions: <a href="#terminators">terminator
1868instructions</a>, <a href="#binaryops">binary instructions</a>,
1869<a href="#bitwiseops">bitwise binary instructions</a>, <a
1870 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1871instructions</a>.</p>
1872
1873</div>
1874
1875<!-- ======================================================================= -->
1876<div class="doc_subsection"> <a name="terminators">Terminator
1877Instructions</a> </div>
1878
1879<div class="doc_text">
1880
1881<p>As mentioned <a href="#functionstructure">previously</a>, every
1882basic block in a program ends with a "Terminator" instruction, which
1883indicates which block should be executed after the current block is
1884finished. These terminator instructions typically yield a '<tt>void</tt>'
1885value: they produce control flow, not values (the one exception being
1886the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1887<p>There are six different terminator instructions: the '<a
1888 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1889instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1890the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1891 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1892 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1893
1894</div>
1895
1896<!-- _______________________________________________________________________ -->
1897<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1898Instruction</a> </div>
1899<div class="doc_text">
1900<h5>Syntax:</h5>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1901<pre>
1902 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]1903 ret void <i>; Return from void function</i>
1904</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1905
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1906<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1907
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1908<p>The '<tt>ret</tt>' instruction is used to return control flow (and
1909optionally a value) from a function back to the caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1910<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1911returns a value and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1912control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1913
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1914<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1915
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1916<p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
1917the return value. The type of the return value must be a
1918'<a href="#t_firstclass">first class</a>' type.</p>
1919
1920<p>A function is not <a href="#wellformed">well formed</a> if
1921it it has a non-void return type and contains a '<tt>ret</tt>'
1922instruction with no return value or a return value with a type that
1923does not match its type, or if it has a void return type and contains
1924a '<tt>ret</tt>' instruction with a return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1925
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1926<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1927
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1928<p>When the '<tt>ret</tt>' instruction is executed, control flow
1929returns back to the calling function's context. If the caller is a "<a
1930 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1931the instruction after the call. If the caller was an "<a
1932 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1933at the beginning of the "normal" destination block. If the instruction
1934returns a value, that value shall set the call or invoke instruction's
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1935return value.
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1936
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1937<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1938
1939<pre>
1940 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1941 ret void <i>; Return from a void function</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1942 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]1943</pre>
1944</div>
1945<!-- _______________________________________________________________________ -->
1946<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1947<div class="doc_text">
1948<h5>Syntax:</h5>
1949<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1950</pre>
1951<h5>Overview:</h5>
1952<p>The '<tt>br</tt>' instruction is used to cause control flow to
1953transfer to a different basic block in the current function. There are
1954two forms of this instruction, corresponding to a conditional branch
1955and an unconditional branch.</p>
1956<h5>Arguments:</h5>
1957<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1958single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1959unconditional form of the '<tt>br</tt>' instruction takes a single
1960'<tt>label</tt>' value as a target.</p>
1961<h5>Semantics:</h5>
1962<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1963argument is evaluated. If the value is <tt>true</tt>, control flows
1964to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1965control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1966<h5>Example:</h5>
1967<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
1968 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1969</div>
1970<!-- _______________________________________________________________________ -->
1971<div class="doc_subsubsection">
1972 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1973</div>
1974
1975<div class="doc_text">
1976<h5>Syntax:</h5>
1977
1978<pre>
1979 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1980</pre>
1981
1982<h5>Overview:</h5>
1983
1984<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1985several different places. It is a generalization of the '<tt>br</tt>'
1986instruction, allowing a branch to occur to one of many possible
1987destinations.</p>
1988
1989
1990<h5>Arguments:</h5>
1991
1992<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1993comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1994an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1995table is not allowed to contain duplicate constant entries.</p>
1996
1997<h5>Semantics:</h5>
1998
1999<p>The <tt>switch</tt> instruction specifies a table of values and
2000destinations. When the '<tt>switch</tt>' instruction is executed, this
2001table is searched for the given value. If the value is found, control flow is
2002transfered to the corresponding destination; otherwise, control flow is
2003transfered to the default destination.</p>
2004
2005<h5>Implementation:</h5>
2006
2007<p>Depending on properties of the target machine and the particular
2008<tt>switch</tt> instruction, this instruction may be code generated in different
2009ways. For example, it could be generated as a series of chained conditional
2010branches or with a lookup table.</p>
2011
2012<h5>Example:</h5>
2013
2014<pre>
2015 <i>; Emulate a conditional br instruction</i>
2016 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2017 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
2018
2019 <i>; Emulate an unconditional br instruction</i>
2020 switch i32 0, label %dest [ ]
2021
2022 <i>; Implement a jump table:</i>
2023 switch i32 %val, label %otherwise [ i32 0, label %onzero
2024 i32 1, label %onone
2025 i32 2, label %ontwo ]
2026</pre>
2027</div>
2028
2029<!-- _______________________________________________________________________ -->
2030<div class="doc_subsubsection">
2031 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2032</div>
2033
2034<div class="doc_text">
2035
2036<h5>Syntax:</h5>
2037
2038<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2039 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; &lt;function ptr val&gt;(&lt;function args&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2040 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2041</pre>
2042
2043<h5>Overview:</h5>
2044
2045<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2046function, with the possibility of control flow transfer to either the
2047'<tt>normal</tt>' label or the
2048'<tt>exception</tt>' label. If the callee function returns with the
2049"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2050"normal" label. If the callee (or any indirect callees) returns with the "<a
2051href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2052continued at the dynamically nearest "exception" label.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2053
2054<h5>Arguments:</h5>
2055
2056<p>This instruction requires several arguments:</p>
2057
2058<ol>
2059 <li>
2060 The optional "cconv" marker indicates which <a href="#callingconv">calling
2061 convention</a> the call should use. If none is specified, the call defaults
2062 to using C calling conventions.
2063 </li>
2064 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2065 function value being invoked. In most cases, this is a direct function
2066 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2067 an arbitrary pointer to function value.
2068 </li>
2069
2070 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2071 function to be invoked. </li>
2072
2073 <li>'<tt>function args</tt>': argument list whose types match the function
2074 signature argument types. If the function signature indicates the function
2075 accepts a variable number of arguments, the extra arguments can be
2076 specified. </li>
2077
2078 <li>'<tt>normal label</tt>': the label reached when the called function
2079 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2080
2081 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2082 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2083
2084</ol>
2085
2086<h5>Semantics:</h5>
2087
2088<p>This instruction is designed to operate as a standard '<tt><a
2089href="#i_call">call</a></tt>' instruction in most regards. The primary
2090difference is that it establishes an association with a label, which is used by
2091the runtime library to unwind the stack.</p>
2092
2093<p>This instruction is used in languages with destructors to ensure that proper
2094cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2095exception. Additionally, this is important for implementation of
2096'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2097
2098<h5>Example:</h5>
2099<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2100 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2101 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2102 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2103 unwind label %TestCleanup <i>; {i32}:retval set</i>
2104</pre>
2105</div>
2106
2107
2108<!-- _______________________________________________________________________ -->
2109
2110<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2111Instruction</a> </div>
2112
2113<div class="doc_text">
2114
2115<h5>Syntax:</h5>
2116<pre>
2117 unwind
2118</pre>
2119
2120<h5>Overview:</h5>
2121
2122<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2123at the first callee in the dynamic call stack which used an <a
2124href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2125primarily used to implement exception handling.</p>
2126
2127<h5>Semantics:</h5>
2128
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2129<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2130immediately halt. The dynamic call stack is then searched for the first <a
2131href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2132execution continues at the "exceptional" destination block specified by the
2133<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2134dynamic call chain, undefined behavior results.</p>
2135</div>
2136
2137<!-- _______________________________________________________________________ -->
2138
2139<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2140Instruction</a> </div>
2141
2142<div class="doc_text">
2143
2144<h5>Syntax:</h5>
2145<pre>
2146 unreachable
2147</pre>
2148
2149<h5>Overview:</h5>
2150
2151<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2152instruction is used to inform the optimizer that a particular portion of the
2153code is not reachable. This can be used to indicate that the code after a
2154no-return function cannot be reached, and other facts.</p>
2155
2156<h5>Semantics:</h5>
2157
2158<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2159</div>
2160
2161
2162
2163<!-- ======================================================================= -->
2164<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2165<div class="doc_text">
2166<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2167program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2168produce a single value. The operands might represent
2169multiple 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]2170The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2171<p>There are several different binary operators:</p>
2172</div>
2173<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2174<div class="doc_subsubsection">
2175 <a name="i_add">'<tt>add</tt>' Instruction</a>
2176</div>
2177
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2178<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2179
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2180<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2181
2182<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2183 &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]2184</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2185
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2186<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2187
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2188<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2189
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2190<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2191
2192<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2193 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2194 <a href="#t_vector">vector</a> values. Both arguments must have identical
2195 types.</p>
2196
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2197<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2198
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2199<p>The value produced is the integer or floating point sum of the two
2200operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2201
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2202<p>If an integer sum has unsigned overflow, the result returned is the
2203mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2204the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2205
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2206<p>Because LLVM integers use a two's complement representation, this
2207instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2208
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2209<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2210
2211<pre>
2212 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2213</pre>
2214</div>
2215<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2216<div class="doc_subsubsection">
2217 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2218</div>
2219
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2220<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2221
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2222<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2223
2224<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2225 &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]2226</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2227
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2228<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2229
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2230<p>The '<tt>sub</tt>' instruction returns the difference of its two
2231operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2232
2233<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2234'<tt>neg</tt>' instruction present in most other intermediate
2235representations.</p>
2236
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2237<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2238
2239<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2240 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2241 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2242 types.</p>
2243
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2244<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2245
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2246<p>The value produced is the integer or floating point difference of
2247the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2248
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2249<p>If an integer difference has unsigned overflow, the result returned is the
2250mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2251the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2252
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2253<p>Because LLVM integers use a two's complement representation, this
2254instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2255
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2256<h5>Example:</h5>
2257<pre>
2258 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2259 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2260</pre>
2261</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2262
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2263<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2264<div class="doc_subsubsection">
2265 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2266</div>
2267
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2268<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2269
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2270<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2271<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]2272</pre>
2273<h5>Overview:</h5>
2274<p>The '<tt>mul</tt>' instruction returns the product of its two
2275operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2276
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2277<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2278
2279<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2280href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2281or <a href="#t_vector">vector</a> values. Both arguments must have identical
2282types.</p>
2283
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2284<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2285
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2286<p>The value produced is the integer or floating point product of the
2287two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2288
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2289<p>If the result of an integer multiplication has unsigned overflow,
2290the result returned is the mathematical result modulo
22912<sup>n</sup>, where n is the bit width of the result.</p>
2292<p>Because LLVM integers use a two's complement representation, and the
2293result is the same width as the operands, this instruction returns the
2294correct result for both signed and unsigned integers. If a full product
2295(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2296should be sign-extended or zero-extended as appropriate to the
2297width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2298<h5>Example:</h5>
2299<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2300</pre>
2301</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2302
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2303<!-- _______________________________________________________________________ -->
2304<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2305</a></div>
2306<div class="doc_text">
2307<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2308<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]2309</pre>
2310<h5>Overview:</h5>
2311<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2312operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2313
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2314<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2315
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2316<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2317<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2318values. Both arguments must have identical types.</p>
2319
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2320<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2321
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2322<p>The value produced is the unsigned integer quotient of the two operands.</p>
2323<p>Note that unsigned integer division and signed integer division are distinct
2324operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2325<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2326<h5>Example:</h5>
2327<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2328</pre>
2329</div>
2330<!-- _______________________________________________________________________ -->
2331<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2332</a> </div>
2333<div class="doc_text">
2334<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2335<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2336 &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]2337</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2338
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2339<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2340
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2341<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2342operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2343
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2344<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2345
2346<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2347<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2348values. Both arguments must have identical types.</p>
2349
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2350<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2351<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]2352<p>Note that signed integer division and unsigned integer division are distinct
2353operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2354<p>Division by zero leads to undefined behavior. Overflow also leads to
2355undefined behavior; this is a rare case, but can occur, for example,
2356by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2357<h5>Example:</h5>
2358<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2359</pre>
2360</div>
2361<!-- _______________________________________________________________________ -->
2362<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2363Instruction</a> </div>
2364<div class="doc_text">
2365<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2366<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2367 &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]2368</pre>
2369<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2370
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2371<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2372operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2373
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2374<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2375
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2376<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2377<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2378of floating point values. Both arguments must have identical types.</p>
2379
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2380<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2381
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2382<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2383
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2384<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2385
2386<pre>
2387 &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]2388</pre>
2389</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2390
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2391<!-- _______________________________________________________________________ -->
2392<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2393</div>
2394<div class="doc_text">
2395<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2396<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]2397</pre>
2398<h5>Overview:</h5>
2399<p>The '<tt>urem</tt>' instruction returns the remainder from the
2400unsigned division of its two arguments.</p>
2401<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2402<p>The two arguments to the '<tt>urem</tt>' instruction must be
2403<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2404values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2405<h5>Semantics:</h5>
2406<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2407This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2408<p>Note that unsigned integer remainder and signed integer remainder are
2409distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2410<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2411<h5>Example:</h5>
2412<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2413</pre>
2414
2415</div>
2416<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2417<div class="doc_subsubsection">
2418 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2419</div>
2420
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2421<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2422
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2423<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2424
2425<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2426 &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]2427</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2428
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2429<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2430
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2431<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2432signed division of its two operands. This instruction can also take
2433<a href="#t_vector">vector</a> versions of the values in which case
2434the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2435
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2436<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2437
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2438<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2439<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2440values. Both arguments must have identical types.</p>
2441
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2442<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2443
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2444<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2445has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2446operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2447a value. For more information about the difference, see <a
2448 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2449Math Forum</a>. For a table of how this is implemented in various languages,
2450please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2451Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2452<p>Note that signed integer remainder and unsigned integer remainder are
2453distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2454<p>Taking the remainder of a division by zero leads to undefined behavior.
2455Overflow also leads to undefined behavior; this is a rare case, but can occur,
2456for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2457(The remainder doesn't actually overflow, but this rule lets srem be
2458implemented using instructions that return both the result of the division
2459and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2460<h5>Example:</h5>
2461<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2462</pre>
2463
2464</div>
2465<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2466<div class="doc_subsubsection">
2467 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2468
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2469<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2470
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2471<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2472<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]2473</pre>
2474<h5>Overview:</h5>
2475<p>The '<tt>frem</tt>' instruction returns the remainder from the
2476division of its two operands.</p>
2477<h5>Arguments:</h5>
2478<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2479<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2480of floating point values. Both arguments must have identical types.</p>
2481
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2482<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2483
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2484<p>This instruction returns the <i>remainder</i> of a division.
2485The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2486
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2487<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2488
2489<pre>
2490 &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]2491</pre>
2492</div>
2493
2494<!-- ======================================================================= -->
2495<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2496Operations</a> </div>
2497<div class="doc_text">
2498<p>Bitwise binary operators are used to do various forms of
2499bit-twiddling in a program. They are generally very efficient
2500instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2501instructions. They require two operands of the same type, execute an operation on them,
2502and 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]2503</div>
2504
2505<!-- _______________________________________________________________________ -->
2506<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2507Instruction</a> </div>
2508<div class="doc_text">
2509<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2510<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]2511</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2512
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2513<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2514
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2515<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2516the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2517
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2518<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2519
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2520<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2521 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2522type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2523
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2524<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2525
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2526<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2527where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2528equal 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]2529
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2530<h5>Example:</h5><pre>
2531 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2532 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2533 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2534 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2535</pre>
2536</div>
2537<!-- _______________________________________________________________________ -->
2538<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2539Instruction</a> </div>
2540<div class="doc_text">
2541<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2542<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]2543</pre>
2544
2545<h5>Overview:</h5>
2546<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2547operand shifted to the right a specified number of bits with zero fill.</p>
2548
2549<h5>Arguments:</h5>
2550<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2551<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]2552type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2553
2554<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2555
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2556<p>This instruction always performs a logical shift right operation. The most
2557significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2558shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2559the number of bits in <tt>op1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2560
2561<h5>Example:</h5>
2562<pre>
2563 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2564 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2565 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2566 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2567 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2568</pre>
2569</div>
2570
2571<!-- _______________________________________________________________________ -->
2572<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2573Instruction</a> </div>
2574<div class="doc_text">
2575
2576<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2577<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]2578</pre>
2579
2580<h5>Overview:</h5>
2581<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2582operand shifted to the right a specified number of bits with sign extension.</p>
2583
2584<h5>Arguments:</h5>
2585<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2586<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]2587type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2588
2589<h5>Semantics:</h5>
2590<p>This instruction always performs an arithmetic shift right operation,
2591The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2592of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2593larger than the number of bits in <tt>op1</tt>, the result is undefined.
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2594</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2595
2596<h5>Example:</h5>
2597<pre>
2598 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2599 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2600 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2601 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2602 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2603</pre>
2604</div>
2605
2606<!-- _______________________________________________________________________ -->
2607<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2608Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2609
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2610<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2611
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2612<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2613
2614<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2615 &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]2616</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2617
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2618<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2619
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2620<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2621its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2622
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2623<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2624
2625<p>The two arguments to the '<tt>and</tt>' instruction must be
2626<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2627values. Both arguments must have identical types.</p>
2628
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2629<h5>Semantics:</h5>
2630<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2631<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2632<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2633<table border="1" cellspacing="0" cellpadding="4">
2634 <tbody>
2635 <tr>
2636 <td>In0</td>
2637 <td>In1</td>
2638 <td>Out</td>
2639 </tr>
2640 <tr>
2641 <td>0</td>
2642 <td>0</td>
2643 <td>0</td>
2644 </tr>
2645 <tr>
2646 <td>0</td>
2647 <td>1</td>
2648 <td>0</td>
2649 </tr>
2650 <tr>
2651 <td>1</td>
2652 <td>0</td>
2653 <td>0</td>
2654 </tr>
2655 <tr>
2656 <td>1</td>
2657 <td>1</td>
2658 <td>1</td>
2659 </tr>
2660 </tbody>
2661</table>
2662</div>
2663<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2664<pre>
2665 &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]2666 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2667 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2668</pre>
2669</div>
2670<!-- _______________________________________________________________________ -->
2671<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2672<div class="doc_text">
2673<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2674<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]2675</pre>
2676<h5>Overview:</h5>
2677<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2678or of its two operands.</p>
2679<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2680
2681<p>The two arguments to the '<tt>or</tt>' instruction must be
2682<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2683values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2684<h5>Semantics:</h5>
2685<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2686<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2687<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2688<table border="1" cellspacing="0" cellpadding="4">
2689 <tbody>
2690 <tr>
2691 <td>In0</td>
2692 <td>In1</td>
2693 <td>Out</td>
2694 </tr>
2695 <tr>
2696 <td>0</td>
2697 <td>0</td>
2698 <td>0</td>
2699 </tr>
2700 <tr>
2701 <td>0</td>
2702 <td>1</td>
2703 <td>1</td>
2704 </tr>
2705 <tr>
2706 <td>1</td>
2707 <td>0</td>
2708 <td>1</td>
2709 </tr>
2710 <tr>
2711 <td>1</td>
2712 <td>1</td>
2713 <td>1</td>
2714 </tr>
2715 </tbody>
2716</table>
2717</div>
2718<h5>Example:</h5>
2719<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2720 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2721 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2722</pre>
2723</div>
2724<!-- _______________________________________________________________________ -->
2725<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2726Instruction</a> </div>
2727<div class="doc_text">
2728<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2729<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]2730</pre>
2731<h5>Overview:</h5>
2732<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2733or of its two operands. The <tt>xor</tt> is used to implement the
2734"one's complement" operation, which is the "~" operator in C.</p>
2735<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2736<p>The two arguments to the '<tt>xor</tt>' instruction must be
2737<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2738values. Both arguments must have identical types.</p>
2739
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2740<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2741
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2742<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2743<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2744<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2745<table border="1" cellspacing="0" cellpadding="4">
2746 <tbody>
2747 <tr>
2748 <td>In0</td>
2749 <td>In1</td>
2750 <td>Out</td>
2751 </tr>
2752 <tr>
2753 <td>0</td>
2754 <td>0</td>
2755 <td>0</td>
2756 </tr>
2757 <tr>
2758 <td>0</td>
2759 <td>1</td>
2760 <td>1</td>
2761 </tr>
2762 <tr>
2763 <td>1</td>
2764 <td>0</td>
2765 <td>1</td>
2766 </tr>
2767 <tr>
2768 <td>1</td>
2769 <td>1</td>
2770 <td>0</td>
2771 </tr>
2772 </tbody>
2773</table>
2774</div>
2775<p> </p>
2776<h5>Example:</h5>
2777<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2778 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2779 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2780 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2781</pre>
2782</div>
2783
2784<!-- ======================================================================= -->
2785<div class="doc_subsection">
2786 <a name="vectorops">Vector Operations</a>
2787</div>
2788
2789<div class="doc_text">
2790
2791<p>LLVM supports several instructions to represent vector operations in a
2792target-independent manner. These instructions cover the element-access and
2793vector-specific operations needed to process vectors effectively. While LLVM
2794does directly support these vector operations, many sophisticated algorithms
2795will want to use target-specific intrinsics to take full advantage of a specific
2796target.</p>
2797
2798</div>
2799
2800<!-- _______________________________________________________________________ -->
2801<div class="doc_subsubsection">
2802 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2803</div>
2804
2805<div class="doc_text">
2806
2807<h5>Syntax:</h5>
2808
2809<pre>
2810 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2811</pre>
2812
2813<h5>Overview:</h5>
2814
2815<p>
2816The '<tt>extractelement</tt>' instruction extracts a single scalar
2817element from a vector at a specified index.
2818</p>
2819
2820
2821<h5>Arguments:</h5>
2822
2823<p>
2824The first operand of an '<tt>extractelement</tt>' instruction is a
2825value of <a href="#t_vector">vector</a> type. The second operand is
2826an index indicating the position from which to extract the element.
2827The index may be a variable.</p>
2828
2829<h5>Semantics:</h5>
2830
2831<p>
2832The result is a scalar of the same type as the element type of
2833<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2834<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2835results are undefined.
2836</p>
2837
2838<h5>Example:</h5>
2839
2840<pre>
2841 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2842</pre>
2843</div>
2844
2845
2846<!-- _______________________________________________________________________ -->
2847<div class="doc_subsubsection">
2848 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2849</div>
2850
2851<div class="doc_text">
2852
2853<h5>Syntax:</h5>
2854
2855<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]2856 &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]2857</pre>
2858
2859<h5>Overview:</h5>
2860
2861<p>
2862The '<tt>insertelement</tt>' instruction inserts a scalar
2863element into a vector at a specified index.
2864</p>
2865
2866
2867<h5>Arguments:</h5>
2868
2869<p>
2870The first operand of an '<tt>insertelement</tt>' instruction is a
2871value of <a href="#t_vector">vector</a> type. The second operand is a
2872scalar value whose type must equal the element type of the first
2873operand. The third operand is an index indicating the position at
2874which to insert the value. The index may be a variable.</p>
2875
2876<h5>Semantics:</h5>
2877
2878<p>
2879The result is a vector of the same type as <tt>val</tt>. Its
2880element values are those of <tt>val</tt> except at position
2881<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2882exceeds the length of <tt>val</tt>, the results are undefined.
2883</p>
2884
2885<h5>Example:</h5>
2886
2887<pre>
2888 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2889</pre>
2890</div>
2891
2892<!-- _______________________________________________________________________ -->
2893<div class="doc_subsubsection">
2894 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2895</div>
2896
2897<div class="doc_text">
2898
2899<h5>Syntax:</h5>
2900
2901<pre>
2902 &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>
2903</pre>
2904
2905<h5>Overview:</h5>
2906
2907<p>
2908The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2909from two input vectors, returning a vector of the same type.
2910</p>
2911
2912<h5>Arguments:</h5>
2913
2914<p>
2915The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2916with types that match each other and types that match the result of the
2917instruction. The third argument is a shuffle mask, which has the same number
2918of elements as the other vector type, but whose element type is always 'i32'.
2919</p>
2920
2921<p>
2922The shuffle mask operand is required to be a constant vector with either
2923constant integer or undef values.
2924</p>
2925
2926<h5>Semantics:</h5>
2927
2928<p>
2929The elements of the two input vectors are numbered from left to right across
2930both of the vectors. The shuffle mask operand specifies, for each element of
2931the result vector, which element of the two input registers the result element
2932gets. The element selector may be undef (meaning "don't care") and the second
2933operand may be undef if performing a shuffle from only one vector.
2934</p>
2935
2936<h5>Example:</h5>
2937
2938<pre>
2939 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2940 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2941 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2942 &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.
2943</pre>
2944</div>
2945
2946
2947<!-- ======================================================================= -->
2948<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2949 <a name="aggregateops">Aggregate Operations</a>
2950</div>
2951
2952<div class="doc_text">
2953
2954<p>LLVM supports several instructions for working with aggregate values.
2955</p>
2956
2957</div>
2958
2959<!-- _______________________________________________________________________ -->
2960<div class="doc_subsubsection">
2961 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
2962</div>
2963
2964<div class="doc_text">
2965
2966<h5>Syntax:</h5>
2967
2968<pre>
2969 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
2970</pre>
2971
2972<h5>Overview:</h5>
2973
2974<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2975The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
2976or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2977</p>
2978
2979
2980<h5>Arguments:</h5>
2981
2982<p>
2983The first operand of an '<tt>extractvalue</tt>' instruction is a
2984value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2985type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2986in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2987'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2988</p>
2989
2990<h5>Semantics:</h5>
2991
2992<p>
2993The result is the value at the position in the aggregate specified by
2994the index operands.
2995</p>
2996
2997<h5>Example:</h5>
2998
2999<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3000 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3001</pre>
3002</div>
3003
3004
3005<!-- _______________________________________________________________________ -->
3006<div class="doc_subsubsection">
3007 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3008</div>
3009
3010<div class="doc_text">
3011
3012<h5>Syntax:</h5>
3013
3014<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3015 &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]3016</pre>
3017
3018<h5>Overview:</h5>
3019
3020<p>
3021The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3022into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3023</p>
3024
3025
3026<h5>Arguments:</h5>
3027
3028<p>
3029The first operand of an '<tt>insertvalue</tt>' instruction is a
3030value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3031The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3032The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3033indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3034indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3035'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3036The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3037by the indices.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3038
3039<h5>Semantics:</h5>
3040
3041<p>
3042The result is an aggregate of the same type as <tt>val</tt>. Its
3043value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3044specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3045</p>
3046
3047<h5>Example:</h5>
3048
3049<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3050 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3051</pre>
3052</div>
3053
3054
3055<!-- ======================================================================= -->
3056<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3057 <a name="memoryops">Memory Access and Addressing Operations</a>
3058</div>
3059
3060<div class="doc_text">
3061
3062<p>A key design point of an SSA-based representation is how it
3063represents memory. In LLVM, no memory locations are in SSA form, which
3064makes things very simple. This section describes how to read, write,
3065allocate, and free memory in LLVM.</p>
3066
3067</div>
3068
3069<!-- _______________________________________________________________________ -->
3070<div class="doc_subsubsection">
3071 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3072</div>
3073
3074<div class="doc_text">
3075
3076<h5>Syntax:</h5>
3077
3078<pre>
3079 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3080</pre>
3081
3082<h5>Overview:</h5>
3083
3084<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3085heap and returns a pointer to it. The object is always allocated in the generic
3086address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3087
3088<h5>Arguments:</h5>
3089
3090<p>The '<tt>malloc</tt>' instruction allocates
3091<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3092bytes of memory from the operating system and returns a pointer of the
3093appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3094number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3095If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3096be aligned to at least that boundary. If not specified, or if zero, the target can
3097choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3098
3099<p>'<tt>type</tt>' must be a sized type.</p>
3100
3101<h5>Semantics:</h5>
3102
3103<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3104a pointer is returned. The result of a zero byte allocattion is undefined. The
3105result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3106
3107<h5>Example:</h5>
3108
3109<pre>
3110 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
3111
3112 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3113 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3114 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3115 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3116 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3117</pre>
3118</div>
3119
3120<!-- _______________________________________________________________________ -->
3121<div class="doc_subsubsection">
3122 <a name="i_free">'<tt>free</tt>' Instruction</a>
3123</div>
3124
3125<div class="doc_text">
3126
3127<h5>Syntax:</h5>
3128
3129<pre>
3130 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
3131</pre>
3132
3133<h5>Overview:</h5>
3134
3135<p>The '<tt>free</tt>' instruction returns memory back to the unused
3136memory heap to be reallocated in the future.</p>
3137
3138<h5>Arguments:</h5>
3139
3140<p>'<tt>value</tt>' shall be a pointer value that points to a value
3141that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3142instruction.</p>
3143
3144<h5>Semantics:</h5>
3145
3146<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3147after this instruction executes. If the pointer is null, the operation
3148is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3149
3150<h5>Example:</h5>
3151
3152<pre>
3153 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3154 free [4 x i8]* %array
3155</pre>
3156</div>
3157
3158<!-- _______________________________________________________________________ -->
3159<div class="doc_subsubsection">
3160 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3161</div>
3162
3163<div class="doc_text">
3164
3165<h5>Syntax:</h5>
3166
3167<pre>
3168 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3169</pre>
3170
3171<h5>Overview:</h5>
3172
3173<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3174currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3175returns to its caller. The object is always allocated in the generic address
3176space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3177
3178<h5>Arguments:</h5>
3179
3180<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3181bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3182appropriate type to the program. If "NumElements" is specified, it is the
3183number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3184If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3185to be aligned to at least that boundary. If not specified, or if zero, the target
3186can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3187
3188<p>'<tt>type</tt>' may be any sized type.</p>
3189
3190<h5>Semantics:</h5>
3191
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3192<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3193there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3194memory is automatically released when the function returns. The '<tt>alloca</tt>'
3195instruction is commonly used to represent automatic variables that must
3196have an address available. When the function returns (either with the <tt><a
3197 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]3198instructions), the memory is reclaimed. Allocating zero bytes
3199is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3200
3201<h5>Example:</h5>
3202
3203<pre>
3204 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3205 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3206 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3207 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3208</pre>
3209</div>
3210
3211<!-- _______________________________________________________________________ -->
3212<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3213Instruction</a> </div>
3214<div class="doc_text">
3215<h5>Syntax:</h5>
3216<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>
3217<h5>Overview:</h5>
3218<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3219<h5>Arguments:</h5>
3220<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3221address from which to load. The pointer must point to a <a
3222 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3223marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3224the number or order of execution of this <tt>load</tt> with other
3225volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3226instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3227<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3228The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3229(that is, the alignment of the memory address). A value of 0 or an
3230omitted "align" argument means that the operation has the preferential
3231alignment for the target. It is the responsibility of the code emitter
3232to ensure that the alignment information is correct. Overestimating
3233the alignment results in an undefined behavior. Underestimating the
3234alignment may produce less efficient code. An alignment of 1 is always
3235safe.
3236</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3237<h5>Semantics:</h5>
3238<p>The location of memory pointed to is loaded.</p>
3239<h5>Examples:</h5>
3240<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3241 <a
3242 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3243 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3244</pre>
3245</div>
3246<!-- _______________________________________________________________________ -->
3247<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3248Instruction</a> </div>
3249<div class="doc_text">
3250<h5>Syntax:</h5>
3251<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3252 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3253</pre>
3254<h5>Overview:</h5>
3255<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3256<h5>Arguments:</h5>
3257<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3258to 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]3259operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3260of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3261operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3262optimizer is not allowed to modify the number or order of execution of
3263this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3264 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3265<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3266The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3267(that is, the alignment of the memory address). A value of 0 or an
3268omitted "align" argument means that the operation has the preferential
3269alignment for the target. It is the responsibility of the code emitter
3270to ensure that the alignment information is correct. Overestimating
3271the alignment results in an undefined behavior. Underestimating the
3272alignment may produce less efficient code. An alignment of 1 is always
3273safe.
3274</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3275<h5>Semantics:</h5>
3276<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3277at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3278<h5>Example:</h5>
3279<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3280 store i32 3, i32* %ptr <i>; yields {void}</i>
3281 %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]3282</pre>
3283</div>
3284
3285<!-- _______________________________________________________________________ -->
3286<div class="doc_subsubsection">
3287 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3288</div>
3289
3290<div class="doc_text">
3291<h5>Syntax:</h5>
3292<pre>
3293 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
3294</pre>
3295
3296<h5>Overview:</h5>
3297
3298<p>
3299The '<tt>getelementptr</tt>' instruction is used to get the address of a
3300subelement of an aggregate data structure.</p>
3301
3302<h5>Arguments:</h5>
3303
3304<p>This instruction takes a list of integer operands that indicate what
3305elements of the aggregate object to index to. The actual types of the arguments
3306provided depend on the type of the first pointer argument. The
3307'<tt>getelementptr</tt>' instruction is used to index down through the type
3308levels of a structure or to a specific index in an array. When indexing into a
3309structure, only <tt>i32</tt> integer constants are allowed. When indexing
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3310into an array or pointer, only integers of 32 or 64 bits are allowed; 32-bit
3311values will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3312
3313<p>For example, let's consider a C code fragment and how it gets
3314compiled to LLVM:</p>
3315
3316<div class="doc_code">
3317<pre>
3318struct RT {
3319 char A;
3320 int B[10][20];
3321 char C;
3322};
3323struct ST {
3324 int X;
3325 double Y;
3326 struct RT Z;
3327};
3328
3329int *foo(struct ST *s) {
3330 return &amp;s[1].Z.B[5][13];
3331}
3332</pre>
3333</div>
3334
3335<p>The LLVM code generated by the GCC frontend is:</p>
3336
3337<div class="doc_code">
3338<pre>
3339%RT = type { i8 , [10 x [20 x i32]], i8 }
3340%ST = type { i32, double, %RT }
3341
3342define i32* %foo(%ST* %s) {
3343entry:
3344 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3345 ret i32* %reg
3346}
3347</pre>
3348</div>
3349
3350<h5>Semantics:</h5>
3351
3352<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3353on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3354and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3355<a href="#t_integer">integer</a> type but the value will always be sign extended
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3356to 64-bits. <a href="#t_struct">Structure</a> and <a href="#t_pstruct">packed
3357structure</a> types require <tt>i32</tt> <b>constants</b>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3358
3359<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3360type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3361}</tt>' type, a structure. The second index indexes into the third element of
3362the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3363i8 }</tt>' type, another structure. The third index indexes into the second
3364element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3365array. The two dimensions of the array are subscripted into, yielding an
3366'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3367to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3368
3369<p>Note that it is perfectly legal to index partially through a
3370structure, returning a pointer to an inner element. Because of this,
3371the LLVM code for the given testcase is equivalent to:</p>
3372
3373<pre>
3374 define i32* %foo(%ST* %s) {
3375 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3376 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3377 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3378 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3379 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3380 ret i32* %t5
3381 }
3382</pre>
3383
3384<p>Note that it is undefined to access an array out of bounds: array and
3385pointer indexes must always be within the defined bounds of the array type.
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3386The one exception for this rule is zero length arrays. These arrays are
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3387defined to be accessible as variable length arrays, which requires access
3388beyond the zero'th element.</p>
3389
3390<p>The getelementptr instruction is often confusing. For some more insight
3391into how it works, see <a href="GetElementPtr.html">the getelementptr
3392FAQ</a>.</p>
3393
3394<h5>Example:</h5>
3395
3396<pre>
3397 <i>; yields [12 x i8]*:aptr</i>
3398 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3399</pre>
3400</div>
3401
3402<!-- ======================================================================= -->
3403<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3404</div>
3405<div class="doc_text">
3406<p>The instructions in this category are the conversion instructions (casting)
3407which all take a single operand and a type. They perform various bit conversions
3408on the operand.</p>
3409</div>
3410
3411<!-- _______________________________________________________________________ -->
3412<div class="doc_subsubsection">
3413 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3414</div>
3415<div class="doc_text">
3416
3417<h5>Syntax:</h5>
3418<pre>
3419 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3420</pre>
3421
3422<h5>Overview:</h5>
3423<p>
3424The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3425</p>
3426
3427<h5>Arguments:</h5>
3428<p>
3429The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3430be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3431and type of the result, which must be an <a href="#t_integer">integer</a>
3432type. The bit size of <tt>value</tt> must be larger than the bit size of
3433<tt>ty2</tt>. Equal sized types are not allowed.</p>
3434
3435<h5>Semantics:</h5>
3436<p>
3437The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3438and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3439larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3440It will always truncate bits.</p>
3441
3442<h5>Example:</h5>
3443<pre>
3444 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3445 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3446 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3447</pre>
3448</div>
3449
3450<!-- _______________________________________________________________________ -->
3451<div class="doc_subsubsection">
3452 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3453</div>
3454<div class="doc_text">
3455
3456<h5>Syntax:</h5>
3457<pre>
3458 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3459</pre>
3460
3461<h5>Overview:</h5>
3462<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3463<tt>ty2</tt>.</p>
3464
3465
3466<h5>Arguments:</h5>
3467<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3468<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3469also be of <a href="#t_integer">integer</a> type. The bit size of the
3470<tt>value</tt> must be smaller than the bit size of the destination type,
3471<tt>ty2</tt>.</p>
3472
3473<h5>Semantics:</h5>
3474<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3475bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3476
3477<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3478
3479<h5>Example:</h5>
3480<pre>
3481 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3482 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3483</pre>
3484</div>
3485
3486<!-- _______________________________________________________________________ -->
3487<div class="doc_subsubsection">
3488 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3489</div>
3490<div class="doc_text">
3491
3492<h5>Syntax:</h5>
3493<pre>
3494 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3495</pre>
3496
3497<h5>Overview:</h5>
3498<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3499
3500<h5>Arguments:</h5>
3501<p>
3502The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3503<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3504also be of <a href="#t_integer">integer</a> type. The bit size of the
3505<tt>value</tt> must be smaller than the bit size of the destination type,
3506<tt>ty2</tt>.</p>
3507
3508<h5>Semantics:</h5>
3509<p>
3510The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3511bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3512the type <tt>ty2</tt>.</p>
3513
3514<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3515
3516<h5>Example:</h5>
3517<pre>
3518 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3519 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3520</pre>
3521</div>
3522
3523<!-- _______________________________________________________________________ -->
3524<div class="doc_subsubsection">
3525 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3526</div>
3527
3528<div class="doc_text">
3529
3530<h5>Syntax:</h5>
3531
3532<pre>
3533 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3534</pre>
3535
3536<h5>Overview:</h5>
3537<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3538<tt>ty2</tt>.</p>
3539
3540
3541<h5>Arguments:</h5>
3542<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3543 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3544cast it to. The size of <tt>value</tt> must be larger than the size of
3545<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3546<i>no-op cast</i>.</p>
3547
3548<h5>Semantics:</h5>
3549<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3550<a href="#t_floating">floating point</a> type to a smaller
3551<a href="#t_floating">floating point</a> type. If the value cannot fit within
3552the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3553
3554<h5>Example:</h5>
3555<pre>
3556 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3557 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3558</pre>
3559</div>
3560
3561<!-- _______________________________________________________________________ -->
3562<div class="doc_subsubsection">
3563 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3564</div>
3565<div class="doc_text">
3566
3567<h5>Syntax:</h5>
3568<pre>
3569 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3570</pre>
3571
3572<h5>Overview:</h5>
3573<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3574floating point value.</p>
3575
3576<h5>Arguments:</h5>
3577<p>The '<tt>fpext</tt>' instruction takes a
3578<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3579and a <a href="#t_floating">floating point</a> type to cast it to. The source
3580type must be smaller than the destination type.</p>
3581
3582<h5>Semantics:</h5>
3583<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3584<a href="#t_floating">floating point</a> type to a larger
3585<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3586used to make a <i>no-op cast</i> because it always changes bits. Use
3587<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3588
3589<h5>Example:</h5>
3590<pre>
3591 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3592 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3593</pre>
3594</div>
3595
3596<!-- _______________________________________________________________________ -->
3597<div class="doc_subsubsection">
3598 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3599</div>
3600<div class="doc_text">
3601
3602<h5>Syntax:</h5>
3603<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3604 &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]3605</pre>
3606
3607<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3608<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3609unsigned integer equivalent of type <tt>ty2</tt>.
3610</p>
3611
3612<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3613<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]3614scalar or vector <a href="#t_floating">floating point</a> value, and a type
3615to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3616type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3617vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3618
3619<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3620<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3621<a href="#t_floating">floating point</a> operand into the nearest (rounding
3622towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3623the results are undefined.</p>
3624
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3625<h5>Example:</h5>
3626<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3627 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3628 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3629 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3630</pre>
3631</div>
3632
3633<!-- _______________________________________________________________________ -->
3634<div class="doc_subsubsection">
3635 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3636</div>
3637<div class="doc_text">
3638
3639<h5>Syntax:</h5>
3640<pre>
3641 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3642</pre>
3643
3644<h5>Overview:</h5>
3645<p>The '<tt>fptosi</tt>' instruction converts
3646<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3647</p>
3648
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3649<h5>Arguments:</h5>
3650<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]3651scalar or vector <a href="#t_floating">floating point</a> value, and a type
3652to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3653type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3654vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3655
3656<h5>Semantics:</h5>
3657<p>The '<tt>fptosi</tt>' instruction converts its
3658<a href="#t_floating">floating point</a> operand into the nearest (rounding
3659towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3660the results are undefined.</p>
3661
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3662<h5>Example:</h5>
3663<pre>
3664 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3665 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3666 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3667</pre>
3668</div>
3669
3670<!-- _______________________________________________________________________ -->
3671<div class="doc_subsubsection">
3672 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3673</div>
3674<div class="doc_text">
3675
3676<h5>Syntax:</h5>
3677<pre>
3678 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3679</pre>
3680
3681<h5>Overview:</h5>
3682<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3683integer and converts that value to the <tt>ty2</tt> type.</p>
3684
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3685<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3686<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3687scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3688to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3689type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3690floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3691
3692<h5>Semantics:</h5>
3693<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3694integer quantity and converts it to the corresponding floating point value. If
3695the value cannot fit in the floating point value, the results are undefined.</p>
3696
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3697<h5>Example:</h5>
3698<pre>
3699 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3700 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3701</pre>
3702</div>
3703
3704<!-- _______________________________________________________________________ -->
3705<div class="doc_subsubsection">
3706 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3707</div>
3708<div class="doc_text">
3709
3710<h5>Syntax:</h5>
3711<pre>
3712 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3713</pre>
3714
3715<h5>Overview:</h5>
3716<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3717integer and converts that value to the <tt>ty2</tt> type.</p>
3718
3719<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3720<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3721scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3722to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3723type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3724floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3725
3726<h5>Semantics:</h5>
3727<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3728integer quantity and converts it to the corresponding floating point value. If
3729the value cannot fit in the floating point value, the results are undefined.</p>
3730
3731<h5>Example:</h5>
3732<pre>
3733 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3734 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3735</pre>
3736</div>
3737
3738<!-- _______________________________________________________________________ -->
3739<div class="doc_subsubsection">
3740 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3741</div>
3742<div class="doc_text">
3743
3744<h5>Syntax:</h5>
3745<pre>
3746 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3747</pre>
3748
3749<h5>Overview:</h5>
3750<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3751the integer type <tt>ty2</tt>.</p>
3752
3753<h5>Arguments:</h5>
3754<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3755must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3756<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3757
3758<h5>Semantics:</h5>
3759<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3760<tt>ty2</tt> by interpreting the pointer value as an integer and either
3761truncating or zero extending that value to the size of the integer type. If
3762<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3763<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3764are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3765change.</p>
3766
3767<h5>Example:</h5>
3768<pre>
3769 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3770 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3771</pre>
3772</div>
3773
3774<!-- _______________________________________________________________________ -->
3775<div class="doc_subsubsection">
3776 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3777</div>
3778<div class="doc_text">
3779
3780<h5>Syntax:</h5>
3781<pre>
3782 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3783</pre>
3784
3785<h5>Overview:</h5>
3786<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3787a pointer type, <tt>ty2</tt>.</p>
3788
3789<h5>Arguments:</h5>
3790<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3791value to cast, and a type to cast it to, which must be a
3792<a href="#t_pointer">pointer</a> type.
3793
3794<h5>Semantics:</h5>
3795<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3796<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3797the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3798size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3799the size of a pointer then a zero extension is done. If they are the same size,
3800nothing is done (<i>no-op cast</i>).</p>
3801
3802<h5>Example:</h5>
3803<pre>
3804 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3805 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3806 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3807</pre>
3808</div>
3809
3810<!-- _______________________________________________________________________ -->
3811<div class="doc_subsubsection">
3812 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3813</div>
3814<div class="doc_text">
3815
3816<h5>Syntax:</h5>
3817<pre>
3818 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3819</pre>
3820
3821<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3822
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3823<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3824<tt>ty2</tt> without changing any bits.</p>
3825
3826<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3827
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3828<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]3829a non-aggregate first class value, and a type to cast it to, which must also be
3830a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
3831<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3832and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3833type is a pointer, the destination type must also be a pointer. This
3834instruction supports bitwise conversion of vectors to integers and to vectors
3835of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3836
3837<h5>Semantics:</h5>
3838<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3839<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3840this conversion. The conversion is done as if the <tt>value</tt> had been
3841stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3842converted to other pointer types with this instruction. To convert pointers to
3843other types, use the <a href="#i_inttoptr">inttoptr</a> or
3844<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3845
3846<h5>Example:</h5>
3847<pre>
3848 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3849 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3850 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3851</pre>
3852</div>
3853
3854<!-- ======================================================================= -->
3855<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3856<div class="doc_text">
3857<p>The instructions in this category are the "miscellaneous"
3858instructions, which defy better classification.</p>
3859</div>
3860
3861<!-- _______________________________________________________________________ -->
3862<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3863</div>
3864<div class="doc_text">
3865<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3866<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]3867</pre>
3868<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3869<p>The '<tt>icmp</tt>' instruction returns a boolean value or
3870a vector of boolean values based on comparison
3871of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3872<h5>Arguments:</h5>
3873<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3874the condition code indicating the kind of comparison to perform. It is not
3875a value, just a keyword. The possible condition code are:
3876<ol>
3877 <li><tt>eq</tt>: equal</li>
3878 <li><tt>ne</tt>: not equal </li>
3879 <li><tt>ugt</tt>: unsigned greater than</li>
3880 <li><tt>uge</tt>: unsigned greater or equal</li>
3881 <li><tt>ult</tt>: unsigned less than</li>
3882 <li><tt>ule</tt>: unsigned less or equal</li>
3883 <li><tt>sgt</tt>: signed greater than</li>
3884 <li><tt>sge</tt>: signed greater or equal</li>
3885 <li><tt>slt</tt>: signed less than</li>
3886 <li><tt>sle</tt>: signed less or equal</li>
3887</ol>
3888<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3889<a href="#t_pointer">pointer</a>
3890or integer <a href="#t_vector">vector</a> typed.
3891They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3892<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3893<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]3894the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3895yields 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]3896<ol>
3897 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3898 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3899 </li>
3900 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3901 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3902 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3903 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3904 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3905 <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]3906 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3907 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3908 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3909 <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]3910 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3911 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3912 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3913 <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]3914 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3915 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3916 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3917 <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]3918</ol>
3919<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3920values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3921<p>If the operands are integer vectors, then they are compared
3922element by element. The result is an <tt>i1</tt> vector with
3923the same number of elements as the values being compared.
3924Otherwise, the result is an <tt>i1</tt>.
3925</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3926
3927<h5>Example:</h5>
3928<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3929 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3930 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3931 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3932 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3933 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3934</pre>
3935</div>
3936
3937<!-- _______________________________________________________________________ -->
3938<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3939</div>
3940<div class="doc_text">
3941<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3942<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]3943</pre>
3944<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3945<p>The '<tt>fcmp</tt>' instruction returns a boolean value
3946or vector of boolean values based on comparison
3947of its operands.
3948<p>
3949If the operands are floating point scalars, then the result
3950type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
3951</p>
3952<p>If the operands are floating point vectors, then the result type
3953is a vector of boolean with the same number of elements as the
3954operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3955<h5>Arguments:</h5>
3956<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3957the condition code indicating the kind of comparison to perform. It is not
3958a value, just a keyword. The possible condition code are:
3959<ol>
3960 <li><tt>false</tt>: no comparison, always returns false</li>
3961 <li><tt>oeq</tt>: ordered and equal</li>
3962 <li><tt>ogt</tt>: ordered and greater than </li>
3963 <li><tt>oge</tt>: ordered and greater than or equal</li>
3964 <li><tt>olt</tt>: ordered and less than </li>
3965 <li><tt>ole</tt>: ordered and less than or equal</li>
3966 <li><tt>one</tt>: ordered and not equal</li>
3967 <li><tt>ord</tt>: ordered (no nans)</li>
3968 <li><tt>ueq</tt>: unordered or equal</li>
3969 <li><tt>ugt</tt>: unordered or greater than </li>
3970 <li><tt>uge</tt>: unordered or greater than or equal</li>
3971 <li><tt>ult</tt>: unordered or less than </li>
3972 <li><tt>ule</tt>: unordered or less than or equal</li>
3973 <li><tt>une</tt>: unordered or not equal</li>
3974 <li><tt>uno</tt>: unordered (either nans)</li>
3975 <li><tt>true</tt>: no comparison, always returns true</li>
3976</ol>
3977<p><i>Ordered</i> means that neither operand is a QNAN while
3978<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3979<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
3980either a <a href="#t_floating">floating point</a> type
3981or a <a href="#t_vector">vector</a> of floating point type.
3982They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3983<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3984<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3985according to the condition code given as <tt>cond</tt>.
3986If the operands are vectors, then the vectors are compared
3987element by element.
3988Each comparison performed
3989always yields an <a href="#t_primitive">i1</a> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3990<ol>
3991 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3992 <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]3993 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3994 <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]3995 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3996 <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]3997 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3998 <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]3999 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4000 <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]4001 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4002 <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]4003 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4004 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4005 <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]4006 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4007 <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]4008 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4009 <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]4010 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4011 <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]4012 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4013 <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]4014 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4015 <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]4016 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4017 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4018 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4019</ol>
4020
4021<h5>Example:</h5>
4022<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]4023 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4024 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4025 &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]4026</pre>
4027</div>
4028
4029<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4030<div class="doc_subsubsection">
4031 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4032</div>
4033<div class="doc_text">
4034<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4035<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]4036</pre>
4037<h5>Overview:</h5>
4038<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4039element-wise comparison of its two integer vector operands.</p>
4040<h5>Arguments:</h5>
4041<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4042the condition code indicating the kind of comparison to perform. It is not
4043a value, just a keyword. The possible condition code are:
4044<ol>
4045 <li><tt>eq</tt>: equal</li>
4046 <li><tt>ne</tt>: not equal </li>
4047 <li><tt>ugt</tt>: unsigned greater than</li>
4048 <li><tt>uge</tt>: unsigned greater or equal</li>
4049 <li><tt>ult</tt>: unsigned less than</li>
4050 <li><tt>ule</tt>: unsigned less or equal</li>
4051 <li><tt>sgt</tt>: signed greater than</li>
4052 <li><tt>sge</tt>: signed greater or equal</li>
4053 <li><tt>slt</tt>: signed less than</li>
4054 <li><tt>sle</tt>: signed less or equal</li>
4055</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4056<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4057<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4058<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4059<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4060according to the condition code given as <tt>cond</tt>. The comparison yields a
4061<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4062identical type as the values being compared. The most significant bit in each
4063element is 1 if the element-wise comparison evaluates to true, and is 0
4064otherwise. All other bits of the result are undefined. The condition codes
4065are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4066instruction</a>.
4067
4068<h5>Example:</h5>
4069<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4070 &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>
4071 &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]4072</pre>
4073</div>
4074
4075<!-- _______________________________________________________________________ -->
4076<div class="doc_subsubsection">
4077 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4078</div>
4079<div class="doc_text">
4080<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4081<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]4082<h5>Overview:</h5>
4083<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4084element-wise comparison of its two floating point vector operands. The output
4085elements have the same width as the input elements.</p>
4086<h5>Arguments:</h5>
4087<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4088the condition code indicating the kind of comparison to perform. It is not
4089a value, just a keyword. The possible condition code are:
4090<ol>
4091 <li><tt>false</tt>: no comparison, always returns false</li>
4092 <li><tt>oeq</tt>: ordered and equal</li>
4093 <li><tt>ogt</tt>: ordered and greater than </li>
4094 <li><tt>oge</tt>: ordered and greater than or equal</li>
4095 <li><tt>olt</tt>: ordered and less than </li>
4096 <li><tt>ole</tt>: ordered and less than or equal</li>
4097 <li><tt>one</tt>: ordered and not equal</li>
4098 <li><tt>ord</tt>: ordered (no nans)</li>
4099 <li><tt>ueq</tt>: unordered or equal</li>
4100 <li><tt>ugt</tt>: unordered or greater than </li>
4101 <li><tt>uge</tt>: unordered or greater than or equal</li>
4102 <li><tt>ult</tt>: unordered or less than </li>
4103 <li><tt>ule</tt>: unordered or less than or equal</li>
4104 <li><tt>une</tt>: unordered or not equal</li>
4105 <li><tt>uno</tt>: unordered (either nans)</li>
4106 <li><tt>true</tt>: no comparison, always returns true</li>
4107</ol>
4108<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4109<a href="#t_floating">floating point</a> typed. They must also be identical
4110types.</p>
4111<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4112<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4113according to the condition code given as <tt>cond</tt>. The comparison yields a
4114<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4115an identical number of elements as the values being compared, and each element
4116having identical with to the width of the floating point elements. The most
4117significant bit in each element is 1 if the element-wise comparison evaluates to
4118true, and is 0 otherwise. All other bits of the result are undefined. The
4119condition codes are evaluated identically to the
4120<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.
4121
4122<h5>Example:</h5>
4123<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4124 &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>
4125 &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]4126</pre>
4127</div>
4128
4129<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4130<div class="doc_subsubsection">
4131 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4132</div>
4133
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4134<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4135
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4136<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4137
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4138<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4139<h5>Overview:</h5>
4140<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4141the SSA graph representing the function.</p>
4142<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4143
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4144<p>The type of the incoming values is specified with the first type
4145field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4146as arguments, with one pair for each predecessor basic block of the
4147current block. Only values of <a href="#t_firstclass">first class</a>
4148type may be used as the value arguments to the PHI node. Only labels
4149may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4150
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4151<p>There must be no non-phi instructions between the start of a basic
4152block and the PHI instructions: i.e. PHI instructions must be first in
4153a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4154
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4155<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4156
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4157<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4158specified by the pair corresponding to the predecessor basic block that executed
4159just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4160
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4161<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4162<pre>
4163Loop: ; Infinite loop that counts from 0 on up...
4164 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4165 %nextindvar = add i32 %indvar, 1
4166 br label %Loop
4167</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4168</div>
4169
4170<!-- _______________________________________________________________________ -->
4171<div class="doc_subsubsection">
4172 <a name="i_select">'<tt>select</tt>' Instruction</a>
4173</div>
4174
4175<div class="doc_text">
4176
4177<h5>Syntax:</h5>
4178
4179<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4180 &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>
4181
4182 <i>selty</i> is either i1 or {&lt;N x i1&gt}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4183</pre>
4184
4185<h5>Overview:</h5>
4186
4187<p>
4188The '<tt>select</tt>' instruction is used to choose one value based on a
4189condition, without branching.
4190</p>
4191
4192
4193<h5>Arguments:</h5>
4194
4195<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4196The '<tt>select</tt>' instruction requires an 'i1' value or
4197a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4198condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4199type. If the val1/val2 are vectors and
4200the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4201individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4202</p>
4203
4204<h5>Semantics:</h5>
4205
4206<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4207If 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]4208value argument; otherwise, it returns the second value argument.
4209</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4210<p>
4211If the condition is a vector of i1, then the value arguments must
4212be vectors of the same size, and the selection is done element
4213by element.
4214</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4215
4216<h5>Example:</h5>
4217
4218<pre>
4219 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4220</pre>
4221</div>
4222
4223
4224<!-- _______________________________________________________________________ -->
4225<div class="doc_subsubsection">
4226 <a name="i_call">'<tt>call</tt>' Instruction</a>
4227</div>
4228
4229<div class="doc_text">
4230
4231<h5>Syntax:</h5>
4232<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4233 &lt;result&gt; = [tail] call [<a href="#callingconv">cconv</a>] &lt;ty&gt; [&lt;fnty&gt;*] &lt;fnptrval&gt;(&lt;param list&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4234</pre>
4235
4236<h5>Overview:</h5>
4237
4238<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4239
4240<h5>Arguments:</h5>
4241
4242<p>This instruction requires several arguments:</p>
4243
4244<ol>
4245 <li>
4246 <p>The optional "tail" marker indicates whether the callee function accesses
4247 any allocas or varargs in the caller. If the "tail" marker is present, the
4248 function call is eligible for tail call optimization. Note that calls may
4249 be marked "tail" even if they do not occur before a <a
4250 href="#i_ret"><tt>ret</tt></a> instruction.
4251 </li>
4252 <li>
4253 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4254 convention</a> the call should use. If none is specified, the call defaults
4255 to using C calling conventions.
4256 </li>
4257 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4258 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4259 the type of the return value. Functions that return no value are marked
4260 <tt><a href="#t_void">void</a></tt>.</p>
4261 </li>
4262 <li>
4263 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4264 value being invoked. The argument types must match the types implied by
4265 this signature. This type can be omitted if the function is not varargs
4266 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4267 </li>
4268 <li>
4269 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4270 be invoked. In most cases, this is a direct function invocation, but
4271 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4272 to function value.</p>
4273 </li>
4274 <li>
4275 <p>'<tt>function args</tt>': argument list whose types match the
4276 function signature argument types. All arguments must be of
4277 <a href="#t_firstclass">first class</a> type. If the function signature
4278 indicates the function accepts a variable number of arguments, the extra
4279 arguments can be specified.</p>
4280 </li>
4281</ol>
4282
4283<h5>Semantics:</h5>
4284
4285<p>The '<tt>call</tt>' instruction is used to cause control flow to
4286transfer to a specified function, with its incoming arguments bound to
4287the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4288instruction in the called function, control flow continues with the
4289instruction after the function call, and the return value of the
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4290function is bound to the result argument.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4291
4292<h5>Example:</h5>
4293
4294<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4295 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4296 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4297 %X = tail call i32 @foo() <i>; yields i32</i>
4298 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4299 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4300
4301 %struct.A = type { i32, i8 }
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4302 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4303 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4304 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4305</pre>
4306
4307</div>
4308
4309<!-- _______________________________________________________________________ -->
4310<div class="doc_subsubsection">
4311 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4312</div>
4313
4314<div class="doc_text">
4315
4316<h5>Syntax:</h5>
4317
4318<pre>
4319 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4320</pre>
4321
4322<h5>Overview:</h5>
4323
4324<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4325the "variable argument" area of a function call. It is used to implement the
4326<tt>va_arg</tt> macro in C.</p>
4327
4328<h5>Arguments:</h5>
4329
4330<p>This instruction takes a <tt>va_list*</tt> value and the type of
4331the argument. It returns a value of the specified argument type and
4332increments the <tt>va_list</tt> to point to the next argument. The
4333actual type of <tt>va_list</tt> is target specific.</p>
4334
4335<h5>Semantics:</h5>
4336
4337<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4338type from the specified <tt>va_list</tt> and causes the
4339<tt>va_list</tt> to point to the next argument. For more information,
4340see the variable argument handling <a href="#int_varargs">Intrinsic
4341Functions</a>.</p>
4342
4343<p>It is legal for this instruction to be called in a function which does not
4344take a variable number of arguments, for example, the <tt>vfprintf</tt>
4345function.</p>
4346
4347<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4348href="#intrinsics">intrinsic function</a> because it takes a type as an
4349argument.</p>
4350
4351<h5>Example:</h5>
4352
4353<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4354
4355</div>
4356
4357<!-- *********************************************************************** -->
4358<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4359<!-- *********************************************************************** -->
4360
4361<div class="doc_text">
4362
4363<p>LLVM supports the notion of an "intrinsic function". These functions have
4364well known names and semantics and are required to follow certain restrictions.
4365Overall, these intrinsics represent an extension mechanism for the LLVM
4366language that does not require changing all of the transformations in LLVM when
4367adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4368
4369<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4370prefix is reserved in LLVM for intrinsic names; thus, function names may not
4371begin with this prefix. Intrinsic functions must always be external functions:
4372you cannot define the body of intrinsic functions. Intrinsic functions may
4373only be used in call or invoke instructions: it is illegal to take the address
4374of an intrinsic function. Additionally, because intrinsic functions are part
4375of the LLVM language, it is required if any are added that they be documented
4376here.</p>
4377
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4378<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4379a family of functions that perform the same operation but on different data
4380types. Because LLVM can represent over 8 million different integer types,
4381overloading is used commonly to allow an intrinsic function to operate on any
4382integer type. One or more of the argument types or the result type can be
4383overloaded to accept any integer type. Argument types may also be defined as
4384exactly matching a previous argument's type or the result type. This allows an
4385intrinsic function which accepts multiple arguments, but needs all of them to
4386be of the same type, to only be overloaded with respect to a single argument or
4387the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4388
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4389<p>Overloaded intrinsics will have the names of its overloaded argument types
4390encoded into its function name, each preceded by a period. Only those types
4391which are overloaded result in a name suffix. Arguments whose type is matched
4392against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4393take an integer of any width and returns an integer of exactly the same integer
4394width. This leads to a family of functions such as
4395<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4396Only one type, the return type, is overloaded, and only one type suffix is
4397required. Because the argument's type is matched against the return type, it
4398does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4399
4400<p>To learn how to add an intrinsic function, please see the
4401<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4402</p>
4403
4404</div>
4405
4406<!-- ======================================================================= -->
4407<div class="doc_subsection">
4408 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4409</div>
4410
4411<div class="doc_text">
4412
4413<p>Variable argument support is defined in LLVM with the <a
4414 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4415intrinsic functions. These functions are related to the similarly
4416named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4417
4418<p>All of these functions operate on arguments that use a
4419target-specific value type "<tt>va_list</tt>". The LLVM assembly
4420language reference manual does not define what this type is, so all
4421transformations should be prepared to handle these functions regardless of
4422the type used.</p>
4423
4424<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4425instruction and the variable argument handling intrinsic functions are
4426used.</p>
4427
4428<div class="doc_code">
4429<pre>
4430define i32 @test(i32 %X, ...) {
4431 ; Initialize variable argument processing
4432 %ap = alloca i8*
4433 %ap2 = bitcast i8** %ap to i8*
4434 call void @llvm.va_start(i8* %ap2)
4435
4436 ; Read a single integer argument
4437 %tmp = va_arg i8** %ap, i32
4438
4439 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4440 %aq = alloca i8*
4441 %aq2 = bitcast i8** %aq to i8*
4442 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4443 call void @llvm.va_end(i8* %aq2)
4444
4445 ; Stop processing of arguments.
4446 call void @llvm.va_end(i8* %ap2)
4447 ret i32 %tmp
4448}
4449
4450declare void @llvm.va_start(i8*)
4451declare void @llvm.va_copy(i8*, i8*)
4452declare void @llvm.va_end(i8*)
4453</pre>
4454</div>
4455
4456</div>
4457
4458<!-- _______________________________________________________________________ -->
4459<div class="doc_subsubsection">
4460 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4461</div>
4462
4463
4464<div class="doc_text">
4465<h5>Syntax:</h5>
4466<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4467<h5>Overview:</h5>
4468<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
4469<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4470href="#i_va_arg">va_arg</a></tt>.</p>
4471
4472<h5>Arguments:</h5>
4473
4474<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4475
4476<h5>Semantics:</h5>
4477
4478<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4479macro available in C. In a target-dependent way, it initializes the
4480<tt>va_list</tt> element to which the argument points, so that the next call to
4481<tt>va_arg</tt> will produce the first variable argument passed to the function.
4482Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4483last argument of the function as the compiler can figure that out.</p>
4484
4485</div>
4486
4487<!-- _______________________________________________________________________ -->
4488<div class="doc_subsubsection">
4489 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4490</div>
4491
4492<div class="doc_text">
4493<h5>Syntax:</h5>
4494<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4495<h5>Overview:</h5>
4496
4497<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4498which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4499or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4500
4501<h5>Arguments:</h5>
4502
4503<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4504
4505<h5>Semantics:</h5>
4506
4507<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4508macro available in C. In a target-dependent way, it destroys the
4509<tt>va_list</tt> element to which the argument points. Calls to <a
4510href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4511<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4512<tt>llvm.va_end</tt>.</p>
4513
4514</div>
4515
4516<!-- _______________________________________________________________________ -->
4517<div class="doc_subsubsection">
4518 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4519</div>
4520
4521<div class="doc_text">
4522
4523<h5>Syntax:</h5>
4524
4525<pre>
4526 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4527</pre>
4528
4529<h5>Overview:</h5>
4530
4531<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4532from the source argument list to the destination argument list.</p>
4533
4534<h5>Arguments:</h5>
4535
4536<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4537The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4538
4539
4540<h5>Semantics:</h5>
4541
4542<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4543macro available in C. In a target-dependent way, it copies the source
4544<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4545intrinsic is necessary because the <tt><a href="#int_va_start">
4546llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4547example, memory allocation.</p>
4548
4549</div>
4550
4551<!-- ======================================================================= -->
4552<div class="doc_subsection">
4553 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4554</div>
4555
4556<div class="doc_text">
4557
4558<p>
4559LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4560Collection</a> (GC) requires the implementation and generation of these
4561intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4562These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4563stack</a>, as well as garbage collector implementations that require <a
4564href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4565Front-ends for type-safe garbage collected languages should generate these
4566intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4567href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4568</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4569
4570<p>The garbage collection intrinsics only operate on objects in the generic
4571 address space (address space zero).</p>
4572
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4573</div>
4574
4575<!-- _______________________________________________________________________ -->
4576<div class="doc_subsubsection">
4577 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4578</div>
4579
4580<div class="doc_text">
4581
4582<h5>Syntax:</h5>
4583
4584<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4585 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4586</pre>
4587
4588<h5>Overview:</h5>
4589
4590<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4591the code generator, and allows some metadata to be associated with it.</p>
4592
4593<h5>Arguments:</h5>
4594
4595<p>The first argument specifies the address of a stack object that contains the
4596root pointer. The second pointer (which must be either a constant or a global
4597value address) contains the meta-data to be associated with the root.</p>
4598
4599<h5>Semantics:</h5>
4600
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4601<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]4602location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4603the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4604intrinsic may only be used in a function which <a href="#gc">specifies a GC
4605algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4606
4607</div>
4608
4609
4610<!-- _______________________________________________________________________ -->
4611<div class="doc_subsubsection">
4612 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4613</div>
4614
4615<div class="doc_text">
4616
4617<h5>Syntax:</h5>
4618
4619<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4620 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4621</pre>
4622
4623<h5>Overview:</h5>
4624
4625<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4626locations, allowing garbage collector implementations that require read
4627barriers.</p>
4628
4629<h5>Arguments:</h5>
4630
4631<p>The second argument is the address to read from, which should be an address
4632allocated from the garbage collector. The first object is a pointer to the
4633start of the referenced object, if needed by the language runtime (otherwise
4634null).</p>
4635
4636<h5>Semantics:</h5>
4637
4638<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4639instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4640garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4641may only be used in a function which <a href="#gc">specifies a GC
4642algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4643
4644</div>
4645
4646
4647<!-- _______________________________________________________________________ -->
4648<div class="doc_subsubsection">
4649 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4650</div>
4651
4652<div class="doc_text">
4653
4654<h5>Syntax:</h5>
4655
4656<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4657 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4658</pre>
4659
4660<h5>Overview:</h5>
4661
4662<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4663locations, allowing garbage collector implementations that require write
4664barriers (such as generational or reference counting collectors).</p>
4665
4666<h5>Arguments:</h5>
4667
4668<p>The first argument is the reference to store, the second is the start of the
4669object to store it to, and the third is the address of the field of Obj to
4670store to. If the runtime does not require a pointer to the object, Obj may be
4671null.</p>
4672
4673<h5>Semantics:</h5>
4674
4675<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4676instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4677garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4678may only be used in a function which <a href="#gc">specifies a GC
4679algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4680
4681</div>
4682
4683
4684
4685<!-- ======================================================================= -->
4686<div class="doc_subsection">
4687 <a name="int_codegen">Code Generator Intrinsics</a>
4688</div>
4689
4690<div class="doc_text">
4691<p>
4692These intrinsics are provided by LLVM to expose special features that may only
4693be implemented with code generator support.
4694</p>
4695
4696</div>
4697
4698<!-- _______________________________________________________________________ -->
4699<div class="doc_subsubsection">
4700 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4701</div>
4702
4703<div class="doc_text">
4704
4705<h5>Syntax:</h5>
4706<pre>
4707 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4708</pre>
4709
4710<h5>Overview:</h5>
4711
4712<p>
4713The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4714target-specific value indicating the return address of the current function
4715or one of its callers.
4716</p>
4717
4718<h5>Arguments:</h5>
4719
4720<p>
4721The argument to this intrinsic indicates which function to return the address
4722for. Zero indicates the calling function, one indicates its caller, etc. The
4723argument is <b>required</b> to be a constant integer value.
4724</p>
4725
4726<h5>Semantics:</h5>
4727
4728<p>
4729The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4730the return address of the specified call frame, or zero if it cannot be
4731identified. The value returned by this intrinsic is likely to be incorrect or 0
4732for arguments other than zero, so it should only be used for debugging purposes.
4733</p>
4734
4735<p>
4736Note that calling this intrinsic does not prevent function inlining or other
4737aggressive transformations, so the value returned may not be that of the obvious
4738source-language caller.
4739</p>
4740</div>
4741
4742
4743<!-- _______________________________________________________________________ -->
4744<div class="doc_subsubsection">
4745 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4746</div>
4747
4748<div class="doc_text">
4749
4750<h5>Syntax:</h5>
4751<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4752 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4753</pre>
4754
4755<h5>Overview:</h5>
4756
4757<p>
4758The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4759target-specific frame pointer value for the specified stack frame.
4760</p>
4761
4762<h5>Arguments:</h5>
4763
4764<p>
4765The argument to this intrinsic indicates which function to return the frame
4766pointer for. Zero indicates the calling function, one indicates its caller,
4767etc. The argument is <b>required</b> to be a constant integer value.
4768</p>
4769
4770<h5>Semantics:</h5>
4771
4772<p>
4773The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4774the frame address of the specified call frame, or zero if it cannot be
4775identified. The value returned by this intrinsic is likely to be incorrect or 0
4776for arguments other than zero, so it should only be used for debugging purposes.
4777</p>
4778
4779<p>
4780Note that calling this intrinsic does not prevent function inlining or other
4781aggressive transformations, so the value returned may not be that of the obvious
4782source-language caller.
4783</p>
4784</div>
4785
4786<!-- _______________________________________________________________________ -->
4787<div class="doc_subsubsection">
4788 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4789</div>
4790
4791<div class="doc_text">
4792
4793<h5>Syntax:</h5>
4794<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4795 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4796</pre>
4797
4798<h5>Overview:</h5>
4799
4800<p>
4801The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4802the function stack, for use with <a href="#int_stackrestore">
4803<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4804features like scoped automatic variable sized arrays in C99.
4805</p>
4806
4807<h5>Semantics:</h5>
4808
4809<p>
4810This intrinsic returns a opaque pointer value that can be passed to <a
4811href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4812<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4813<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4814state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4815practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4816that were allocated after the <tt>llvm.stacksave</tt> was executed.
4817</p>
4818
4819</div>
4820
4821<!-- _______________________________________________________________________ -->
4822<div class="doc_subsubsection">
4823 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4824</div>
4825
4826<div class="doc_text">
4827
4828<h5>Syntax:</h5>
4829<pre>
4830 declare void @llvm.stackrestore(i8 * %ptr)
4831</pre>
4832
4833<h5>Overview:</h5>
4834
4835<p>
4836The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4837the function stack to the state it was in when the corresponding <a
4838href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4839useful for implementing language features like scoped automatic variable sized
4840arrays in C99.
4841</p>
4842
4843<h5>Semantics:</h5>
4844
4845<p>
4846See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4847</p>
4848
4849</div>
4850
4851
4852<!-- _______________________________________________________________________ -->
4853<div class="doc_subsubsection">
4854 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4855</div>
4856
4857<div class="doc_text">
4858
4859<h5>Syntax:</h5>
4860<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4861 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]4862</pre>
4863
4864<h5>Overview:</h5>
4865
4866
4867<p>
4868The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4869a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4870no
4871effect on the behavior of the program but can change its performance
4872characteristics.
4873</p>
4874
4875<h5>Arguments:</h5>
4876
4877<p>
4878<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4879determining if the fetch should be for a read (0) or write (1), and
4880<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4881locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4882<tt>locality</tt> arguments must be constant integers.
4883</p>
4884
4885<h5>Semantics:</h5>
4886
4887<p>
4888This intrinsic does not modify the behavior of the program. In particular,
4889prefetches cannot trap and do not produce a value. On targets that support this
4890intrinsic, the prefetch can provide hints to the processor cache for better
4891performance.
4892</p>
4893
4894</div>
4895
4896<!-- _______________________________________________________________________ -->
4897<div class="doc_subsubsection">
4898 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4899</div>
4900
4901<div class="doc_text">
4902
4903<h5>Syntax:</h5>
4904<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4905 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4906</pre>
4907
4908<h5>Overview:</h5>
4909
4910
4911<p>
4912The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4913(PC) in a region of
4914code to simulators and other tools. The method is target specific, but it is
4915expected that the marker will use exported symbols to transmit the PC of the
4916marker.
4917The marker makes no guarantees that it will remain with any specific instruction
4918after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4919optimizations. The intended use is to be inserted after optimizations to allow
4920correlations of simulation runs.
4921</p>
4922
4923<h5>Arguments:</h5>
4924
4925<p>
4926<tt>id</tt> is a numerical id identifying the marker.
4927</p>
4928
4929<h5>Semantics:</h5>
4930
4931<p>
4932This intrinsic does not modify the behavior of the program. Backends that do not
4933support this intrinisic may ignore it.
4934</p>
4935
4936</div>
4937
4938<!-- _______________________________________________________________________ -->
4939<div class="doc_subsubsection">
4940 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4941</div>
4942
4943<div class="doc_text">
4944
4945<h5>Syntax:</h5>
4946<pre>
4947 declare i64 @llvm.readcyclecounter( )
4948</pre>
4949
4950<h5>Overview:</h5>
4951
4952
4953<p>
4954The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4955counter register (or similar low latency, high accuracy clocks) on those targets
4956that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4957As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4958should only be used for small timings.
4959</p>
4960
4961<h5>Semantics:</h5>
4962
4963<p>
4964When directly supported, reading the cycle counter should not modify any memory.
4965Implementations are allowed to either return a application specific value or a
4966system wide value. On backends without support, this is lowered to a constant 0.
4967</p>
4968
4969</div>
4970
4971<!-- ======================================================================= -->
4972<div class="doc_subsection">
4973 <a name="int_libc">Standard C Library Intrinsics</a>
4974</div>
4975
4976<div class="doc_text">
4977<p>
4978LLVM provides intrinsics for a few important standard C library functions.
4979These intrinsics allow source-language front-ends to pass information about the
4980alignment of the pointer arguments to the code generator, providing opportunity
4981for more efficient code generation.
4982</p>
4983
4984</div>
4985
4986<!-- _______________________________________________________________________ -->
4987<div class="doc_subsubsection">
4988 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4989</div>
4990
4991<div class="doc_text">
4992
4993<h5>Syntax:</h5>
4994<pre>
4995 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4996 i32 &lt;len&gt;, i32 &lt;align&gt;)
4997 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4998 i64 &lt;len&gt;, i32 &lt;align&gt;)
4999</pre>
5000
5001<h5>Overview:</h5>
5002
5003<p>
5004The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5005location to the destination location.
5006</p>
5007
5008<p>
5009Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5010intrinsics do not return a value, and takes an extra alignment argument.
5011</p>
5012
5013<h5>Arguments:</h5>
5014
5015<p>
5016The first argument is a pointer to the destination, the second is a pointer to
5017the source. The third argument is an integer argument
5018specifying the number of bytes to copy, and the fourth argument is the alignment
5019of the source and destination locations.
5020</p>
5021
5022<p>
5023If the call to this intrinisic has an alignment value that is not 0 or 1, then
5024the caller guarantees that both the source and destination pointers are aligned
5025to that boundary.
5026</p>
5027
5028<h5>Semantics:</h5>
5029
5030<p>
5031The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5032location to the destination location, which are not allowed to overlap. It
5033copies "len" bytes of memory over. If the argument is known to be aligned to
5034some boundary, this can be specified as the fourth argument, otherwise it should
5035be set to 0 or 1.
5036</p>
5037</div>
5038
5039
5040<!-- _______________________________________________________________________ -->
5041<div class="doc_subsubsection">
5042 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5043</div>
5044
5045<div class="doc_text">
5046
5047<h5>Syntax:</h5>
5048<pre>
5049 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5050 i32 &lt;len&gt;, i32 &lt;align&gt;)
5051 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5052 i64 &lt;len&gt;, i32 &lt;align&gt;)
5053</pre>
5054
5055<h5>Overview:</h5>
5056
5057<p>
5058The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5059location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5060'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5061</p>
5062
5063<p>
5064Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5065intrinsics do not return a value, and takes an extra alignment argument.
5066</p>
5067
5068<h5>Arguments:</h5>
5069
5070<p>
5071The first argument is a pointer to the destination, the second is a pointer to
5072the source. The third argument is an integer argument
5073specifying the number of bytes to copy, and the fourth argument is the alignment
5074of the source and destination locations.
5075</p>
5076
5077<p>
5078If the call to this intrinisic has an alignment value that is not 0 or 1, then
5079the caller guarantees that the source and destination pointers are aligned to
5080that boundary.
5081</p>
5082
5083<h5>Semantics:</h5>
5084
5085<p>
5086The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5087location to the destination location, which may overlap. It
5088copies "len" bytes of memory over. If the argument is known to be aligned to
5089some boundary, this can be specified as the fourth argument, otherwise it should
5090be set to 0 or 1.
5091</p>
5092</div>
5093
5094
5095<!-- _______________________________________________________________________ -->
5096<div class="doc_subsubsection">
5097 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5098</div>
5099
5100<div class="doc_text">
5101
5102<h5>Syntax:</h5>
5103<pre>
5104 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5105 i32 &lt;len&gt;, i32 &lt;align&gt;)
5106 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5107 i64 &lt;len&gt;, i32 &lt;align&gt;)
5108</pre>
5109
5110<h5>Overview:</h5>
5111
5112<p>
5113The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5114byte value.
5115</p>
5116
5117<p>
5118Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5119does not return a value, and takes an extra alignment argument.
5120</p>
5121
5122<h5>Arguments:</h5>
5123
5124<p>
5125The first argument is a pointer to the destination to fill, the second is the
5126byte value to fill it with, the third argument is an integer
5127argument specifying the number of bytes to fill, and the fourth argument is the
5128known alignment of destination location.
5129</p>
5130
5131<p>
5132If the call to this intrinisic has an alignment value that is not 0 or 1, then
5133the caller guarantees that the destination pointer is aligned to that boundary.
5134</p>
5135
5136<h5>Semantics:</h5>
5137
5138<p>
5139The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5140the
5141destination location. If the argument is known to be aligned to some boundary,
5142this can be specified as the fourth argument, otherwise it should be set to 0 or
51431.
5144</p>
5145</div>
5146
5147
5148<!-- _______________________________________________________________________ -->
5149<div class="doc_subsubsection">
5150 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5151</div>
5152
5153<div class="doc_text">
5154
5155<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5156<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]5157floating point or vector of floating point type. Not all targets support all
5158types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5159<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5160 declare float @llvm.sqrt.f32(float %Val)
5161 declare double @llvm.sqrt.f64(double %Val)
5162 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5163 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5164 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5165</pre>
5166
5167<h5>Overview:</h5>
5168
5169<p>
5170The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5171returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5172<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5173negative numbers other than -0.0 (which allows for better optimization, because
5174there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5175defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5176</p>
5177
5178<h5>Arguments:</h5>
5179
5180<p>
5181The argument and return value are floating point numbers of the same type.
5182</p>
5183
5184<h5>Semantics:</h5>
5185
5186<p>
5187This function returns the sqrt of the specified operand if it is a nonnegative
5188floating point number.
5189</p>
5190</div>
5191
5192<!-- _______________________________________________________________________ -->
5193<div class="doc_subsubsection">
5194 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5195</div>
5196
5197<div class="doc_text">
5198
5199<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5200<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]5201floating point or vector of floating point type. Not all targets support all
5202types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5203<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5204 declare float @llvm.powi.f32(float %Val, i32 %power)
5205 declare double @llvm.powi.f64(double %Val, i32 %power)
5206 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5207 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5208 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5209</pre>
5210
5211<h5>Overview:</h5>
5212
5213<p>
5214The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5215specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5216multiplications is not defined. When a vector of floating point type is
5217used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5218</p>
5219
5220<h5>Arguments:</h5>
5221
5222<p>
5223The second argument is an integer power, and the first is a value to raise to
5224that power.
5225</p>
5226
5227<h5>Semantics:</h5>
5228
5229<p>
5230This function returns the first value raised to the second power with an
5231unspecified sequence of rounding operations.</p>
5232</div>
5233
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5234<!-- _______________________________________________________________________ -->
5235<div class="doc_subsubsection">
5236 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5237</div>
5238
5239<div class="doc_text">
5240
5241<h5>Syntax:</h5>
5242<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5243floating point or vector of floating point type. Not all targets support all
5244types however.
5245<pre>
5246 declare float @llvm.sin.f32(float %Val)
5247 declare double @llvm.sin.f64(double %Val)
5248 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5249 declare fp128 @llvm.sin.f128(fp128 %Val)
5250 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5251</pre>
5252
5253<h5>Overview:</h5>
5254
5255<p>
5256The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5257</p>
5258
5259<h5>Arguments:</h5>
5260
5261<p>
5262The argument and return value are floating point numbers of the same type.
5263</p>
5264
5265<h5>Semantics:</h5>
5266
5267<p>
5268This function returns the sine of the specified operand, returning the
5269same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5270conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5271</div>
5272
5273<!-- _______________________________________________________________________ -->
5274<div class="doc_subsubsection">
5275 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5276</div>
5277
5278<div class="doc_text">
5279
5280<h5>Syntax:</h5>
5281<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5282floating point or vector of floating point type. Not all targets support all
5283types however.
5284<pre>
5285 declare float @llvm.cos.f32(float %Val)
5286 declare double @llvm.cos.f64(double %Val)
5287 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5288 declare fp128 @llvm.cos.f128(fp128 %Val)
5289 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5290</pre>
5291
5292<h5>Overview:</h5>
5293
5294<p>
5295The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5296</p>
5297
5298<h5>Arguments:</h5>
5299
5300<p>
5301The argument and return value are floating point numbers of the same type.
5302</p>
5303
5304<h5>Semantics:</h5>
5305
5306<p>
5307This function returns the cosine of the specified operand, returning the
5308same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5309conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5310</div>
5311
5312<!-- _______________________________________________________________________ -->
5313<div class="doc_subsubsection">
5314 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5315</div>
5316
5317<div class="doc_text">
5318
5319<h5>Syntax:</h5>
5320<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5321floating point or vector of floating point type. Not all targets support all
5322types however.
5323<pre>
5324 declare float @llvm.pow.f32(float %Val, float %Power)
5325 declare double @llvm.pow.f64(double %Val, double %Power)
5326 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5327 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5328 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5329</pre>
5330
5331<h5>Overview:</h5>
5332
5333<p>
5334The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5335specified (positive or negative) power.
5336</p>
5337
5338<h5>Arguments:</h5>
5339
5340<p>
5341The second argument is a floating point power, and the first is a value to
5342raise to that power.
5343</p>
5344
5345<h5>Semantics:</h5>
5346
5347<p>
5348This function returns the first value raised to the second power,
5349returning the
5350same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5351conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5352</div>
5353
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5354
5355<!-- ======================================================================= -->
5356<div class="doc_subsection">
5357 <a name="int_manip">Bit Manipulation Intrinsics</a>
5358</div>
5359
5360<div class="doc_text">
5361<p>
5362LLVM provides intrinsics for a few important bit manipulation operations.
5363These allow efficient code generation for some algorithms.
5364</p>
5365
5366</div>
5367
5368<!-- _______________________________________________________________________ -->
5369<div class="doc_subsubsection">
5370 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5371</div>
5372
5373<div class="doc_text">
5374
5375<h5>Syntax:</h5>
5376<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5377type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5378<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5379 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5380 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5381 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5382</pre>
5383
5384<h5>Overview:</h5>
5385
5386<p>
5387The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5388values with an even number of bytes (positive multiple of 16 bits). These are
5389useful for performing operations on data that is not in the target's native
5390byte order.
5391</p>
5392
5393<h5>Semantics:</h5>
5394
5395<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5396The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5397and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5398intrinsic returns an i32 value that has the four bytes of the input i32
5399swapped, 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]5400i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5401<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5402additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5403</p>
5404
5405</div>
5406
5407<!-- _______________________________________________________________________ -->
5408<div class="doc_subsubsection">
5409 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5410</div>
5411
5412<div class="doc_text">
5413
5414<h5>Syntax:</h5>
5415<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5416width. Not all targets support all bit widths however.
5417<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5418 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
5419 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5420 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5421 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5422 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5423</pre>
5424
5425<h5>Overview:</h5>
5426
5427<p>
5428The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5429value.
5430</p>
5431
5432<h5>Arguments:</h5>
5433
5434<p>
5435The only argument is the value to be counted. The argument may be of any
5436integer type. The return type must match the argument type.
5437</p>
5438
5439<h5>Semantics:</h5>
5440
5441<p>
5442The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5443</p>
5444</div>
5445
5446<!-- _______________________________________________________________________ -->
5447<div class="doc_subsubsection">
5448 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5449</div>
5450
5451<div class="doc_text">
5452
5453<h5>Syntax:</h5>
5454<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5455integer bit width. Not all targets support all bit widths however.
5456<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5457 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5458 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5459 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5460 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5461 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5462</pre>
5463
5464<h5>Overview:</h5>
5465
5466<p>
5467The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5468leading zeros in a variable.
5469</p>
5470
5471<h5>Arguments:</h5>
5472
5473<p>
5474The only argument is the value to be counted. The argument may be of any
5475integer type. The return type must match the argument type.
5476</p>
5477
5478<h5>Semantics:</h5>
5479
5480<p>
5481The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5482in a variable. If the src == 0 then the result is the size in bits of the type
5483of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5484</p>
5485</div>
5486
5487
5488
5489<!-- _______________________________________________________________________ -->
5490<div class="doc_subsubsection">
5491 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5492</div>
5493
5494<div class="doc_text">
5495
5496<h5>Syntax:</h5>
5497<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5498integer bit width. Not all targets support all bit widths however.
5499<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5500 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5501 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5502 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5503 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5504 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5505</pre>
5506
5507<h5>Overview:</h5>
5508
5509<p>
5510The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5511trailing zeros.
5512</p>
5513
5514<h5>Arguments:</h5>
5515
5516<p>
5517The only argument is the value to be counted. The argument may be of any
5518integer type. The return type must match the argument type.
5519</p>
5520
5521<h5>Semantics:</h5>
5522
5523<p>
5524The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5525in a variable. If the src == 0 then the result is the size in bits of the type
5526of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5527</p>
5528</div>
5529
5530<!-- _______________________________________________________________________ -->
5531<div class="doc_subsubsection">
5532 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5533</div>
5534
5535<div class="doc_text">
5536
5537<h5>Syntax:</h5>
5538<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5539on any integer bit width.
5540<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5541 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5542 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5543</pre>
5544
5545<h5>Overview:</h5>
5546<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5547range of bits from an integer value and returns them in the same bit width as
5548the original value.</p>
5549
5550<h5>Arguments:</h5>
5551<p>The first argument, <tt>%val</tt> and the result may be integer types of
5552any bit width but they must have the same bit width. The second and third
5553arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5554
5555<h5>Semantics:</h5>
5556<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5557of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5558<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5559operates in forward mode.</p>
5560<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5561right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5562only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5563<ol>
5564 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5565 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5566 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5567 to determine the number of bits to retain.</li>
5568 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5569 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5570</ol>
5571<p>In reverse mode, a similar computation is made except that the bits are
5572returned in the reverse order. So, for example, if <tt>X</tt> has the value
5573<tt>i16 0x0ACF (101011001111)</tt> and we apply
5574<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5575<tt>i16 0x0026 (000000100110)</tt>.</p>
5576</div>
5577
5578<div class="doc_subsubsection">
5579 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5580</div>
5581
5582<div class="doc_text">
5583
5584<h5>Syntax:</h5>
5585<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5586on any integer bit width.
5587<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5588 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5589 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]5590</pre>
5591
5592<h5>Overview:</h5>
5593<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5594of bits in an integer value with another integer value. It returns the integer
5595with the replaced bits.</p>
5596
5597<h5>Arguments:</h5>
5598<p>The first argument, <tt>%val</tt> and the result may be integer types of
5599any bit width but they must have the same bit width. <tt>%val</tt> is the value
5600whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5601integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5602type since they specify only a bit index.</p>
5603
5604<h5>Semantics:</h5>
5605<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5606of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5607<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5608operates in forward mode.</p>
5609<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5610truncating it down to the size of the replacement area or zero extending it
5611up to that size.</p>
5612<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5613are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5614in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5615to the <tt>%hi</tt>th bit.
5616<p>In reverse mode, a similar computation is made except that the bits are
5617reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5618<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5619<h5>Examples:</h5>
5620<pre>
5621 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5622 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5623 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5624 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5625 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5626</pre>
5627</div>
5628
5629<!-- ======================================================================= -->
5630<div class="doc_subsection">
5631 <a name="int_debugger">Debugger Intrinsics</a>
5632</div>
5633
5634<div class="doc_text">
5635<p>
5636The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5637are described in the <a
5638href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5639Debugging</a> document.
5640</p>
5641</div>
5642
5643
5644<!-- ======================================================================= -->
5645<div class="doc_subsection">
5646 <a name="int_eh">Exception Handling Intrinsics</a>
5647</div>
5648
5649<div class="doc_text">
5650<p> The LLVM exception handling intrinsics (which all start with
5651<tt>llvm.eh.</tt> prefix), are described in the <a
5652href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5653Handling</a> document. </p>
5654</div>
5655
5656<!-- ======================================================================= -->
5657<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5658 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5659</div>
5660
5661<div class="doc_text">
5662<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5663 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5664 the <tt>nest</tt> attribute, from a function. The result is a callable
5665 function pointer lacking the nest parameter - the caller does not need
5666 to provide a value for it. Instead, the value to use is stored in
5667 advance in a "trampoline", a block of memory usually allocated
5668 on the stack, which also contains code to splice the nest value into the
5669 argument list. This is used to implement the GCC nested function address
5670 extension.
5671</p>
5672<p>
5673 For example, if the function is
5674 <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]5675 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]5676<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5677 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5678 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5679 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5680 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5681</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5682 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5683 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5684</div>
5685
5686<!-- _______________________________________________________________________ -->
5687<div class="doc_subsubsection">
5688 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5689</div>
5690<div class="doc_text">
5691<h5>Syntax:</h5>
5692<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5693declare 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]5694</pre>
5695<h5>Overview:</h5>
5696<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5697 This fills the memory pointed to by <tt>tramp</tt> with code
5698 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5699</p>
5700<h5>Arguments:</h5>
5701<p>
5702 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5703 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5704 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5705 intrinsic. Note that the size and the alignment are target-specific - LLVM
5706 currently provides no portable way of determining them, so a front-end that
5707 generates this intrinsic needs to have some target-specific knowledge.
5708 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]5709</p>
5710<h5>Semantics:</h5>
5711<p>
5712 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5713 dependent code, turning it into a function. A pointer to this function is
5714 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5715 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5716 before being called. The new function's signature is the same as that of
5717 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5718 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5719 of pointer type. Calling the new function is equivalent to calling
5720 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5721 missing <tt>nest</tt> argument. If, after calling
5722 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5723 modified, then the effect of any later call to the returned function pointer is
5724 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5725</p>
5726</div>
5727
5728<!-- ======================================================================= -->
5729<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5730 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5731</div>
5732
5733<div class="doc_text">
5734<p>
5735 These intrinsic functions expand the "universal IR" of LLVM to represent
5736 hardware constructs for atomic operations and memory synchronization. This
5737 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5738 is aimed at a low enough level to allow any programming models or APIs
5739 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5740 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5741 hardware behavior. Just as hardware provides a "universal IR" for source
5742 languages, it also provides a starting point for developing a "universal"
5743 atomic operation and synchronization IR.
5744</p>
5745<p>
5746 These do <em>not</em> form an API such as high-level threading libraries,
5747 software transaction memory systems, atomic primitives, and intrinsic
5748 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5749 application libraries. The hardware interface provided by LLVM should allow
5750 a clean implementation of all of these APIs and parallel programming models.
5751 No one model or paradigm should be selected above others unless the hardware
5752 itself ubiquitously does so.
5753
5754</p>
5755</div>
5756
5757<!-- _______________________________________________________________________ -->
5758<div class="doc_subsubsection">
5759 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5760</div>
5761<div class="doc_text">
5762<h5>Syntax:</h5>
5763<pre>
5764declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5765i1 &lt;device&gt; )
5766
5767</pre>
5768<h5>Overview:</h5>
5769<p>
5770 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5771 specific pairs of memory access types.
5772</p>
5773<h5>Arguments:</h5>
5774<p>
5775 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5776 The first four arguments enables a specific barrier as listed below. The fith
5777 argument specifies that the barrier applies to io or device or uncached memory.
5778
5779</p>
5780 <ul>
5781 <li><tt>ll</tt>: load-load barrier</li>
5782 <li><tt>ls</tt>: load-store barrier</li>
5783 <li><tt>sl</tt>: store-load barrier</li>
5784 <li><tt>ss</tt>: store-store barrier</li>
5785 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5786 </ul>
5787<h5>Semantics:</h5>
5788<p>
5789 This intrinsic causes the system to enforce some ordering constraints upon
5790 the loads and stores of the program. This barrier does not indicate
5791 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5792 which they occur. For any of the specified pairs of load and store operations
5793 (f.ex. load-load, or store-load), all of the first operations preceding the
5794 barrier will complete before any of the second operations succeeding the
5795 barrier begin. Specifically the semantics for each pairing is as follows:
5796</p>
5797 <ul>
5798 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5799 after the barrier begins.</li>
5800
5801 <li><tt>ls</tt>: All loads before the barrier must complete before any
5802 store after the barrier begins.</li>
5803 <li><tt>ss</tt>: All stores before the barrier must complete before any
5804 store after the barrier begins.</li>
5805 <li><tt>sl</tt>: All stores before the barrier must complete before any
5806 load after the barrier begins.</li>
5807 </ul>
5808<p>
5809 These semantics are applied with a logical "and" behavior when more than one
5810 is enabled in a single memory barrier intrinsic.
5811</p>
5812<p>
5813 Backends may implement stronger barriers than those requested when they do not
5814 support as fine grained a barrier as requested. Some architectures do not
5815 need all types of barriers and on such architectures, these become noops.
5816</p>
5817<h5>Example:</h5>
5818<pre>
5819%ptr = malloc i32
5820 store i32 4, %ptr
5821
5822%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5823 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5824 <i>; guarantee the above finishes</i>
5825 store i32 8, %ptr <i>; before this begins</i>
5826</pre>
5827</div>
5828
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5829<!-- _______________________________________________________________________ -->
5830<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5831 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5832</div>
5833<div class="doc_text">
5834<h5>Syntax:</h5>
5835<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5836 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
5837 any integer bit width and for different address spaces. Not all targets
5838 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5839
5840<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5841declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
5842declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
5843declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
5844declare 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]5845
5846</pre>
5847<h5>Overview:</h5>
5848<p>
5849 This loads a value in memory and compares it to a given value. If they are
5850 equal, it stores a new value into the memory.
5851</p>
5852<h5>Arguments:</h5>
5853<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5854 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5855 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5856 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5857 this integer type. While any bit width integer may be used, targets may only
5858 lower representations they support in hardware.
5859
5860</p>
5861<h5>Semantics:</h5>
5862<p>
5863 This entire intrinsic must be executed atomically. It first loads the value
5864 in memory pointed to by <tt>ptr</tt> and compares it with the value
5865 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5866 loaded value is yielded in all cases. This provides the equivalent of an
5867 atomic compare-and-swap operation within the SSA framework.
5868</p>
5869<h5>Examples:</h5>
5870
5871<pre>
5872%ptr = malloc i32
5873 store i32 4, %ptr
5874
5875%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5876%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5877 <i>; yields {i32}:result1 = 4</i>
5878%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5879%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5880
5881%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5882%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5883 <i>; yields {i32}:result2 = 8</i>
5884%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5885
5886%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5887</pre>
5888</div>
5889
5890<!-- _______________________________________________________________________ -->
5891<div class="doc_subsubsection">
5892 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5893</div>
5894<div class="doc_text">
5895<h5>Syntax:</h5>
5896
5897<p>
5898 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5899 integer bit width. Not all targets support all bit widths however.</p>
5900<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5901declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
5902declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
5903declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
5904declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5905
5906</pre>
5907<h5>Overview:</h5>
5908<p>
5909 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
5910 the value from memory. It then stores the value in <tt>val</tt> in the memory
5911 at <tt>ptr</tt>.
5912</p>
5913<h5>Arguments:</h5>
5914
5915<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5916 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5917 <tt>val</tt> argument and the result must be integers of the same bit width.
5918 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
5919 integer type. The targets may only lower integer representations they
5920 support.
5921</p>
5922<h5>Semantics:</h5>
5923<p>
5924 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
5925 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
5926 equivalent of an atomic swap operation within the SSA framework.
5927
5928</p>
5929<h5>Examples:</h5>
5930<pre>
5931%ptr = malloc i32
5932 store i32 4, %ptr
5933
5934%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5935%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5936 <i>; yields {i32}:result1 = 4</i>
5937%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5938%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5939
5940%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5941%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5942 <i>; yields {i32}:result2 = 8</i>
5943
5944%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
5945%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
5946</pre>
5947</div>
5948
5949<!-- _______________________________________________________________________ -->
5950<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5951 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5952
5953</div>
5954<div class="doc_text">
5955<h5>Syntax:</h5>
5956<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5957 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]5958 integer bit width. Not all targets support all bit widths however.</p>
5959<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5960declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
5961declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
5962declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
5963declare 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]5964
5965</pre>
5966<h5>Overview:</h5>
5967<p>
5968 This intrinsic adds <tt>delta</tt> to the value stored in memory at
5969 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
5970</p>
5971<h5>Arguments:</h5>
5972<p>
5973
5974 The intrinsic takes two arguments, the first a pointer to an integer value
5975 and the second an integer value. The result is also an integer value. These
5976 integer types can have any bit width, but they must all have the same bit
5977 width. The targets may only lower integer representations they support.
5978</p>
5979<h5>Semantics:</h5>
5980<p>
5981 This intrinsic does a series of operations atomically. It first loads the
5982 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
5983 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
5984</p>
5985
5986<h5>Examples:</h5>
5987<pre>
5988%ptr = malloc i32
5989 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5990%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5991 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5992%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5993 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5994%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5995 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5996%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5997</pre>
5998</div>
5999
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6000<!-- _______________________________________________________________________ -->
6001<div class="doc_subsubsection">
6002 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6003
6004</div>
6005<div class="doc_text">
6006<h5>Syntax:</h5>
6007<p>
6008 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]6009 any integer bit width and for different address spaces. Not all targets
6010 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6011<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6012declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6013declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6014declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6015declare 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]6016
6017</pre>
6018<h5>Overview:</h5>
6019<p>
6020 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6021 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6022</p>
6023<h5>Arguments:</h5>
6024<p>
6025
6026 The intrinsic takes two arguments, the first a pointer to an integer value
6027 and the second an integer value. The result is also an integer value. These
6028 integer types can have any bit width, but they must all have the same bit
6029 width. The targets may only lower integer representations they support.
6030</p>
6031<h5>Semantics:</h5>
6032<p>
6033 This intrinsic does a series of operations atomically. It first loads the
6034 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6035 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6036</p>
6037
6038<h5>Examples:</h5>
6039<pre>
6040%ptr = malloc i32
6041 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6042%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6043 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6044%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6045 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6046%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6047 <i>; yields {i32}:result3 = 2</i>
6048%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6049</pre>
6050</div>
6051
6052<!-- _______________________________________________________________________ -->
6053<div class="doc_subsubsection">
6054 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6055 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6056 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6057 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6058
6059</div>
6060<div class="doc_text">
6061<h5>Syntax:</h5>
6062<p>
6063 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6064 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6065 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6066 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6067<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6068declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6069declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6070declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6071declare 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]6072
6073</pre>
6074
6075<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6076declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6077declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6078declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6079declare 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]6080
6081</pre>
6082
6083<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6084declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6085declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6086declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6087declare 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]6088
6089</pre>
6090
6091<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6092declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6093declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6094declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6095declare 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]6096
6097</pre>
6098<h5>Overview:</h5>
6099<p>
6100 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6101 the value stored in memory at <tt>ptr</tt>. It yields the original value
6102 at <tt>ptr</tt>.
6103</p>
6104<h5>Arguments:</h5>
6105<p>
6106
6107 These intrinsics take two arguments, the first a pointer to an integer value
6108 and the second an integer value. The result is also an integer value. These
6109 integer types can have any bit width, but they must all have the same bit
6110 width. The targets may only lower integer representations they support.
6111</p>
6112<h5>Semantics:</h5>
6113<p>
6114 These intrinsics does a series of operations atomically. They first load the
6115 value stored at <tt>ptr</tt>. They then do the bitwise operation
6116 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6117 value stored at <tt>ptr</tt>.
6118</p>
6119
6120<h5>Examples:</h5>
6121<pre>
6122%ptr = malloc i32
6123 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6124%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6125 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6126%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6127 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6128%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6129 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6130%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6131 <i>; yields {i32}:result3 = FF</i>
6132%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6133</pre>
6134</div>
6135
6136
6137<!-- _______________________________________________________________________ -->
6138<div class="doc_subsubsection">
6139 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6140 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6141 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6142 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6143
6144</div>
6145<div class="doc_text">
6146<h5>Syntax:</h5>
6147<p>
6148 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6149 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6150 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6151 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6152 support all bit widths however.</p>
6153<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6154declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6155declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6156declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6157declare 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]6158
6159</pre>
6160
6161<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6162declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6163declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6164declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6165declare 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]6166
6167</pre>
6168
6169<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6170declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6171declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6172declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6173declare 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]6174
6175</pre>
6176
6177<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6178declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6179declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6180declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6181declare 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]6182
6183</pre>
6184<h5>Overview:</h5>
6185<p>
6186 These intrinsics takes the signed or unsigned minimum or maximum of
6187 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6188 original value at <tt>ptr</tt>.
6189</p>
6190<h5>Arguments:</h5>
6191<p>
6192
6193 These intrinsics take two arguments, the first a pointer to an integer value
6194 and the second an integer value. The result is also an integer value. These
6195 integer types can have any bit width, but they must all have the same bit
6196 width. The targets may only lower integer representations they support.
6197</p>
6198<h5>Semantics:</h5>
6199<p>
6200 These intrinsics does a series of operations atomically. They first load the
6201 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6202 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6203 the original value stored at <tt>ptr</tt>.
6204</p>
6205
6206<h5>Examples:</h5>
6207<pre>
6208%ptr = malloc i32
6209 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6210%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6211 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6212%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6213 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6214%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6215 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6216%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6217 <i>; yields {i32}:result3 = 8</i>
6218%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6219</pre>
6220</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6221
6222<!-- ======================================================================= -->
6223<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6224 <a name="int_general">General Intrinsics</a>
6225</div>
6226
6227<div class="doc_text">
6228<p> This class of intrinsics is designed to be generic and has
6229no specific purpose. </p>
6230</div>
6231
6232<!-- _______________________________________________________________________ -->
6233<div class="doc_subsubsection">
6234 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6235</div>
6236
6237<div class="doc_text">
6238
6239<h5>Syntax:</h5>
6240<pre>
6241 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6242</pre>
6243
6244<h5>Overview:</h5>
6245
6246<p>
6247The '<tt>llvm.var.annotation</tt>' intrinsic
6248</p>
6249
6250<h5>Arguments:</h5>
6251
6252<p>
6253The first argument is a pointer to a value, the second is a pointer to a
6254global string, the third is a pointer to a global string which is the source
6255file name, and the last argument is the line number.
6256</p>
6257
6258<h5>Semantics:</h5>
6259
6260<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6261This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6262This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6263annotations. These have no other defined use, they are ignored by code
6264generation and optimization.
6265</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6266</div>
6267
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6268<!-- _______________________________________________________________________ -->
6269<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6270 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6271</div>
6272
6273<div class="doc_text">
6274
6275<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6276<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6277any integer bit width.
6278</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6279<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6280 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6281 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6282 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6283 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6284 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]6285</pre>
6286
6287<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6288
6289<p>
6290The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6291</p>
6292
6293<h5>Arguments:</h5>
6294
6295<p>
6296The first argument is an integer value (result of some expression),
6297the second is a pointer to a global string, the third is a pointer to a global
6298string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6299It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6300</p>
6301
6302<h5>Semantics:</h5>
6303
6304<p>
6305This intrinsic allows annotations to be put on arbitrary expressions
6306with arbitrary strings. This can be useful for special purpose optimizations
6307that want to look for these annotations. These have no other defined use, they
6308are ignored by code generation and optimization.
6309</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6310
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6311<!-- _______________________________________________________________________ -->
6312<div class="doc_subsubsection">
6313 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6314</div>
6315
6316<div class="doc_text">
6317
6318<h5>Syntax:</h5>
6319<pre>
6320 declare void @llvm.trap()
6321</pre>
6322
6323<h5>Overview:</h5>
6324
6325<p>
6326The '<tt>llvm.trap</tt>' intrinsic
6327</p>
6328
6329<h5>Arguments:</h5>
6330
6331<p>
6332None
6333</p>
6334
6335<h5>Semantics:</h5>
6336
6337<p>
6338This intrinsics is lowered to the target dependent trap instruction. If the
6339target does not have a trap instruction, this intrinsic will be lowered to the
6340call of the abort() function.
6341</p>
6342</div>
6343
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6344<!-- *********************************************************************** -->
6345<hr>
6346<address>
6347 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6348 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
6349 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6350 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6351
6352 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6353 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6354 Last modified: $Date$
6355</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6356
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6357</body>
6358</html>