Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 0872f0874a94200ee8f7666bf667f8910c595fe0 / . / docs / LangRef.html
blob: b8b2add31e6d30df82fb20cda0df226df2202380 [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
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]855 <dt><tt><a name="byval">byval</a></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
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]861 value, but is also valid on pointers to scalars. The copy is considered to
862 belong to the caller not the callee (for example,
863 <tt><a href="#readonly">readonly</a></tt> functions should not write to
864 <tt>byval</tt> parameters).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]865
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]866 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]867 <dd>This indicates that the pointer parameter specifies the address of a
868 structure that is the return value of the function in the source program.
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]869 This pointer must be guaranteed by the caller to be valid: loads and stores
870 to the structure may be assumed by the callee to not to trap. This may only
871 be applied to the first parameter.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]872
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]873 <dt><tt>noalias</tt></dt>
Owen Andersonc4fc4cd2008-02-18 04:09:01 +0000[diff] [blame]874 <dd>This indicates that the parameter does not alias any global or any other
875 parameter. The caller is responsible for ensuring that this is the case,
876 usually by placing the value in a stack allocation.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]877
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]878 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]879 <dd>This indicates that the pointer parameter can be excised using the
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]880 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]881 </dl>
882
883</div>
884
885<!-- ======================================================================= -->
886<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]887 <a name="gc">Garbage Collector Names</a>
888</div>
889
890<div class="doc_text">
891<p>Each function may specify a garbage collector name, which is simply a
892string.</p>
893
894<div class="doc_code"><pre
895>define void @f() gc "name" { ...</pre></div>
896
897<p>The compiler declares the supported values of <i>name</i>. Specifying a
898collector which will cause the compiler to alter its output in order to support
899the named garbage collection algorithm.</p>
900</div>
901
902<!-- ======================================================================= -->
903<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]904 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]905</div>
906
907<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]908
909<p>Function attributes are set to communicate additional information about
910 a function. Function attributes are considered to be part of the function,
911 not of the function type, so functions with different parameter attributes
912 can have the same function type.</p>
913
914 <p>Function attributes are simple keywords that follow the type specified. If
915 multiple attributes are needed, they are space separated. For
916 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]917
918<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]919<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]920define void @f() noinline { ... }
921define void @f() alwaysinline { ... }
922define void @f() alwaysinline optsize { ... }
923define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]924</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]925</div>
926
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]927<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]928<dt><tt>alwaysinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]929<dd>This attribute indicates that the inliner should attempt to inline this
930function into callers whenever possible, ignoring any active inlining size
931threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]932
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]933<dt><tt>noinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]934<dd>This attribute indicates that the inliner should never inline this function
Chris Lattner377f3ae2008-10-05 17:14:59 +0000[diff] [blame]935in any situation. This attribute may not be used together with the
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]936<tt>alwaysinline</tt> attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]937
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]938<dt><tt>optsize</tt></dt>
Devang Patelc29099b2008-09-29 18:34:44 +0000[diff] [blame]939<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]940make choices that keep the code size of this function low, and otherwise do
941optimizations specifically to reduce code size.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]942
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]943<dt><tt>noreturn</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]944<dd>This function attribute indicates that the function never returns normally.
945This produces undefined behavior at runtime if the function ever does
946dynamically return.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]947
948<dt><tt>nounwind</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]949<dd>This function attribute indicates that the function never returns with an
950unwind or exceptional control flow. If the function does unwind, its runtime
951behavior is undefined.</dd>
952
953<dt><tt>readnone</tt></dt>
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]954<dd>This attribute indicates that the function computes its result (or the
955exception it throws) based strictly on its arguments, without dereferencing any
956pointer arguments or otherwise accessing any mutable state (e.g. memory, control
957registers, etc) visible to caller functions. It does not write through any
958pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
959never changes any state visible to callers.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]960
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]961<dt><tt><a name="readonly">readonly</a></tt></dt>
962<dd>This attribute indicates that the function does not write through any
963pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
964or otherwise modify any state (e.g. memory, control registers, etc) visible to
965caller functions. It may dereference pointer arguments and read state that may
966be set in the caller. A readonly function always returns the same value (or
967throws the same exception) when called with the same set of arguments and global
968state.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]969</dl>
970
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]971</div>
972
973<!-- ======================================================================= -->
974<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]975 <a name="moduleasm">Module-Level Inline Assembly</a>
976</div>
977
978<div class="doc_text">
979<p>
980Modules may contain "module-level inline asm" blocks, which corresponds to the
981GCC "file scope inline asm" blocks. These blocks are internally concatenated by
982LLVM and treated as a single unit, but may be separated in the .ll file if
983desired. The syntax is very simple:
984</p>
985
986<div class="doc_code">
987<pre>
988module asm "inline asm code goes here"
989module asm "more can go here"
990</pre>
991</div>
992
993<p>The strings can contain any character by escaping non-printable characters.
994 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
995 for the number.
996</p>
997
998<p>
999 The inline asm code is simply printed to the machine code .s file when
1000 assembly code is generated.
1001</p>
1002</div>
1003
1004<!-- ======================================================================= -->
1005<div class="doc_subsection">
1006 <a name="datalayout">Data Layout</a>
1007</div>
1008
1009<div class="doc_text">
1010<p>A module may specify a target specific data layout string that specifies how
1011data is to be laid out in memory. The syntax for the data layout is simply:</p>
1012<pre> target datalayout = "<i>layout specification</i>"</pre>
1013<p>The <i>layout specification</i> consists of a list of specifications
1014separated by the minus sign character ('-'). Each specification starts with a
1015letter and may include other information after the letter to define some
1016aspect of the data layout. The specifications accepted are as follows: </p>
1017<dl>
1018 <dt><tt>E</tt></dt>
1019 <dd>Specifies that the target lays out data in big-endian form. That is, the
1020 bits with the most significance have the lowest address location.</dd>
1021 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1022 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1023 the bits with the least significance have the lowest address location.</dd>
1024 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1025 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1026 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1027 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1028 too.</dd>
1029 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1030 <dd>This specifies the alignment for an integer type of a given bit
1031 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1032 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1033 <dd>This specifies the alignment for a vector type of a given bit
1034 <i>size</i>.</dd>
1035 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1036 <dd>This specifies the alignment for a floating point type of a given bit
1037 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1038 (double).</dd>
1039 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1040 <dd>This specifies the alignment for an aggregate type of a given bit
1041 <i>size</i>.</dd>
1042</dl>
1043<p>When constructing the data layout for a given target, LLVM starts with a
1044default set of specifications which are then (possibly) overriden by the
1045specifications in the <tt>datalayout</tt> keyword. The default specifications
1046are given in this list:</p>
1047<ul>
1048 <li><tt>E</tt> - big endian</li>
1049 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1050 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1051 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1052 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1053 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1054 <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]1055 alignment of 64-bits</li>
1056 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1057 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1058 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1059 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1060 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1061</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1062<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1063following rules:
1064<ol>
1065 <li>If the type sought is an exact match for one of the specifications, that
1066 specification is used.</li>
1067 <li>If no match is found, and the type sought is an integer type, then the
1068 smallest integer type that is larger than the bitwidth of the sought type is
1069 used. If none of the specifications are larger than the bitwidth then the the
1070 largest integer type is used. For example, given the default specifications
1071 above, the i7 type will use the alignment of i8 (next largest) while both
1072 i65 and i256 will use the alignment of i64 (largest specified).</li>
1073 <li>If no match is found, and the type sought is a vector type, then the
1074 largest vector type that is smaller than the sought vector type will be used
1075 as a fall back. This happens because <128 x double> can be implemented in
1076 terms of 64 <2 x double>, for example.</li>
1077</ol>
1078</div>
1079
1080<!-- *********************************************************************** -->
1081<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1082<!-- *********************************************************************** -->
1083
1084<div class="doc_text">
1085
1086<p>The LLVM type system is one of the most important features of the
1087intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1088optimizations to be performed on the intermediate representation directly,
1089without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1090extra analyses on the side before the transformation. A strong type
1091system makes it easier to read the generated code and enables novel
1092analyses and transformations that are not feasible to perform on normal
1093three address code representations.</p>
1094
1095</div>
1096
1097<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1098<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1099Classifications</a> </div>
1100<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1101<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1102classifications:</p>
1103
1104<table border="1" cellspacing="0" cellpadding="4">
1105 <tbody>
1106 <tr><th>Classification</th><th>Types</th></tr>
1107 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1108 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1109 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1110 </tr>
1111 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1112 <td><a href="#t_floating">floating point</a></td>
1113 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1114 </tr>
1115 <tr>
1116 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1117 <td><a href="#t_integer">integer</a>,
1118 <a href="#t_floating">floating point</a>,
1119 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1120 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1121 <a href="#t_struct">structure</a>,
1122 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1123 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1124 </td>
1125 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1126 <tr>
1127 <td><a href="#t_primitive">primitive</a></td>
1128 <td><a href="#t_label">label</a>,
1129 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1130 <a href="#t_floating">floating point</a>.</td>
1131 </tr>
1132 <tr>
1133 <td><a href="#t_derived">derived</a></td>
1134 <td><a href="#t_integer">integer</a>,
1135 <a href="#t_array">array</a>,
1136 <a href="#t_function">function</a>,
1137 <a href="#t_pointer">pointer</a>,
1138 <a href="#t_struct">structure</a>,
1139 <a href="#t_pstruct">packed structure</a>,
1140 <a href="#t_vector">vector</a>,
1141 <a href="#t_opaque">opaque</a>.
1142 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1143 </tbody>
1144</table>
1145
1146<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1147most important. Values of these types are the only ones which can be
1148produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1149instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1150</div>
1151
1152<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1153<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1154
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1155<div class="doc_text">
1156<p>The primitive types are the fundamental building blocks of the LLVM
1157system.</p>
1158
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1159</div>
1160
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1161<!-- _______________________________________________________________________ -->
1162<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1163
1164<div class="doc_text">
1165 <table>
1166 <tbody>
1167 <tr><th>Type</th><th>Description</th></tr>
1168 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1169 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1170 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1171 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1172 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1173 </tbody>
1174 </table>
1175</div>
1176
1177<!-- _______________________________________________________________________ -->
1178<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1179
1180<div class="doc_text">
1181<h5>Overview:</h5>
1182<p>The void type does not represent any value and has no size.</p>
1183
1184<h5>Syntax:</h5>
1185
1186<pre>
1187 void
1188</pre>
1189</div>
1190
1191<!-- _______________________________________________________________________ -->
1192<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1193
1194<div class="doc_text">
1195<h5>Overview:</h5>
1196<p>The label type represents code labels.</p>
1197
1198<h5>Syntax:</h5>
1199
1200<pre>
1201 label
1202</pre>
1203</div>
1204
1205
1206<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1207<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1208
1209<div class="doc_text">
1210
1211<p>The real power in LLVM comes from the derived types in the system.
1212This is what allows a programmer to represent arrays, functions,
1213pointers, and other useful types. Note that these derived types may be
1214recursive: For example, it is possible to have a two dimensional array.</p>
1215
1216</div>
1217
1218<!-- _______________________________________________________________________ -->
1219<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1220
1221<div class="doc_text">
1222
1223<h5>Overview:</h5>
1224<p>The integer type is a very simple derived type that simply specifies an
1225arbitrary bit width for the integer type desired. Any bit width from 1 bit to
12262^23-1 (about 8 million) can be specified.</p>
1227
1228<h5>Syntax:</h5>
1229
1230<pre>
1231 iN
1232</pre>
1233
1234<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1235value.</p>
1236
1237<h5>Examples:</h5>
1238<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1239 <tbody>
1240 <tr>
1241 <td><tt>i1</tt></td>
1242 <td>a single-bit integer.</td>
1243 </tr><tr>
1244 <td><tt>i32</tt></td>
1245 <td>a 32-bit integer.</td>
1246 </tr><tr>
1247 <td><tt>i1942652</tt></td>
1248 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1249 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1250 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1251</table>
1252</div>
1253
1254<!-- _______________________________________________________________________ -->
1255<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1256
1257<div class="doc_text">
1258
1259<h5>Overview:</h5>
1260
1261<p>The array type is a very simple derived type that arranges elements
1262sequentially in memory. The array type requires a size (number of
1263elements) and an underlying data type.</p>
1264
1265<h5>Syntax:</h5>
1266
1267<pre>
1268 [&lt;# elements&gt; x &lt;elementtype&gt;]
1269</pre>
1270
1271<p>The number of elements is a constant integer value; elementtype may
1272be any type with a size.</p>
1273
1274<h5>Examples:</h5>
1275<table class="layout">
1276 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1277 <td class="left"><tt>[40 x i32]</tt></td>
1278 <td class="left">Array of 40 32-bit integer values.</td>
1279 </tr>
1280 <tr class="layout">
1281 <td class="left"><tt>[41 x i32]</tt></td>
1282 <td class="left">Array of 41 32-bit integer values.</td>
1283 </tr>
1284 <tr class="layout">
1285 <td class="left"><tt>[4 x i8]</tt></td>
1286 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1287 </tr>
1288</table>
1289<p>Here are some examples of multidimensional arrays:</p>
1290<table class="layout">
1291 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1292 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1293 <td class="left">3x4 array of 32-bit integer values.</td>
1294 </tr>
1295 <tr class="layout">
1296 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1297 <td class="left">12x10 array of single precision floating point values.</td>
1298 </tr>
1299 <tr class="layout">
1300 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1301 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1302 </tr>
1303</table>
1304
1305<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1306length array. Normally, accesses past the end of an array are undefined in
1307LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1308As a special case, however, zero length arrays are recognized to be variable
1309length. This allows implementation of 'pascal style arrays' with the LLVM
1310type "{ i32, [0 x float]}", for example.</p>
1311
1312</div>
1313
1314<!-- _______________________________________________________________________ -->
1315<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1316<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1317
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1318<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1319
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1320<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1321consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1322return 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]1323If 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]1324class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1325
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1326<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1327
1328<pre>
1329 &lt;returntype list&gt; (&lt;parameter list&gt;)
1330</pre>
1331
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1332<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1333specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1334which indicates that the function takes a variable number of arguments.
1335Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1336 href="#int_varargs">variable argument handling intrinsic</a> functions.
1337'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1338<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1339
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1340<h5>Examples:</h5>
1341<table class="layout">
1342 <tr class="layout">
1343 <td class="left"><tt>i32 (i32)</tt></td>
1344 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1345 </td>
1346 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1347 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1348 </tt></td>
1349 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1350 an <tt>i16</tt> that should be sign extended and a
1351 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1352 <tt>float</tt>.
1353 </td>
1354 </tr><tr class="layout">
1355 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1356 <td class="left">A vararg function that takes at least one
1357 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1358 which returns an integer. This is the signature for <tt>printf</tt> in
1359 LLVM.
1360 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1361 </tr><tr class="layout">
1362 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Devang Patel6dddba22008-03-24 18:10:52 +0000[diff] [blame]1363 <td class="left">A function taking an <tt>i32></tt>, returning two
1364 <tt> i32 </tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1365 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1366 </tr>
1367</table>
1368
1369</div>
1370<!-- _______________________________________________________________________ -->
1371<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1372<div class="doc_text">
1373<h5>Overview:</h5>
1374<p>The structure type is used to represent a collection of data members
1375together in memory. The packing of the field types is defined to match
1376the ABI of the underlying processor. The elements of a structure may
1377be any type that has a size.</p>
1378<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1379and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1380field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1381instruction.</p>
1382<h5>Syntax:</h5>
1383<pre> { &lt;type list&gt; }<br></pre>
1384<h5>Examples:</h5>
1385<table class="layout">
1386 <tr class="layout">
1387 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1388 <td class="left">A triple of three <tt>i32</tt> values</td>
1389 </tr><tr class="layout">
1390 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1391 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1392 second element is a <a href="#t_pointer">pointer</a> to a
1393 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1394 an <tt>i32</tt>.</td>
1395 </tr>
1396</table>
1397</div>
1398
1399<!-- _______________________________________________________________________ -->
1400<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1401</div>
1402<div class="doc_text">
1403<h5>Overview:</h5>
1404<p>The packed structure type is used to represent a collection of data members
1405together in memory. There is no padding between fields. Further, the alignment
1406of a packed structure is 1 byte. The elements of a packed structure may
1407be any type that has a size.</p>
1408<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1409and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1410field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1411instruction.</p>
1412<h5>Syntax:</h5>
1413<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1414<h5>Examples:</h5>
1415<table class="layout">
1416 <tr class="layout">
1417 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1418 <td class="left">A triple of three <tt>i32</tt> values</td>
1419 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1420 <td class="left">
1421<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1422 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1423 second element is a <a href="#t_pointer">pointer</a> to a
1424 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1425 an <tt>i32</tt>.</td>
1426 </tr>
1427</table>
1428</div>
1429
1430<!-- _______________________________________________________________________ -->
1431<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1432<div class="doc_text">
1433<h5>Overview:</h5>
1434<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1435reference to another object, which must live in memory. Pointer types may have
1436an optional address space attribute defining the target-specific numbered
1437address space where the pointed-to object resides. The default address space is
1438zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1439<h5>Syntax:</h5>
1440<pre> &lt;type&gt; *<br></pre>
1441<h5>Examples:</h5>
1442<table class="layout">
1443 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1444 <td class="left"><tt>[4x i32]*</tt></td>
1445 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1446 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1447 </tr>
1448 <tr class="layout">
1449 <td class="left"><tt>i32 (i32 *) *</tt></td>
1450 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1451 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1452 <tt>i32</tt>.</td>
1453 </tr>
1454 <tr class="layout">
1455 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1456 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1457 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1458 </tr>
1459</table>
1460</div>
1461
1462<!-- _______________________________________________________________________ -->
1463<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1464<div class="doc_text">
1465
1466<h5>Overview:</h5>
1467
1468<p>A vector type is a simple derived type that represents a vector
1469of elements. Vector types are used when multiple primitive data
1470are operated in parallel using a single instruction (SIMD).
1471A vector type requires a size (number of
1472elements) and an underlying primitive data type. Vectors must have a power
1473of two length (1, 2, 4, 8, 16 ...). Vector types are
1474considered <a href="#t_firstclass">first class</a>.</p>
1475
1476<h5>Syntax:</h5>
1477
1478<pre>
1479 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1480</pre>
1481
1482<p>The number of elements is a constant integer value; elementtype may
1483be any integer or floating point type.</p>
1484
1485<h5>Examples:</h5>
1486
1487<table class="layout">
1488 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1489 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1490 <td class="left">Vector of 4 32-bit integer values.</td>
1491 </tr>
1492 <tr class="layout">
1493 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1494 <td class="left">Vector of 8 32-bit floating-point values.</td>
1495 </tr>
1496 <tr class="layout">
1497 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1498 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1499 </tr>
1500</table>
1501</div>
1502
1503<!-- _______________________________________________________________________ -->
1504<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1505<div class="doc_text">
1506
1507<h5>Overview:</h5>
1508
1509<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1510corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1511In LLVM, opaque types can eventually be resolved to any type (not just a
1512structure type).</p>
1513
1514<h5>Syntax:</h5>
1515
1516<pre>
1517 opaque
1518</pre>
1519
1520<h5>Examples:</h5>
1521
1522<table class="layout">
1523 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1524 <td class="left"><tt>opaque</tt></td>
1525 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1526 </tr>
1527</table>
1528</div>
1529
1530
1531<!-- *********************************************************************** -->
1532<div class="doc_section"> <a name="constants">Constants</a> </div>
1533<!-- *********************************************************************** -->
1534
1535<div class="doc_text">
1536
1537<p>LLVM has several different basic types of constants. This section describes
1538them all and their syntax.</p>
1539
1540</div>
1541
1542<!-- ======================================================================= -->
1543<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1544
1545<div class="doc_text">
1546
1547<dl>
1548 <dt><b>Boolean constants</b></dt>
1549
1550 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1551 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1552 </dd>
1553
1554 <dt><b>Integer constants</b></dt>
1555
1556 <dd>Standard integers (such as '4') are constants of the <a
1557 href="#t_integer">integer</a> type. Negative numbers may be used with
1558 integer types.
1559 </dd>
1560
1561 <dt><b>Floating point constants</b></dt>
1562
1563 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1564 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1565 notation (see below). The assembler requires the exact decimal value of
1566 a floating-point constant. For example, the assembler accepts 1.25 but
1567 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1568 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1569
1570 <dt><b>Null pointer constants</b></dt>
1571
1572 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1573 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1574
1575</dl>
1576
1577<p>The one non-intuitive notation for constants is the optional hexadecimal form
1578of floating point constants. For example, the form '<tt>double
15790x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
15804.5e+15</tt>'. The only time hexadecimal floating point constants are required
1581(and the only time that they are generated by the disassembler) is when a
1582floating point constant must be emitted but it cannot be represented as a
1583decimal floating point number. For example, NaN's, infinities, and other
1584special values are represented in their IEEE hexadecimal format so that
1585assembly and disassembly do not cause any bits to change in the constants.</p>
1586
1587</div>
1588
1589<!-- ======================================================================= -->
1590<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1591</div>
1592
1593<div class="doc_text">
1594<p>Aggregate constants arise from aggregation of simple constants
1595and smaller aggregate constants.</p>
1596
1597<dl>
1598 <dt><b>Structure constants</b></dt>
1599
1600 <dd>Structure constants are represented with notation similar to structure
1601 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1602 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1603 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]1604 must have <a href="#t_struct">structure type</a>, and the number and
1605 types of elements must match those specified by the type.
1606 </dd>
1607
1608 <dt><b>Array constants</b></dt>
1609
1610 <dd>Array constants are represented with notation similar to array type
1611 definitions (a comma separated list of elements, surrounded by square brackets
1612 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1613 constants must have <a href="#t_array">array type</a>, and the number and
1614 types of elements must match those specified by the type.
1615 </dd>
1616
1617 <dt><b>Vector constants</b></dt>
1618
1619 <dd>Vector constants are represented with notation similar to vector type
1620 definitions (a comma separated list of elements, surrounded by
1621 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1622 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1623 href="#t_vector">vector type</a>, and the number and types of elements must
1624 match those specified by the type.
1625 </dd>
1626
1627 <dt><b>Zero initialization</b></dt>
1628
1629 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1630 value to zero of <em>any</em> type, including scalar and aggregate types.
1631 This is often used to avoid having to print large zero initializers (e.g. for
1632 large arrays) and is always exactly equivalent to using explicit zero
1633 initializers.
1634 </dd>
1635</dl>
1636
1637</div>
1638
1639<!-- ======================================================================= -->
1640<div class="doc_subsection">
1641 <a name="globalconstants">Global Variable and Function Addresses</a>
1642</div>
1643
1644<div class="doc_text">
1645
1646<p>The addresses of <a href="#globalvars">global variables</a> and <a
1647href="#functionstructure">functions</a> are always implicitly valid (link-time)
1648constants. These constants are explicitly referenced when the <a
1649href="#identifiers">identifier for the global</a> is used and always have <a
1650href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1651file:</p>
1652
1653<div class="doc_code">
1654<pre>
1655@X = global i32 17
1656@Y = global i32 42
1657@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1658</pre>
1659</div>
1660
1661</div>
1662
1663<!-- ======================================================================= -->
1664<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1665<div class="doc_text">
1666 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1667 no specific value. Undefined values may be of any type and be used anywhere
1668 a constant is permitted.</p>
1669
1670 <p>Undefined values indicate to the compiler that the program is well defined
1671 no matter what value is used, giving the compiler more freedom to optimize.
1672 </p>
1673</div>
1674
1675<!-- ======================================================================= -->
1676<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1677</div>
1678
1679<div class="doc_text">
1680
1681<p>Constant expressions are used to allow expressions involving other constants
1682to be used as constants. Constant expressions may be of any <a
1683href="#t_firstclass">first class</a> type and may involve any LLVM operation
1684that does not have side effects (e.g. load and call are not supported). The
1685following is the syntax for constant expressions:</p>
1686
1687<dl>
1688 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1689 <dd>Truncate a constant to another type. The bit size of CST must be larger
1690 than the bit size of TYPE. Both types must be integers.</dd>
1691
1692 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1693 <dd>Zero extend a constant to another type. The bit size of CST must be
1694 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1695
1696 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1697 <dd>Sign extend a constant to another type. The bit size of CST must be
1698 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1699
1700 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1701 <dd>Truncate a floating point constant to another floating point type. The
1702 size of CST must be larger than the size of TYPE. Both types must be
1703 floating point.</dd>
1704
1705 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1706 <dd>Floating point extend a constant to another type. The size of CST must be
1707 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1708
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1709 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1710 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1711 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1712 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1713 of the same number of elements. If the value won't fit in the integer type,
1714 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1715
1716 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1717 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1718 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1719 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1720 of the same number of elements. If the value won't fit in the integer type,
1721 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1722
1723 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1724 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1725 constant. TYPE must be a scalar or vector floating point type. CST must be of
1726 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1727 of the same number of elements. If the value won't fit in the floating point
1728 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1729
1730 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1731 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1732 constant. TYPE must be a scalar or vector floating point type. CST must be of
1733 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1734 of the same number of elements. If the value won't fit in the floating point
1735 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1736
1737 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1738 <dd>Convert a pointer typed constant to the corresponding integer constant
1739 TYPE must be an integer type. CST must be of pointer type. The CST value is
1740 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1741
1742 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1743 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1744 pointer type. CST must be of integer type. The CST value is zero extended,
1745 truncated, or unchanged to make it fit in a pointer size. This one is
1746 <i>really</i> dangerous!</dd>
1747
1748 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1749 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1750 identical (same number of bits). The conversion is done as if the CST value
1751 was stored to memory and read back as TYPE. In other words, no bits change
1752 with this operator, just the type. This can be used for conversion of
1753 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]1754 pointers it is only valid to cast to another pointer type. It is not valid
1755 to bitcast to or from an aggregate type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1756 </dd>
1757
1758 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1759
1760 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1761 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1762 instruction, the index list may have zero or more indexes, which are required
1763 to make sense for the type of "CSTPTR".</dd>
1764
1765 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1766
1767 <dd>Perform the <a href="#i_select">select operation</a> on
1768 constants.</dd>
1769
1770 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1771 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1772
1773 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1774 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1775
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1776 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1777 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1778
1779 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1780 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1781
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1782 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1783
1784 <dd>Perform the <a href="#i_extractelement">extractelement
1785 operation</a> on constants.
1786
1787 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1788
1789 <dd>Perform the <a href="#i_insertelement">insertelement
1790 operation</a> on constants.</dd>
1791
1792
1793 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1794
1795 <dd>Perform the <a href="#i_shufflevector">shufflevector
1796 operation</a> on constants.</dd>
1797
1798 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1799
1800 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1801 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1802 binary</a> operations. The constraints on operands are the same as those for
1803 the corresponding instruction (e.g. no bitwise operations on floating point
1804 values are allowed).</dd>
1805</dl>
1806</div>
1807
1808<!-- *********************************************************************** -->
1809<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1810<!-- *********************************************************************** -->
1811
1812<!-- ======================================================================= -->
1813<div class="doc_subsection">
1814<a name="inlineasm">Inline Assembler Expressions</a>
1815</div>
1816
1817<div class="doc_text">
1818
1819<p>
1820LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1821Module-Level Inline Assembly</a>) through the use of a special value. This
1822value represents the inline assembler as a string (containing the instructions
1823to emit), a list of operand constraints (stored as a string), and a flag that
1824indicates whether or not the inline asm expression has side effects. An example
1825inline assembler expression is:
1826</p>
1827
1828<div class="doc_code">
1829<pre>
1830i32 (i32) asm "bswap $0", "=r,r"
1831</pre>
1832</div>
1833
1834<p>
1835Inline assembler expressions may <b>only</b> be used as the callee operand of
1836a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1837</p>
1838
1839<div class="doc_code">
1840<pre>
1841%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1842</pre>
1843</div>
1844
1845<p>
1846Inline asms with side effects not visible in the constraint list must be marked
1847as having side effects. This is done through the use of the
1848'<tt>sideeffect</tt>' keyword, like so:
1849</p>
1850
1851<div class="doc_code">
1852<pre>
1853call void asm sideeffect "eieio", ""()
1854</pre>
1855</div>
1856
1857<p>TODO: The format of the asm and constraints string still need to be
1858documented here. Constraints on what can be done (e.g. duplication, moving, etc
Chris Lattner1cdd64e2008-10-04 18:36:02 +0000[diff] [blame]1859need to be documented). This is probably best done by reference to another
1860document that covers inline asm from a holistic perspective.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1861</p>
1862
1863</div>
1864
1865<!-- *********************************************************************** -->
1866<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1867<!-- *********************************************************************** -->
1868
1869<div class="doc_text">
1870
1871<p>The LLVM instruction set consists of several different
1872classifications of instructions: <a href="#terminators">terminator
1873instructions</a>, <a href="#binaryops">binary instructions</a>,
1874<a href="#bitwiseops">bitwise binary instructions</a>, <a
1875 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1876instructions</a>.</p>
1877
1878</div>
1879
1880<!-- ======================================================================= -->
1881<div class="doc_subsection"> <a name="terminators">Terminator
1882Instructions</a> </div>
1883
1884<div class="doc_text">
1885
1886<p>As mentioned <a href="#functionstructure">previously</a>, every
1887basic block in a program ends with a "Terminator" instruction, which
1888indicates which block should be executed after the current block is
1889finished. These terminator instructions typically yield a '<tt>void</tt>'
1890value: they produce control flow, not values (the one exception being
1891the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1892<p>There are six different terminator instructions: the '<a
1893 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1894instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1895the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1896 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1897 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1898
1899</div>
1900
1901<!-- _______________________________________________________________________ -->
1902<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1903Instruction</a> </div>
1904<div class="doc_text">
1905<h5>Syntax:</h5>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1906<pre>
1907 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]1908 ret void <i>; Return from void function</i>
1909</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1910
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1911<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1912
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1913<p>The '<tt>ret</tt>' instruction is used to return control flow (and
1914optionally a value) from a function back to the caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1915<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1916returns a value and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1917control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1918
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1919<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1920
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1921<p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
1922the return value. The type of the return value must be a
1923'<a href="#t_firstclass">first class</a>' type.</p>
1924
1925<p>A function is not <a href="#wellformed">well formed</a> if
1926it it has a non-void return type and contains a '<tt>ret</tt>'
1927instruction with no return value or a return value with a type that
1928does not match its type, or if it has a void return type and contains
1929a '<tt>ret</tt>' instruction with a return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1930
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1931<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1932
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1933<p>When the '<tt>ret</tt>' instruction is executed, control flow
1934returns back to the calling function's context. If the caller is a "<a
1935 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1936the instruction after the call. If the caller was an "<a
1937 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1938at the beginning of the "normal" destination block. If the instruction
1939returns a value, that value shall set the call or invoke instruction's
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1940return value.
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1941
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1942<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1943
1944<pre>
1945 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1946 ret void <i>; Return from a void function</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]1947 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]1948</pre>
1949</div>
1950<!-- _______________________________________________________________________ -->
1951<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1952<div class="doc_text">
1953<h5>Syntax:</h5>
1954<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1955</pre>
1956<h5>Overview:</h5>
1957<p>The '<tt>br</tt>' instruction is used to cause control flow to
1958transfer to a different basic block in the current function. There are
1959two forms of this instruction, corresponding to a conditional branch
1960and an unconditional branch.</p>
1961<h5>Arguments:</h5>
1962<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1963single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1964unconditional form of the '<tt>br</tt>' instruction takes a single
1965'<tt>label</tt>' value as a target.</p>
1966<h5>Semantics:</h5>
1967<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1968argument is evaluated. If the value is <tt>true</tt>, control flows
1969to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1970control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1971<h5>Example:</h5>
1972<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
1973 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1974</div>
1975<!-- _______________________________________________________________________ -->
1976<div class="doc_subsubsection">
1977 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1978</div>
1979
1980<div class="doc_text">
1981<h5>Syntax:</h5>
1982
1983<pre>
1984 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1985</pre>
1986
1987<h5>Overview:</h5>
1988
1989<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1990several different places. It is a generalization of the '<tt>br</tt>'
1991instruction, allowing a branch to occur to one of many possible
1992destinations.</p>
1993
1994
1995<h5>Arguments:</h5>
1996
1997<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1998comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1999an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2000table is not allowed to contain duplicate constant entries.</p>
2001
2002<h5>Semantics:</h5>
2003
2004<p>The <tt>switch</tt> instruction specifies a table of values and
2005destinations. When the '<tt>switch</tt>' instruction is executed, this
2006table is searched for the given value. If the value is found, control flow is
2007transfered to the corresponding destination; otherwise, control flow is
2008transfered to the default destination.</p>
2009
2010<h5>Implementation:</h5>
2011
2012<p>Depending on properties of the target machine and the particular
2013<tt>switch</tt> instruction, this instruction may be code generated in different
2014ways. For example, it could be generated as a series of chained conditional
2015branches or with a lookup table.</p>
2016
2017<h5>Example:</h5>
2018
2019<pre>
2020 <i>; Emulate a conditional br instruction</i>
2021 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2022 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
2023
2024 <i>; Emulate an unconditional br instruction</i>
2025 switch i32 0, label %dest [ ]
2026
2027 <i>; Implement a jump table:</i>
2028 switch i32 %val, label %otherwise [ i32 0, label %onzero
2029 i32 1, label %onone
2030 i32 2, label %ontwo ]
2031</pre>
2032</div>
2033
2034<!-- _______________________________________________________________________ -->
2035<div class="doc_subsubsection">
2036 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2037</div>
2038
2039<div class="doc_text">
2040
2041<h5>Syntax:</h5>
2042
2043<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2044 &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]2045 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2046</pre>
2047
2048<h5>Overview:</h5>
2049
2050<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2051function, with the possibility of control flow transfer to either the
2052'<tt>normal</tt>' label or the
2053'<tt>exception</tt>' label. If the callee function returns with the
2054"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2055"normal" label. If the callee (or any indirect callees) returns with the "<a
2056href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2057continued at the dynamically nearest "exception" label.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2058
2059<h5>Arguments:</h5>
2060
2061<p>This instruction requires several arguments:</p>
2062
2063<ol>
2064 <li>
2065 The optional "cconv" marker indicates which <a href="#callingconv">calling
2066 convention</a> the call should use. If none is specified, the call defaults
2067 to using C calling conventions.
2068 </li>
2069 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2070 function value being invoked. In most cases, this is a direct function
2071 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2072 an arbitrary pointer to function value.
2073 </li>
2074
2075 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2076 function to be invoked. </li>
2077
2078 <li>'<tt>function args</tt>': argument list whose types match the function
2079 signature argument types. If the function signature indicates the function
2080 accepts a variable number of arguments, the extra arguments can be
2081 specified. </li>
2082
2083 <li>'<tt>normal label</tt>': the label reached when the called function
2084 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2085
2086 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2087 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2088
2089</ol>
2090
2091<h5>Semantics:</h5>
2092
2093<p>This instruction is designed to operate as a standard '<tt><a
2094href="#i_call">call</a></tt>' instruction in most regards. The primary
2095difference is that it establishes an association with a label, which is used by
2096the runtime library to unwind the stack.</p>
2097
2098<p>This instruction is used in languages with destructors to ensure that proper
2099cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2100exception. Additionally, this is important for implementation of
2101'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2102
2103<h5>Example:</h5>
2104<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2105 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2106 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2107 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2108 unwind label %TestCleanup <i>; {i32}:retval set</i>
2109</pre>
2110</div>
2111
2112
2113<!-- _______________________________________________________________________ -->
2114
2115<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2116Instruction</a> </div>
2117
2118<div class="doc_text">
2119
2120<h5>Syntax:</h5>
2121<pre>
2122 unwind
2123</pre>
2124
2125<h5>Overview:</h5>
2126
2127<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2128at the first callee in the dynamic call stack which used an <a
2129href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2130primarily used to implement exception handling.</p>
2131
2132<h5>Semantics:</h5>
2133
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2134<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2135immediately halt. The dynamic call stack is then searched for the first <a
2136href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2137execution continues at the "exceptional" destination block specified by the
2138<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2139dynamic call chain, undefined behavior results.</p>
2140</div>
2141
2142<!-- _______________________________________________________________________ -->
2143
2144<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2145Instruction</a> </div>
2146
2147<div class="doc_text">
2148
2149<h5>Syntax:</h5>
2150<pre>
2151 unreachable
2152</pre>
2153
2154<h5>Overview:</h5>
2155
2156<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2157instruction is used to inform the optimizer that a particular portion of the
2158code is not reachable. This can be used to indicate that the code after a
2159no-return function cannot be reached, and other facts.</p>
2160
2161<h5>Semantics:</h5>
2162
2163<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2164</div>
2165
2166
2167
2168<!-- ======================================================================= -->
2169<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2170<div class="doc_text">
2171<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2172program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2173produce a single value. The operands might represent
2174multiple 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]2175The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2176<p>There are several different binary operators:</p>
2177</div>
2178<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2179<div class="doc_subsubsection">
2180 <a name="i_add">'<tt>add</tt>' Instruction</a>
2181</div>
2182
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2183<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2184
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2185<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2186
2187<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2188 &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]2189</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2190
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2191<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2192
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2193<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2194
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2195<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2196
2197<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2198 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2199 <a href="#t_vector">vector</a> values. Both arguments must have identical
2200 types.</p>
2201
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2202<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2203
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2204<p>The value produced is the integer or floating point sum of the two
2205operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2206
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2207<p>If an integer sum has unsigned overflow, the result returned is the
2208mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2209the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2210
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2211<p>Because LLVM integers use a two's complement representation, this
2212instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2213
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2214<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2215
2216<pre>
2217 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2218</pre>
2219</div>
2220<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2221<div class="doc_subsubsection">
2222 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2223</div>
2224
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2225<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2226
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2227<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2228
2229<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2230 &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]2231</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2232
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2233<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2234
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2235<p>The '<tt>sub</tt>' instruction returns the difference of its two
2236operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2237
2238<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2239'<tt>neg</tt>' instruction present in most other intermediate
2240representations.</p>
2241
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2242<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2243
2244<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2245 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2246 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2247 types.</p>
2248
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2249<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2250
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2251<p>The value produced is the integer or floating point difference of
2252the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2253
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2254<p>If an integer difference has unsigned overflow, the result returned is the
2255mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2256the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2257
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2258<p>Because LLVM integers use a two's complement representation, this
2259instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2260
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2261<h5>Example:</h5>
2262<pre>
2263 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2264 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2265</pre>
2266</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2267
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2268<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2269<div class="doc_subsubsection">
2270 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2271</div>
2272
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2273<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2274
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2275<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2276<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]2277</pre>
2278<h5>Overview:</h5>
2279<p>The '<tt>mul</tt>' instruction returns the product of its two
2280operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2281
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2282<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2283
2284<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2285href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2286or <a href="#t_vector">vector</a> values. Both arguments must have identical
2287types.</p>
2288
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2289<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2290
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2291<p>The value produced is the integer or floating point product of the
2292two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2293
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2294<p>If the result of an integer multiplication has unsigned overflow,
2295the result returned is the mathematical result modulo
22962<sup>n</sup>, where n is the bit width of the result.</p>
2297<p>Because LLVM integers use a two's complement representation, and the
2298result is the same width as the operands, this instruction returns the
2299correct result for both signed and unsigned integers. If a full product
2300(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2301should be sign-extended or zero-extended as appropriate to the
2302width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2303<h5>Example:</h5>
2304<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2305</pre>
2306</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2307
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2308<!-- _______________________________________________________________________ -->
2309<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2310</a></div>
2311<div class="doc_text">
2312<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2313<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]2314</pre>
2315<h5>Overview:</h5>
2316<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2317operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2318
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2319<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2320
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2321<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2322<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2323values. Both arguments must have identical types.</p>
2324
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2325<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2326
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2327<p>The value produced is the unsigned integer quotient of the two operands.</p>
2328<p>Note that unsigned integer division and signed integer division are distinct
2329operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2330<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2331<h5>Example:</h5>
2332<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2333</pre>
2334</div>
2335<!-- _______________________________________________________________________ -->
2336<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2337</a> </div>
2338<div class="doc_text">
2339<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2340<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2341 &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]2342</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2343
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2344<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2345
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2346<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2347operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2348
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2349<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2350
2351<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2352<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2353values. Both arguments must have identical types.</p>
2354
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2355<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2356<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]2357<p>Note that signed integer division and unsigned integer division are distinct
2358operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2359<p>Division by zero leads to undefined behavior. Overflow also leads to
2360undefined behavior; this is a rare case, but can occur, for example,
2361by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2362<h5>Example:</h5>
2363<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2364</pre>
2365</div>
2366<!-- _______________________________________________________________________ -->
2367<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2368Instruction</a> </div>
2369<div class="doc_text">
2370<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2371<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2372 &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]2373</pre>
2374<h5>Overview:</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 '<tt>fdiv</tt>' instruction returns the quotient of its two
2377operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2378
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2379<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2380
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2381<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2382<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2383of floating point values. Both arguments must have identical types.</p>
2384
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2385<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2386
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2387<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2388
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2389<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2390
2391<pre>
2392 &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]2393</pre>
2394</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2395
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2396<!-- _______________________________________________________________________ -->
2397<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2398</div>
2399<div class="doc_text">
2400<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2401<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]2402</pre>
2403<h5>Overview:</h5>
2404<p>The '<tt>urem</tt>' instruction returns the remainder from the
2405unsigned division of its two arguments.</p>
2406<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2407<p>The two arguments to the '<tt>urem</tt>' instruction must be
2408<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2409values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2410<h5>Semantics:</h5>
2411<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2412This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2413<p>Note that unsigned integer remainder and signed integer remainder are
2414distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2415<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2416<h5>Example:</h5>
2417<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2418</pre>
2419
2420</div>
2421<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2422<div class="doc_subsubsection">
2423 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2424</div>
2425
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2426<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2427
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2428<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2429
2430<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2431 &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]2432</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2433
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2434<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2435
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2436<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2437signed division of its two operands. This instruction can also take
2438<a href="#t_vector">vector</a> versions of the values in which case
2439the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2440
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2441<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2442
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2443<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2444<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2445values. Both arguments must have identical types.</p>
2446
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2447<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2448
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2449<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2450has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2451operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2452a value. For more information about the difference, see <a
2453 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2454Math Forum</a>. For a table of how this is implemented in various languages,
2455please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2456Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2457<p>Note that signed integer remainder and unsigned integer remainder are
2458distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2459<p>Taking the remainder of a division by zero leads to undefined behavior.
2460Overflow also leads to undefined behavior; this is a rare case, but can occur,
2461for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2462(The remainder doesn't actually overflow, but this rule lets srem be
2463implemented using instructions that return both the result of the division
2464and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2465<h5>Example:</h5>
2466<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2467</pre>
2468
2469</div>
2470<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2471<div class="doc_subsubsection">
2472 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2473
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2474<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2475
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2476<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2477<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]2478</pre>
2479<h5>Overview:</h5>
2480<p>The '<tt>frem</tt>' instruction returns the remainder from the
2481division of its two operands.</p>
2482<h5>Arguments:</h5>
2483<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2484<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2485of floating point values. Both arguments must have identical types.</p>
2486
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2487<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2488
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2489<p>This instruction returns the <i>remainder</i> of a division.
2490The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2491
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2492<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2493
2494<pre>
2495 &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]2496</pre>
2497</div>
2498
2499<!-- ======================================================================= -->
2500<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2501Operations</a> </div>
2502<div class="doc_text">
2503<p>Bitwise binary operators are used to do various forms of
2504bit-twiddling in a program. They are generally very efficient
2505instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2506instructions. They require two operands of the same type, execute an operation on them,
2507and 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]2508</div>
2509
2510<!-- _______________________________________________________________________ -->
2511<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2512Instruction</a> </div>
2513<div class="doc_text">
2514<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2515<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]2516</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2517
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2518<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2519
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2520<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2521the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2522
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2523<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2524
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2525<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2526 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2527type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2528
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2529<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2530
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2531<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2532where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2533equal 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]2534
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2535<h5>Example:</h5><pre>
2536 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2537 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2538 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2539 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2540</pre>
2541</div>
2542<!-- _______________________________________________________________________ -->
2543<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2544Instruction</a> </div>
2545<div class="doc_text">
2546<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2547<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]2548</pre>
2549
2550<h5>Overview:</h5>
2551<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2552operand shifted to the right a specified number of bits with zero fill.</p>
2553
2554<h5>Arguments:</h5>
2555<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2556<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]2557type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2558
2559<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2560
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2561<p>This instruction always performs a logical shift right operation. The most
2562significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2563shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2564the number of bits in <tt>op1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2565
2566<h5>Example:</h5>
2567<pre>
2568 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2569 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2570 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2571 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2572 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2573</pre>
2574</div>
2575
2576<!-- _______________________________________________________________________ -->
2577<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2578Instruction</a> </div>
2579<div class="doc_text">
2580
2581<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2582<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]2583</pre>
2584
2585<h5>Overview:</h5>
2586<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2587operand shifted to the right a specified number of bits with sign extension.</p>
2588
2589<h5>Arguments:</h5>
2590<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2591<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]2592type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2593
2594<h5>Semantics:</h5>
2595<p>This instruction always performs an arithmetic shift right operation,
2596The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2597of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2598larger than the number of bits in <tt>op1</tt>, the result is undefined.
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2599</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2600
2601<h5>Example:</h5>
2602<pre>
2603 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2604 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2605 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2606 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2607 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2608</pre>
2609</div>
2610
2611<!-- _______________________________________________________________________ -->
2612<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2613Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2614
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2615<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2616
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2617<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2618
2619<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2620 &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]2621</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2622
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2623<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2624
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2625<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2626its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2627
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2628<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2629
2630<p>The two arguments to the '<tt>and</tt>' instruction must be
2631<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2632values. Both arguments must have identical types.</p>
2633
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2634<h5>Semantics:</h5>
2635<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2636<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2637<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2638<table border="1" cellspacing="0" cellpadding="4">
2639 <tbody>
2640 <tr>
2641 <td>In0</td>
2642 <td>In1</td>
2643 <td>Out</td>
2644 </tr>
2645 <tr>
2646 <td>0</td>
2647 <td>0</td>
2648 <td>0</td>
2649 </tr>
2650 <tr>
2651 <td>0</td>
2652 <td>1</td>
2653 <td>0</td>
2654 </tr>
2655 <tr>
2656 <td>1</td>
2657 <td>0</td>
2658 <td>0</td>
2659 </tr>
2660 <tr>
2661 <td>1</td>
2662 <td>1</td>
2663 <td>1</td>
2664 </tr>
2665 </tbody>
2666</table>
2667</div>
2668<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2669<pre>
2670 &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]2671 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2672 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2673</pre>
2674</div>
2675<!-- _______________________________________________________________________ -->
2676<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2677<div class="doc_text">
2678<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2679<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]2680</pre>
2681<h5>Overview:</h5>
2682<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2683or of its two operands.</p>
2684<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2685
2686<p>The two arguments to the '<tt>or</tt>' instruction must be
2687<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2688values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2689<h5>Semantics:</h5>
2690<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2691<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2692<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2693<table border="1" cellspacing="0" cellpadding="4">
2694 <tbody>
2695 <tr>
2696 <td>In0</td>
2697 <td>In1</td>
2698 <td>Out</td>
2699 </tr>
2700 <tr>
2701 <td>0</td>
2702 <td>0</td>
2703 <td>0</td>
2704 </tr>
2705 <tr>
2706 <td>0</td>
2707 <td>1</td>
2708 <td>1</td>
2709 </tr>
2710 <tr>
2711 <td>1</td>
2712 <td>0</td>
2713 <td>1</td>
2714 </tr>
2715 <tr>
2716 <td>1</td>
2717 <td>1</td>
2718 <td>1</td>
2719 </tr>
2720 </tbody>
2721</table>
2722</div>
2723<h5>Example:</h5>
2724<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2725 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2726 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2727</pre>
2728</div>
2729<!-- _______________________________________________________________________ -->
2730<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2731Instruction</a> </div>
2732<div class="doc_text">
2733<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2734<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]2735</pre>
2736<h5>Overview:</h5>
2737<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2738or of its two operands. The <tt>xor</tt> is used to implement the
2739"one's complement" operation, which is the "~" operator in C.</p>
2740<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2741<p>The two arguments to the '<tt>xor</tt>' instruction must be
2742<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2743values. Both arguments must have identical types.</p>
2744
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2745<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2746
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2747<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2748<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2749<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2750<table border="1" cellspacing="0" cellpadding="4">
2751 <tbody>
2752 <tr>
2753 <td>In0</td>
2754 <td>In1</td>
2755 <td>Out</td>
2756 </tr>
2757 <tr>
2758 <td>0</td>
2759 <td>0</td>
2760 <td>0</td>
2761 </tr>
2762 <tr>
2763 <td>0</td>
2764 <td>1</td>
2765 <td>1</td>
2766 </tr>
2767 <tr>
2768 <td>1</td>
2769 <td>0</td>
2770 <td>1</td>
2771 </tr>
2772 <tr>
2773 <td>1</td>
2774 <td>1</td>
2775 <td>0</td>
2776 </tr>
2777 </tbody>
2778</table>
2779</div>
2780<p> </p>
2781<h5>Example:</h5>
2782<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2783 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2784 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2785 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2786</pre>
2787</div>
2788
2789<!-- ======================================================================= -->
2790<div class="doc_subsection">
2791 <a name="vectorops">Vector Operations</a>
2792</div>
2793
2794<div class="doc_text">
2795
2796<p>LLVM supports several instructions to represent vector operations in a
2797target-independent manner. These instructions cover the element-access and
2798vector-specific operations needed to process vectors effectively. While LLVM
2799does directly support these vector operations, many sophisticated algorithms
2800will want to use target-specific intrinsics to take full advantage of a specific
2801target.</p>
2802
2803</div>
2804
2805<!-- _______________________________________________________________________ -->
2806<div class="doc_subsubsection">
2807 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2808</div>
2809
2810<div class="doc_text">
2811
2812<h5>Syntax:</h5>
2813
2814<pre>
2815 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2816</pre>
2817
2818<h5>Overview:</h5>
2819
2820<p>
2821The '<tt>extractelement</tt>' instruction extracts a single scalar
2822element from a vector at a specified index.
2823</p>
2824
2825
2826<h5>Arguments:</h5>
2827
2828<p>
2829The first operand of an '<tt>extractelement</tt>' instruction is a
2830value of <a href="#t_vector">vector</a> type. The second operand is
2831an index indicating the position from which to extract the element.
2832The index may be a variable.</p>
2833
2834<h5>Semantics:</h5>
2835
2836<p>
2837The result is a scalar of the same type as the element type of
2838<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2839<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2840results are undefined.
2841</p>
2842
2843<h5>Example:</h5>
2844
2845<pre>
2846 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2847</pre>
2848</div>
2849
2850
2851<!-- _______________________________________________________________________ -->
2852<div class="doc_subsubsection">
2853 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2854</div>
2855
2856<div class="doc_text">
2857
2858<h5>Syntax:</h5>
2859
2860<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]2861 &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]2862</pre>
2863
2864<h5>Overview:</h5>
2865
2866<p>
2867The '<tt>insertelement</tt>' instruction inserts a scalar
2868element into a vector at a specified index.
2869</p>
2870
2871
2872<h5>Arguments:</h5>
2873
2874<p>
2875The first operand of an '<tt>insertelement</tt>' instruction is a
2876value of <a href="#t_vector">vector</a> type. The second operand is a
2877scalar value whose type must equal the element type of the first
2878operand. The third operand is an index indicating the position at
2879which to insert the value. The index may be a variable.</p>
2880
2881<h5>Semantics:</h5>
2882
2883<p>
2884The result is a vector of the same type as <tt>val</tt>. Its
2885element values are those of <tt>val</tt> except at position
2886<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2887exceeds the length of <tt>val</tt>, the results are undefined.
2888</p>
2889
2890<h5>Example:</h5>
2891
2892<pre>
2893 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2894</pre>
2895</div>
2896
2897<!-- _______________________________________________________________________ -->
2898<div class="doc_subsubsection">
2899 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2900</div>
2901
2902<div class="doc_text">
2903
2904<h5>Syntax:</h5>
2905
2906<pre>
2907 &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>
2908</pre>
2909
2910<h5>Overview:</h5>
2911
2912<p>
2913The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2914from two input vectors, returning a vector of the same type.
2915</p>
2916
2917<h5>Arguments:</h5>
2918
2919<p>
2920The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2921with types that match each other and types that match the result of the
2922instruction. The third argument is a shuffle mask, which has the same number
2923of elements as the other vector type, but whose element type is always 'i32'.
2924</p>
2925
2926<p>
2927The shuffle mask operand is required to be a constant vector with either
2928constant integer or undef values.
2929</p>
2930
2931<h5>Semantics:</h5>
2932
2933<p>
2934The elements of the two input vectors are numbered from left to right across
2935both of the vectors. The shuffle mask operand specifies, for each element of
2936the result vector, which element of the two input registers the result element
2937gets. The element selector may be undef (meaning "don't care") and the second
2938operand may be undef if performing a shuffle from only one vector.
2939</p>
2940
2941<h5>Example:</h5>
2942
2943<pre>
2944 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2945 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2946 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2947 &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.
2948</pre>
2949</div>
2950
2951
2952<!-- ======================================================================= -->
2953<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2954 <a name="aggregateops">Aggregate Operations</a>
2955</div>
2956
2957<div class="doc_text">
2958
2959<p>LLVM supports several instructions for working with aggregate values.
2960</p>
2961
2962</div>
2963
2964<!-- _______________________________________________________________________ -->
2965<div class="doc_subsubsection">
2966 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
2967</div>
2968
2969<div class="doc_text">
2970
2971<h5>Syntax:</h5>
2972
2973<pre>
2974 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
2975</pre>
2976
2977<h5>Overview:</h5>
2978
2979<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2980The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
2981or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2982</p>
2983
2984
2985<h5>Arguments:</h5>
2986
2987<p>
2988The first operand of an '<tt>extractvalue</tt>' instruction is a
2989value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2990type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2991in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2992'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2993</p>
2994
2995<h5>Semantics:</h5>
2996
2997<p>
2998The result is the value at the position in the aggregate specified by
2999the index operands.
3000</p>
3001
3002<h5>Example:</h5>
3003
3004<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3005 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3006</pre>
3007</div>
3008
3009
3010<!-- _______________________________________________________________________ -->
3011<div class="doc_subsubsection">
3012 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3013</div>
3014
3015<div class="doc_text">
3016
3017<h5>Syntax:</h5>
3018
3019<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3020 &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]3021</pre>
3022
3023<h5>Overview:</h5>
3024
3025<p>
3026The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3027into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3028</p>
3029
3030
3031<h5>Arguments:</h5>
3032
3033<p>
3034The first operand of an '<tt>insertvalue</tt>' instruction is a
3035value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3036The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3037The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3038indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3039indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3040'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3041The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3042by the indices.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3043
3044<h5>Semantics:</h5>
3045
3046<p>
3047The result is an aggregate of the same type as <tt>val</tt>. Its
3048value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3049specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3050</p>
3051
3052<h5>Example:</h5>
3053
3054<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3055 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3056</pre>
3057</div>
3058
3059
3060<!-- ======================================================================= -->
3061<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3062 <a name="memoryops">Memory Access and Addressing Operations</a>
3063</div>
3064
3065<div class="doc_text">
3066
3067<p>A key design point of an SSA-based representation is how it
3068represents memory. In LLVM, no memory locations are in SSA form, which
3069makes things very simple. This section describes how to read, write,
3070allocate, and free memory in LLVM.</p>
3071
3072</div>
3073
3074<!-- _______________________________________________________________________ -->
3075<div class="doc_subsubsection">
3076 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3077</div>
3078
3079<div class="doc_text">
3080
3081<h5>Syntax:</h5>
3082
3083<pre>
3084 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3085</pre>
3086
3087<h5>Overview:</h5>
3088
3089<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3090heap and returns a pointer to it. The object is always allocated in the generic
3091address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3092
3093<h5>Arguments:</h5>
3094
3095<p>The '<tt>malloc</tt>' instruction allocates
3096<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3097bytes of memory from the operating system and returns a pointer of the
3098appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3099number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3100If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3101be aligned to at least that boundary. If not specified, or if zero, the target can
3102choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3103
3104<p>'<tt>type</tt>' must be a sized type.</p>
3105
3106<h5>Semantics:</h5>
3107
3108<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3109a pointer is returned. The result of a zero byte allocattion is undefined. The
3110result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3111
3112<h5>Example:</h5>
3113
3114<pre>
3115 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
3116
3117 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3118 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3119 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3120 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3121 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3122</pre>
3123</div>
3124
3125<!-- _______________________________________________________________________ -->
3126<div class="doc_subsubsection">
3127 <a name="i_free">'<tt>free</tt>' Instruction</a>
3128</div>
3129
3130<div class="doc_text">
3131
3132<h5>Syntax:</h5>
3133
3134<pre>
3135 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
3136</pre>
3137
3138<h5>Overview:</h5>
3139
3140<p>The '<tt>free</tt>' instruction returns memory back to the unused
3141memory heap to be reallocated in the future.</p>
3142
3143<h5>Arguments:</h5>
3144
3145<p>'<tt>value</tt>' shall be a pointer value that points to a value
3146that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3147instruction.</p>
3148
3149<h5>Semantics:</h5>
3150
3151<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3152after this instruction executes. If the pointer is null, the operation
3153is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3154
3155<h5>Example:</h5>
3156
3157<pre>
3158 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3159 free [4 x i8]* %array
3160</pre>
3161</div>
3162
3163<!-- _______________________________________________________________________ -->
3164<div class="doc_subsubsection">
3165 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3166</div>
3167
3168<div class="doc_text">
3169
3170<h5>Syntax:</h5>
3171
3172<pre>
3173 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3174</pre>
3175
3176<h5>Overview:</h5>
3177
3178<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3179currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3180returns to its caller. The object is always allocated in the generic address
3181space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3182
3183<h5>Arguments:</h5>
3184
3185<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3186bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3187appropriate type to the program. If "NumElements" is specified, it is the
3188number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3189If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3190to be aligned to at least that boundary. If not specified, or if zero, the target
3191can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3192
3193<p>'<tt>type</tt>' may be any sized type.</p>
3194
3195<h5>Semantics:</h5>
3196
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3197<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3198there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3199memory is automatically released when the function returns. The '<tt>alloca</tt>'
3200instruction is commonly used to represent automatic variables that must
3201have an address available. When the function returns (either with the <tt><a
3202 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]3203instructions), the memory is reclaimed. Allocating zero bytes
3204is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3205
3206<h5>Example:</h5>
3207
3208<pre>
3209 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3210 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3211 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3212 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3213</pre>
3214</div>
3215
3216<!-- _______________________________________________________________________ -->
3217<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3218Instruction</a> </div>
3219<div class="doc_text">
3220<h5>Syntax:</h5>
3221<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>
3222<h5>Overview:</h5>
3223<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3224<h5>Arguments:</h5>
3225<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3226address from which to load. The pointer must point to a <a
3227 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3228marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3229the number or order of execution of this <tt>load</tt> with other
3230volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3231instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3232<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3233The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3234(that is, the alignment of the memory address). A value of 0 or an
3235omitted "align" argument means that the operation has the preferential
3236alignment for the target. It is the responsibility of the code emitter
3237to ensure that the alignment information is correct. Overestimating
3238the alignment results in an undefined behavior. Underestimating the
3239alignment may produce less efficient code. An alignment of 1 is always
3240safe.
3241</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3242<h5>Semantics:</h5>
3243<p>The location of memory pointed to is loaded.</p>
3244<h5>Examples:</h5>
3245<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3246 <a
3247 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3248 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3249</pre>
3250</div>
3251<!-- _______________________________________________________________________ -->
3252<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3253Instruction</a> </div>
3254<div class="doc_text">
3255<h5>Syntax:</h5>
3256<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3257 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3258</pre>
3259<h5>Overview:</h5>
3260<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3261<h5>Arguments:</h5>
3262<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3263to 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]3264operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3265of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3266operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3267optimizer is not allowed to modify the number or order of execution of
3268this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3269 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3270<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3271The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3272(that is, the alignment of the memory address). A value of 0 or an
3273omitted "align" argument means that the operation has the preferential
3274alignment for the target. It is the responsibility of the code emitter
3275to ensure that the alignment information is correct. Overestimating
3276the alignment results in an undefined behavior. Underestimating the
3277alignment may produce less efficient code. An alignment of 1 is always
3278safe.
3279</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3280<h5>Semantics:</h5>
3281<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3282at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3283<h5>Example:</h5>
3284<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3285 store i32 3, i32* %ptr <i>; yields {void}</i>
3286 %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]3287</pre>
3288</div>
3289
3290<!-- _______________________________________________________________________ -->
3291<div class="doc_subsubsection">
3292 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3293</div>
3294
3295<div class="doc_text">
3296<h5>Syntax:</h5>
3297<pre>
3298 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
3299</pre>
3300
3301<h5>Overview:</h5>
3302
3303<p>
3304The '<tt>getelementptr</tt>' instruction is used to get the address of a
3305subelement of an aggregate data structure.</p>
3306
3307<h5>Arguments:</h5>
3308
3309<p>This instruction takes a list of integer operands that indicate what
3310elements of the aggregate object to index to. The actual types of the arguments
3311provided depend on the type of the first pointer argument. The
3312'<tt>getelementptr</tt>' instruction is used to index down through the type
3313levels of a structure or to a specific index in an array. When indexing into a
3314structure, only <tt>i32</tt> integer constants are allowed. When indexing
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3315into an array or pointer, only integers of 32 or 64 bits are allowed; 32-bit
3316values will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3317
3318<p>For example, let's consider a C code fragment and how it gets
3319compiled to LLVM:</p>
3320
3321<div class="doc_code">
3322<pre>
3323struct RT {
3324 char A;
3325 int B[10][20];
3326 char C;
3327};
3328struct ST {
3329 int X;
3330 double Y;
3331 struct RT Z;
3332};
3333
3334int *foo(struct ST *s) {
3335 return &amp;s[1].Z.B[5][13];
3336}
3337</pre>
3338</div>
3339
3340<p>The LLVM code generated by the GCC frontend is:</p>
3341
3342<div class="doc_code">
3343<pre>
3344%RT = type { i8 , [10 x [20 x i32]], i8 }
3345%ST = type { i32, double, %RT }
3346
3347define i32* %foo(%ST* %s) {
3348entry:
3349 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3350 ret i32* %reg
3351}
3352</pre>
3353</div>
3354
3355<h5>Semantics:</h5>
3356
3357<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3358on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3359and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3360<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]3361to 64-bits. <a href="#t_struct">Structure</a> and <a href="#t_pstruct">packed
3362structure</a> types require <tt>i32</tt> <b>constants</b>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3363
3364<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3365type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3366}</tt>' type, a structure. The second index indexes into the third element of
3367the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3368i8 }</tt>' type, another structure. The third index indexes into the second
3369element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3370array. The two dimensions of the array are subscripted into, yielding an
3371'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3372to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3373
3374<p>Note that it is perfectly legal to index partially through a
3375structure, returning a pointer to an inner element. Because of this,
3376the LLVM code for the given testcase is equivalent to:</p>
3377
3378<pre>
3379 define i32* %foo(%ST* %s) {
3380 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3381 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3382 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3383 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3384 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3385 ret i32* %t5
3386 }
3387</pre>
3388
3389<p>Note that it is undefined to access an array out of bounds: array and
3390pointer indexes must always be within the defined bounds of the array type.
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3391The one exception for this rule is zero length arrays. These arrays are
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3392defined to be accessible as variable length arrays, which requires access
3393beyond the zero'th element.</p>
3394
3395<p>The getelementptr instruction is often confusing. For some more insight
3396into how it works, see <a href="GetElementPtr.html">the getelementptr
3397FAQ</a>.</p>
3398
3399<h5>Example:</h5>
3400
3401<pre>
3402 <i>; yields [12 x i8]*:aptr</i>
3403 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3404</pre>
3405</div>
3406
3407<!-- ======================================================================= -->
3408<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3409</div>
3410<div class="doc_text">
3411<p>The instructions in this category are the conversion instructions (casting)
3412which all take a single operand and a type. They perform various bit conversions
3413on the operand.</p>
3414</div>
3415
3416<!-- _______________________________________________________________________ -->
3417<div class="doc_subsubsection">
3418 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3419</div>
3420<div class="doc_text">
3421
3422<h5>Syntax:</h5>
3423<pre>
3424 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3425</pre>
3426
3427<h5>Overview:</h5>
3428<p>
3429The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3430</p>
3431
3432<h5>Arguments:</h5>
3433<p>
3434The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3435be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3436and type of the result, which must be an <a href="#t_integer">integer</a>
3437type. The bit size of <tt>value</tt> must be larger than the bit size of
3438<tt>ty2</tt>. Equal sized types are not allowed.</p>
3439
3440<h5>Semantics:</h5>
3441<p>
3442The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3443and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3444larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3445It will always truncate bits.</p>
3446
3447<h5>Example:</h5>
3448<pre>
3449 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3450 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3451 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3452</pre>
3453</div>
3454
3455<!-- _______________________________________________________________________ -->
3456<div class="doc_subsubsection">
3457 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3458</div>
3459<div class="doc_text">
3460
3461<h5>Syntax:</h5>
3462<pre>
3463 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3464</pre>
3465
3466<h5>Overview:</h5>
3467<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3468<tt>ty2</tt>.</p>
3469
3470
3471<h5>Arguments:</h5>
3472<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3473<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3474also be of <a href="#t_integer">integer</a> type. The bit size of the
3475<tt>value</tt> must be smaller than the bit size of the destination type,
3476<tt>ty2</tt>.</p>
3477
3478<h5>Semantics:</h5>
3479<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3480bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3481
3482<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3483
3484<h5>Example:</h5>
3485<pre>
3486 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3487 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3488</pre>
3489</div>
3490
3491<!-- _______________________________________________________________________ -->
3492<div class="doc_subsubsection">
3493 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3494</div>
3495<div class="doc_text">
3496
3497<h5>Syntax:</h5>
3498<pre>
3499 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3500</pre>
3501
3502<h5>Overview:</h5>
3503<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3504
3505<h5>Arguments:</h5>
3506<p>
3507The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3508<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3509also be of <a href="#t_integer">integer</a> type. The bit size of the
3510<tt>value</tt> must be smaller than the bit size of the destination type,
3511<tt>ty2</tt>.</p>
3512
3513<h5>Semantics:</h5>
3514<p>
3515The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3516bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3517the type <tt>ty2</tt>.</p>
3518
3519<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3520
3521<h5>Example:</h5>
3522<pre>
3523 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3524 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3525</pre>
3526</div>
3527
3528<!-- _______________________________________________________________________ -->
3529<div class="doc_subsubsection">
3530 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3531</div>
3532
3533<div class="doc_text">
3534
3535<h5>Syntax:</h5>
3536
3537<pre>
3538 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3539</pre>
3540
3541<h5>Overview:</h5>
3542<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3543<tt>ty2</tt>.</p>
3544
3545
3546<h5>Arguments:</h5>
3547<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3548 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3549cast it to. The size of <tt>value</tt> must be larger than the size of
3550<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3551<i>no-op cast</i>.</p>
3552
3553<h5>Semantics:</h5>
3554<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3555<a href="#t_floating">floating point</a> type to a smaller
3556<a href="#t_floating">floating point</a> type. If the value cannot fit within
3557the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3558
3559<h5>Example:</h5>
3560<pre>
3561 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3562 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3563</pre>
3564</div>
3565
3566<!-- _______________________________________________________________________ -->
3567<div class="doc_subsubsection">
3568 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3569</div>
3570<div class="doc_text">
3571
3572<h5>Syntax:</h5>
3573<pre>
3574 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3575</pre>
3576
3577<h5>Overview:</h5>
3578<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3579floating point value.</p>
3580
3581<h5>Arguments:</h5>
3582<p>The '<tt>fpext</tt>' instruction takes a
3583<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3584and a <a href="#t_floating">floating point</a> type to cast it to. The source
3585type must be smaller than the destination type.</p>
3586
3587<h5>Semantics:</h5>
3588<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3589<a href="#t_floating">floating point</a> type to a larger
3590<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3591used to make a <i>no-op cast</i> because it always changes bits. Use
3592<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3593
3594<h5>Example:</h5>
3595<pre>
3596 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3597 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3598</pre>
3599</div>
3600
3601<!-- _______________________________________________________________________ -->
3602<div class="doc_subsubsection">
3603 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3604</div>
3605<div class="doc_text">
3606
3607<h5>Syntax:</h5>
3608<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3609 &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]3610</pre>
3611
3612<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3613<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3614unsigned integer equivalent of type <tt>ty2</tt>.
3615</p>
3616
3617<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3618<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]3619scalar or vector <a href="#t_floating">floating point</a> value, and a type
3620to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3621type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3622vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3623
3624<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3625<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3626<a href="#t_floating">floating point</a> operand into the nearest (rounding
3627towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3628the results are undefined.</p>
3629
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3630<h5>Example:</h5>
3631<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3632 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3633 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3634 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3635</pre>
3636</div>
3637
3638<!-- _______________________________________________________________________ -->
3639<div class="doc_subsubsection">
3640 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3641</div>
3642<div class="doc_text">
3643
3644<h5>Syntax:</h5>
3645<pre>
3646 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3647</pre>
3648
3649<h5>Overview:</h5>
3650<p>The '<tt>fptosi</tt>' instruction converts
3651<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3652</p>
3653
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3654<h5>Arguments:</h5>
3655<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]3656scalar or vector <a href="#t_floating">floating point</a> value, and a type
3657to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3658type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3659vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3660
3661<h5>Semantics:</h5>
3662<p>The '<tt>fptosi</tt>' instruction converts its
3663<a href="#t_floating">floating point</a> operand into the nearest (rounding
3664towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3665the results are undefined.</p>
3666
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3667<h5>Example:</h5>
3668<pre>
3669 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3670 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3671 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3672</pre>
3673</div>
3674
3675<!-- _______________________________________________________________________ -->
3676<div class="doc_subsubsection">
3677 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3678</div>
3679<div class="doc_text">
3680
3681<h5>Syntax:</h5>
3682<pre>
3683 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3684</pre>
3685
3686<h5>Overview:</h5>
3687<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3688integer and converts that value to the <tt>ty2</tt> type.</p>
3689
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3690<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3691<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3692scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3693to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3694type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3695floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3696
3697<h5>Semantics:</h5>
3698<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3699integer quantity and converts it to the corresponding floating point value. If
3700the value cannot fit in the floating point value, the results are undefined.</p>
3701
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3702<h5>Example:</h5>
3703<pre>
3704 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3705 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3706</pre>
3707</div>
3708
3709<!-- _______________________________________________________________________ -->
3710<div class="doc_subsubsection">
3711 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3712</div>
3713<div class="doc_text">
3714
3715<h5>Syntax:</h5>
3716<pre>
3717 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3718</pre>
3719
3720<h5>Overview:</h5>
3721<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3722integer and converts that value to the <tt>ty2</tt> type.</p>
3723
3724<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3725<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3726scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3727to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3728type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3729floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3730
3731<h5>Semantics:</h5>
3732<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3733integer quantity and converts it to the corresponding floating point value. If
3734the value cannot fit in the floating point value, the results are undefined.</p>
3735
3736<h5>Example:</h5>
3737<pre>
3738 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3739 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3740</pre>
3741</div>
3742
3743<!-- _______________________________________________________________________ -->
3744<div class="doc_subsubsection">
3745 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3746</div>
3747<div class="doc_text">
3748
3749<h5>Syntax:</h5>
3750<pre>
3751 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3752</pre>
3753
3754<h5>Overview:</h5>
3755<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3756the integer type <tt>ty2</tt>.</p>
3757
3758<h5>Arguments:</h5>
3759<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3760must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3761<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3762
3763<h5>Semantics:</h5>
3764<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3765<tt>ty2</tt> by interpreting the pointer value as an integer and either
3766truncating or zero extending that value to the size of the integer type. If
3767<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3768<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3769are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3770change.</p>
3771
3772<h5>Example:</h5>
3773<pre>
3774 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3775 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3776</pre>
3777</div>
3778
3779<!-- _______________________________________________________________________ -->
3780<div class="doc_subsubsection">
3781 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3782</div>
3783<div class="doc_text">
3784
3785<h5>Syntax:</h5>
3786<pre>
3787 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3788</pre>
3789
3790<h5>Overview:</h5>
3791<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3792a pointer type, <tt>ty2</tt>.</p>
3793
3794<h5>Arguments:</h5>
3795<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3796value to cast, and a type to cast it to, which must be a
3797<a href="#t_pointer">pointer</a> type.
3798
3799<h5>Semantics:</h5>
3800<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3801<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3802the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3803size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3804the size of a pointer then a zero extension is done. If they are the same size,
3805nothing is done (<i>no-op cast</i>).</p>
3806
3807<h5>Example:</h5>
3808<pre>
3809 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3810 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3811 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3812</pre>
3813</div>
3814
3815<!-- _______________________________________________________________________ -->
3816<div class="doc_subsubsection">
3817 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3818</div>
3819<div class="doc_text">
3820
3821<h5>Syntax:</h5>
3822<pre>
3823 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3824</pre>
3825
3826<h5>Overview:</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 converts <tt>value</tt> to type
3829<tt>ty2</tt> without changing any bits.</p>
3830
3831<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3832
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3833<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]3834a non-aggregate first class value, and a type to cast it to, which must also be
3835a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
3836<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3837and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3838type is a pointer, the destination type must also be a pointer. This
3839instruction supports bitwise conversion of vectors to integers and to vectors
3840of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3841
3842<h5>Semantics:</h5>
3843<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3844<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3845this conversion. The conversion is done as if the <tt>value</tt> had been
3846stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3847converted to other pointer types with this instruction. To convert pointers to
3848other types, use the <a href="#i_inttoptr">inttoptr</a> or
3849<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3850
3851<h5>Example:</h5>
3852<pre>
3853 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3854 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3855 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3856</pre>
3857</div>
3858
3859<!-- ======================================================================= -->
3860<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3861<div class="doc_text">
3862<p>The instructions in this category are the "miscellaneous"
3863instructions, which defy better classification.</p>
3864</div>
3865
3866<!-- _______________________________________________________________________ -->
3867<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3868</div>
3869<div class="doc_text">
3870<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3871<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]3872</pre>
3873<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3874<p>The '<tt>icmp</tt>' instruction returns a boolean value or
3875a vector of boolean values based on comparison
3876of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3877<h5>Arguments:</h5>
3878<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3879the condition code indicating the kind of comparison to perform. It is not
3880a value, just a keyword. The possible condition code are:
3881<ol>
3882 <li><tt>eq</tt>: equal</li>
3883 <li><tt>ne</tt>: not equal </li>
3884 <li><tt>ugt</tt>: unsigned greater than</li>
3885 <li><tt>uge</tt>: unsigned greater or equal</li>
3886 <li><tt>ult</tt>: unsigned less than</li>
3887 <li><tt>ule</tt>: unsigned less or equal</li>
3888 <li><tt>sgt</tt>: signed greater than</li>
3889 <li><tt>sge</tt>: signed greater or equal</li>
3890 <li><tt>slt</tt>: signed less than</li>
3891 <li><tt>sle</tt>: signed less or equal</li>
3892</ol>
3893<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3894<a href="#t_pointer">pointer</a>
3895or integer <a href="#t_vector">vector</a> typed.
3896They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3897<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3898<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]3899the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3900yields 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]3901<ol>
3902 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3903 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3904 </li>
3905 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3906 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3907 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3908 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3909 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3910 <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]3911 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3912 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3913 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3914 <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]3915 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3916 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3917 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3918 <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]3919 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3920 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3921 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3922 <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]3923</ol>
3924<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3925values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3926<p>If the operands are integer vectors, then they are compared
3927element by element. The result is an <tt>i1</tt> vector with
3928the same number of elements as the values being compared.
3929Otherwise, the result is an <tt>i1</tt>.
3930</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3931
3932<h5>Example:</h5>
3933<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3934 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3935 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3936 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3937 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3938 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3939</pre>
3940</div>
3941
3942<!-- _______________________________________________________________________ -->
3943<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3944</div>
3945<div class="doc_text">
3946<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3947<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]3948</pre>
3949<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3950<p>The '<tt>fcmp</tt>' instruction returns a boolean value
3951or vector of boolean values based on comparison
3952of its operands.
3953<p>
3954If the operands are floating point scalars, then the result
3955type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
3956</p>
3957<p>If the operands are floating point vectors, then the result type
3958is a vector of boolean with the same number of elements as the
3959operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3960<h5>Arguments:</h5>
3961<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3962the condition code indicating the kind of comparison to perform. It is not
3963a value, just a keyword. The possible condition code are:
3964<ol>
3965 <li><tt>false</tt>: no comparison, always returns false</li>
3966 <li><tt>oeq</tt>: ordered and equal</li>
3967 <li><tt>ogt</tt>: ordered and greater than </li>
3968 <li><tt>oge</tt>: ordered and greater than or equal</li>
3969 <li><tt>olt</tt>: ordered and less than </li>
3970 <li><tt>ole</tt>: ordered and less than or equal</li>
3971 <li><tt>one</tt>: ordered and not equal</li>
3972 <li><tt>ord</tt>: ordered (no nans)</li>
3973 <li><tt>ueq</tt>: unordered or equal</li>
3974 <li><tt>ugt</tt>: unordered or greater than </li>
3975 <li><tt>uge</tt>: unordered or greater than or equal</li>
3976 <li><tt>ult</tt>: unordered or less than </li>
3977 <li><tt>ule</tt>: unordered or less than or equal</li>
3978 <li><tt>une</tt>: unordered or not equal</li>
3979 <li><tt>uno</tt>: unordered (either nans)</li>
3980 <li><tt>true</tt>: no comparison, always returns true</li>
3981</ol>
3982<p><i>Ordered</i> means that neither operand is a QNAN while
3983<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3984<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
3985either a <a href="#t_floating">floating point</a> type
3986or a <a href="#t_vector">vector</a> of floating point type.
3987They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3988<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3989<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3990according to the condition code given as <tt>cond</tt>.
3991If the operands are vectors, then the vectors are compared
3992element by element.
3993Each comparison performed
3994always yields an <a href="#t_primitive">i1</a> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3995<ol>
3996 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3997 <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]3998 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3999 <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]4000 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4001 <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]4002 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4003 <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]4004 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4005 <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]4006 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4007 <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]4008 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4009 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4010 <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]4011 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4012 <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]4013 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4014 <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]4015 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4016 <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]4017 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4018 <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]4019 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4020 <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]4021 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4022 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4023 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4024</ol>
4025
4026<h5>Example:</h5>
4027<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]4028 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4029 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4030 &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]4031</pre>
4032</div>
4033
4034<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4035<div class="doc_subsubsection">
4036 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4037</div>
4038<div class="doc_text">
4039<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4040<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]4041</pre>
4042<h5>Overview:</h5>
4043<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4044element-wise comparison of its two integer vector operands.</p>
4045<h5>Arguments:</h5>
4046<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4047the condition code indicating the kind of comparison to perform. It is not
4048a value, just a keyword. The possible condition code are:
4049<ol>
4050 <li><tt>eq</tt>: equal</li>
4051 <li><tt>ne</tt>: not equal </li>
4052 <li><tt>ugt</tt>: unsigned greater than</li>
4053 <li><tt>uge</tt>: unsigned greater or equal</li>
4054 <li><tt>ult</tt>: unsigned less than</li>
4055 <li><tt>ule</tt>: unsigned less or equal</li>
4056 <li><tt>sgt</tt>: signed greater than</li>
4057 <li><tt>sge</tt>: signed greater or equal</li>
4058 <li><tt>slt</tt>: signed less than</li>
4059 <li><tt>sle</tt>: signed less or equal</li>
4060</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4061<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4062<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4063<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4064<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4065according to the condition code given as <tt>cond</tt>. The comparison yields a
4066<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4067identical type as the values being compared. The most significant bit in each
4068element is 1 if the element-wise comparison evaluates to true, and is 0
4069otherwise. All other bits of the result are undefined. The condition codes
4070are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4071instruction</a>.
4072
4073<h5>Example:</h5>
4074<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4075 &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>
4076 &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]4077</pre>
4078</div>
4079
4080<!-- _______________________________________________________________________ -->
4081<div class="doc_subsubsection">
4082 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4083</div>
4084<div class="doc_text">
4085<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4086<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]4087<h5>Overview:</h5>
4088<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4089element-wise comparison of its two floating point vector operands. The output
4090elements have the same width as the input elements.</p>
4091<h5>Arguments:</h5>
4092<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4093the condition code indicating the kind of comparison to perform. It is not
4094a value, just a keyword. The possible condition code are:
4095<ol>
4096 <li><tt>false</tt>: no comparison, always returns false</li>
4097 <li><tt>oeq</tt>: ordered and equal</li>
4098 <li><tt>ogt</tt>: ordered and greater than </li>
4099 <li><tt>oge</tt>: ordered and greater than or equal</li>
4100 <li><tt>olt</tt>: ordered and less than </li>
4101 <li><tt>ole</tt>: ordered and less than or equal</li>
4102 <li><tt>one</tt>: ordered and not equal</li>
4103 <li><tt>ord</tt>: ordered (no nans)</li>
4104 <li><tt>ueq</tt>: unordered or equal</li>
4105 <li><tt>ugt</tt>: unordered or greater than </li>
4106 <li><tt>uge</tt>: unordered or greater than or equal</li>
4107 <li><tt>ult</tt>: unordered or less than </li>
4108 <li><tt>ule</tt>: unordered or less than or equal</li>
4109 <li><tt>une</tt>: unordered or not equal</li>
4110 <li><tt>uno</tt>: unordered (either nans)</li>
4111 <li><tt>true</tt>: no comparison, always returns true</li>
4112</ol>
4113<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4114<a href="#t_floating">floating point</a> typed. They must also be identical
4115types.</p>
4116<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4117<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4118according to the condition code given as <tt>cond</tt>. The comparison yields a
4119<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4120an identical number of elements as the values being compared, and each element
4121having identical with to the width of the floating point elements. The most
4122significant bit in each element is 1 if the element-wise comparison evaluates to
4123true, and is 0 otherwise. All other bits of the result are undefined. The
4124condition codes are evaluated identically to the
4125<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.
4126
4127<h5>Example:</h5>
4128<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4129 &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>
4130 &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]4131</pre>
4132</div>
4133
4134<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4135<div class="doc_subsubsection">
4136 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4137</div>
4138
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4139<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4140
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4141<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4142
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4143<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4144<h5>Overview:</h5>
4145<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4146the SSA graph representing the function.</p>
4147<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4148
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4149<p>The type of the incoming values is specified with the first type
4150field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4151as arguments, with one pair for each predecessor basic block of the
4152current block. Only values of <a href="#t_firstclass">first class</a>
4153type may be used as the value arguments to the PHI node. Only labels
4154may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4155
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4156<p>There must be no non-phi instructions between the start of a basic
4157block and the PHI instructions: i.e. PHI instructions must be first in
4158a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4159
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4160<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4161
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4162<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4163specified by the pair corresponding to the predecessor basic block that executed
4164just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4165
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4166<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4167<pre>
4168Loop: ; Infinite loop that counts from 0 on up...
4169 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4170 %nextindvar = add i32 %indvar, 1
4171 br label %Loop
4172</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4173</div>
4174
4175<!-- _______________________________________________________________________ -->
4176<div class="doc_subsubsection">
4177 <a name="i_select">'<tt>select</tt>' Instruction</a>
4178</div>
4179
4180<div class="doc_text">
4181
4182<h5>Syntax:</h5>
4183
4184<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4185 &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>
4186
4187 <i>selty</i> is either i1 or {&lt;N x i1&gt}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4188</pre>
4189
4190<h5>Overview:</h5>
4191
4192<p>
4193The '<tt>select</tt>' instruction is used to choose one value based on a
4194condition, without branching.
4195</p>
4196
4197
4198<h5>Arguments:</h5>
4199
4200<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4201The '<tt>select</tt>' instruction requires an 'i1' value or
4202a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4203condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4204type. If the val1/val2 are vectors and
4205the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4206individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4207</p>
4208
4209<h5>Semantics:</h5>
4210
4211<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4212If 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]4213value argument; otherwise, it returns the second value argument.
4214</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4215<p>
4216If the condition is a vector of i1, then the value arguments must
4217be vectors of the same size, and the selection is done element
4218by element.
4219</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4220
4221<h5>Example:</h5>
4222
4223<pre>
4224 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4225</pre>
4226</div>
4227
4228
4229<!-- _______________________________________________________________________ -->
4230<div class="doc_subsubsection">
4231 <a name="i_call">'<tt>call</tt>' Instruction</a>
4232</div>
4233
4234<div class="doc_text">
4235
4236<h5>Syntax:</h5>
4237<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4238 &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]4239</pre>
4240
4241<h5>Overview:</h5>
4242
4243<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4244
4245<h5>Arguments:</h5>
4246
4247<p>This instruction requires several arguments:</p>
4248
4249<ol>
4250 <li>
4251 <p>The optional "tail" marker indicates whether the callee function accesses
4252 any allocas or varargs in the caller. If the "tail" marker is present, the
4253 function call is eligible for tail call optimization. Note that calls may
4254 be marked "tail" even if they do not occur before a <a
4255 href="#i_ret"><tt>ret</tt></a> instruction.
4256 </li>
4257 <li>
4258 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4259 convention</a> the call should use. If none is specified, the call defaults
4260 to using C calling conventions.
4261 </li>
4262 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4263 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4264 the type of the return value. Functions that return no value are marked
4265 <tt><a href="#t_void">void</a></tt>.</p>
4266 </li>
4267 <li>
4268 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4269 value being invoked. The argument types must match the types implied by
4270 this signature. This type can be omitted if the function is not varargs
4271 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4272 </li>
4273 <li>
4274 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4275 be invoked. In most cases, this is a direct function invocation, but
4276 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4277 to function value.</p>
4278 </li>
4279 <li>
4280 <p>'<tt>function args</tt>': argument list whose types match the
4281 function signature argument types. All arguments must be of
4282 <a href="#t_firstclass">first class</a> type. If the function signature
4283 indicates the function accepts a variable number of arguments, the extra
4284 arguments can be specified.</p>
4285 </li>
4286</ol>
4287
4288<h5>Semantics:</h5>
4289
4290<p>The '<tt>call</tt>' instruction is used to cause control flow to
4291transfer to a specified function, with its incoming arguments bound to
4292the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4293instruction in the called function, control flow continues with the
4294instruction after the function call, and the return value of the
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4295function is bound to the result argument.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4296
4297<h5>Example:</h5>
4298
4299<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4300 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4301 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4302 %X = tail call i32 @foo() <i>; yields i32</i>
4303 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4304 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4305
4306 %struct.A = type { i32, i8 }
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4307 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4308 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4309 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4310</pre>
4311
4312</div>
4313
4314<!-- _______________________________________________________________________ -->
4315<div class="doc_subsubsection">
4316 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4317</div>
4318
4319<div class="doc_text">
4320
4321<h5>Syntax:</h5>
4322
4323<pre>
4324 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4325</pre>
4326
4327<h5>Overview:</h5>
4328
4329<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4330the "variable argument" area of a function call. It is used to implement the
4331<tt>va_arg</tt> macro in C.</p>
4332
4333<h5>Arguments:</h5>
4334
4335<p>This instruction takes a <tt>va_list*</tt> value and the type of
4336the argument. It returns a value of the specified argument type and
4337increments the <tt>va_list</tt> to point to the next argument. The
4338actual type of <tt>va_list</tt> is target specific.</p>
4339
4340<h5>Semantics:</h5>
4341
4342<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4343type from the specified <tt>va_list</tt> and causes the
4344<tt>va_list</tt> to point to the next argument. For more information,
4345see the variable argument handling <a href="#int_varargs">Intrinsic
4346Functions</a>.</p>
4347
4348<p>It is legal for this instruction to be called in a function which does not
4349take a variable number of arguments, for example, the <tt>vfprintf</tt>
4350function.</p>
4351
4352<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4353href="#intrinsics">intrinsic function</a> because it takes a type as an
4354argument.</p>
4355
4356<h5>Example:</h5>
4357
4358<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4359
4360</div>
4361
4362<!-- *********************************************************************** -->
4363<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4364<!-- *********************************************************************** -->
4365
4366<div class="doc_text">
4367
4368<p>LLVM supports the notion of an "intrinsic function". These functions have
4369well known names and semantics and are required to follow certain restrictions.
4370Overall, these intrinsics represent an extension mechanism for the LLVM
4371language that does not require changing all of the transformations in LLVM when
4372adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4373
4374<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4375prefix is reserved in LLVM for intrinsic names; thus, function names may not
4376begin with this prefix. Intrinsic functions must always be external functions:
4377you cannot define the body of intrinsic functions. Intrinsic functions may
4378only be used in call or invoke instructions: it is illegal to take the address
4379of an intrinsic function. Additionally, because intrinsic functions are part
4380of the LLVM language, it is required if any are added that they be documented
4381here.</p>
4382
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4383<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4384a family of functions that perform the same operation but on different data
4385types. Because LLVM can represent over 8 million different integer types,
4386overloading is used commonly to allow an intrinsic function to operate on any
4387integer type. One or more of the argument types or the result type can be
4388overloaded to accept any integer type. Argument types may also be defined as
4389exactly matching a previous argument's type or the result type. This allows an
4390intrinsic function which accepts multiple arguments, but needs all of them to
4391be of the same type, to only be overloaded with respect to a single argument or
4392the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4393
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4394<p>Overloaded intrinsics will have the names of its overloaded argument types
4395encoded into its function name, each preceded by a period. Only those types
4396which are overloaded result in a name suffix. Arguments whose type is matched
4397against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4398take an integer of any width and returns an integer of exactly the same integer
4399width. This leads to a family of functions such as
4400<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4401Only one type, the return type, is overloaded, and only one type suffix is
4402required. Because the argument's type is matched against the return type, it
4403does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4404
4405<p>To learn how to add an intrinsic function, please see the
4406<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4407</p>
4408
4409</div>
4410
4411<!-- ======================================================================= -->
4412<div class="doc_subsection">
4413 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4414</div>
4415
4416<div class="doc_text">
4417
4418<p>Variable argument support is defined in LLVM with the <a
4419 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4420intrinsic functions. These functions are related to the similarly
4421named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4422
4423<p>All of these functions operate on arguments that use a
4424target-specific value type "<tt>va_list</tt>". The LLVM assembly
4425language reference manual does not define what this type is, so all
4426transformations should be prepared to handle these functions regardless of
4427the type used.</p>
4428
4429<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4430instruction and the variable argument handling intrinsic functions are
4431used.</p>
4432
4433<div class="doc_code">
4434<pre>
4435define i32 @test(i32 %X, ...) {
4436 ; Initialize variable argument processing
4437 %ap = alloca i8*
4438 %ap2 = bitcast i8** %ap to i8*
4439 call void @llvm.va_start(i8* %ap2)
4440
4441 ; Read a single integer argument
4442 %tmp = va_arg i8** %ap, i32
4443
4444 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4445 %aq = alloca i8*
4446 %aq2 = bitcast i8** %aq to i8*
4447 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4448 call void @llvm.va_end(i8* %aq2)
4449
4450 ; Stop processing of arguments.
4451 call void @llvm.va_end(i8* %ap2)
4452 ret i32 %tmp
4453}
4454
4455declare void @llvm.va_start(i8*)
4456declare void @llvm.va_copy(i8*, i8*)
4457declare void @llvm.va_end(i8*)
4458</pre>
4459</div>
4460
4461</div>
4462
4463<!-- _______________________________________________________________________ -->
4464<div class="doc_subsubsection">
4465 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4466</div>
4467
4468
4469<div class="doc_text">
4470<h5>Syntax:</h5>
4471<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4472<h5>Overview:</h5>
4473<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
4474<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4475href="#i_va_arg">va_arg</a></tt>.</p>
4476
4477<h5>Arguments:</h5>
4478
4479<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4480
4481<h5>Semantics:</h5>
4482
4483<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4484macro available in C. In a target-dependent way, it initializes the
4485<tt>va_list</tt> element to which the argument points, so that the next call to
4486<tt>va_arg</tt> will produce the first variable argument passed to the function.
4487Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4488last argument of the function as the compiler can figure that out.</p>
4489
4490</div>
4491
4492<!-- _______________________________________________________________________ -->
4493<div class="doc_subsubsection">
4494 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4495</div>
4496
4497<div class="doc_text">
4498<h5>Syntax:</h5>
4499<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4500<h5>Overview:</h5>
4501
4502<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4503which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4504or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4505
4506<h5>Arguments:</h5>
4507
4508<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4509
4510<h5>Semantics:</h5>
4511
4512<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4513macro available in C. In a target-dependent way, it destroys the
4514<tt>va_list</tt> element to which the argument points. Calls to <a
4515href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4516<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4517<tt>llvm.va_end</tt>.</p>
4518
4519</div>
4520
4521<!-- _______________________________________________________________________ -->
4522<div class="doc_subsubsection">
4523 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4524</div>
4525
4526<div class="doc_text">
4527
4528<h5>Syntax:</h5>
4529
4530<pre>
4531 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4532</pre>
4533
4534<h5>Overview:</h5>
4535
4536<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4537from the source argument list to the destination argument list.</p>
4538
4539<h5>Arguments:</h5>
4540
4541<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4542The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4543
4544
4545<h5>Semantics:</h5>
4546
4547<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4548macro available in C. In a target-dependent way, it copies the source
4549<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4550intrinsic is necessary because the <tt><a href="#int_va_start">
4551llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4552example, memory allocation.</p>
4553
4554</div>
4555
4556<!-- ======================================================================= -->
4557<div class="doc_subsection">
4558 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4559</div>
4560
4561<div class="doc_text">
4562
4563<p>
4564LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4565Collection</a> (GC) requires the implementation and generation of these
4566intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4567These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4568stack</a>, as well as garbage collector implementations that require <a
4569href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4570Front-ends for type-safe garbage collected languages should generate these
4571intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4572href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4573</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4574
4575<p>The garbage collection intrinsics only operate on objects in the generic
4576 address space (address space zero).</p>
4577
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4578</div>
4579
4580<!-- _______________________________________________________________________ -->
4581<div class="doc_subsubsection">
4582 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4583</div>
4584
4585<div class="doc_text">
4586
4587<h5>Syntax:</h5>
4588
4589<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4590 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4591</pre>
4592
4593<h5>Overview:</h5>
4594
4595<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4596the code generator, and allows some metadata to be associated with it.</p>
4597
4598<h5>Arguments:</h5>
4599
4600<p>The first argument specifies the address of a stack object that contains the
4601root pointer. The second pointer (which must be either a constant or a global
4602value address) contains the meta-data to be associated with the root.</p>
4603
4604<h5>Semantics:</h5>
4605
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4606<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]4607location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4608the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4609intrinsic may only be used in a function which <a href="#gc">specifies a GC
4610algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4611
4612</div>
4613
4614
4615<!-- _______________________________________________________________________ -->
4616<div class="doc_subsubsection">
4617 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4618</div>
4619
4620<div class="doc_text">
4621
4622<h5>Syntax:</h5>
4623
4624<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4625 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4626</pre>
4627
4628<h5>Overview:</h5>
4629
4630<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4631locations, allowing garbage collector implementations that require read
4632barriers.</p>
4633
4634<h5>Arguments:</h5>
4635
4636<p>The second argument is the address to read from, which should be an address
4637allocated from the garbage collector. The first object is a pointer to the
4638start of the referenced object, if needed by the language runtime (otherwise
4639null).</p>
4640
4641<h5>Semantics:</h5>
4642
4643<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4644instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4645garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4646may only be used in a function which <a href="#gc">specifies a GC
4647algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4648
4649</div>
4650
4651
4652<!-- _______________________________________________________________________ -->
4653<div class="doc_subsubsection">
4654 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4655</div>
4656
4657<div class="doc_text">
4658
4659<h5>Syntax:</h5>
4660
4661<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4662 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4663</pre>
4664
4665<h5>Overview:</h5>
4666
4667<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4668locations, allowing garbage collector implementations that require write
4669barriers (such as generational or reference counting collectors).</p>
4670
4671<h5>Arguments:</h5>
4672
4673<p>The first argument is the reference to store, the second is the start of the
4674object to store it to, and the third is the address of the field of Obj to
4675store to. If the runtime does not require a pointer to the object, Obj may be
4676null.</p>
4677
4678<h5>Semantics:</h5>
4679
4680<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4681instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4682garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4683may only be used in a function which <a href="#gc">specifies a GC
4684algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4685
4686</div>
4687
4688
4689
4690<!-- ======================================================================= -->
4691<div class="doc_subsection">
4692 <a name="int_codegen">Code Generator Intrinsics</a>
4693</div>
4694
4695<div class="doc_text">
4696<p>
4697These intrinsics are provided by LLVM to expose special features that may only
4698be implemented with code generator support.
4699</p>
4700
4701</div>
4702
4703<!-- _______________________________________________________________________ -->
4704<div class="doc_subsubsection">
4705 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4706</div>
4707
4708<div class="doc_text">
4709
4710<h5>Syntax:</h5>
4711<pre>
4712 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4713</pre>
4714
4715<h5>Overview:</h5>
4716
4717<p>
4718The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4719target-specific value indicating the return address of the current function
4720or one of its callers.
4721</p>
4722
4723<h5>Arguments:</h5>
4724
4725<p>
4726The argument to this intrinsic indicates which function to return the address
4727for. Zero indicates the calling function, one indicates its caller, etc. The
4728argument is <b>required</b> to be a constant integer value.
4729</p>
4730
4731<h5>Semantics:</h5>
4732
4733<p>
4734The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4735the return address of the specified call frame, or zero if it cannot be
4736identified. The value returned by this intrinsic is likely to be incorrect or 0
4737for arguments other than zero, so it should only be used for debugging purposes.
4738</p>
4739
4740<p>
4741Note that calling this intrinsic does not prevent function inlining or other
4742aggressive transformations, so the value returned may not be that of the obvious
4743source-language caller.
4744</p>
4745</div>
4746
4747
4748<!-- _______________________________________________________________________ -->
4749<div class="doc_subsubsection">
4750 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4751</div>
4752
4753<div class="doc_text">
4754
4755<h5>Syntax:</h5>
4756<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4757 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4758</pre>
4759
4760<h5>Overview:</h5>
4761
4762<p>
4763The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4764target-specific frame pointer value for the specified stack frame.
4765</p>
4766
4767<h5>Arguments:</h5>
4768
4769<p>
4770The argument to this intrinsic indicates which function to return the frame
4771pointer for. Zero indicates the calling function, one indicates its caller,
4772etc. The argument is <b>required</b> to be a constant integer value.
4773</p>
4774
4775<h5>Semantics:</h5>
4776
4777<p>
4778The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4779the frame address of the specified call frame, or zero if it cannot be
4780identified. The value returned by this intrinsic is likely to be incorrect or 0
4781for arguments other than zero, so it should only be used for debugging purposes.
4782</p>
4783
4784<p>
4785Note that calling this intrinsic does not prevent function inlining or other
4786aggressive transformations, so the value returned may not be that of the obvious
4787source-language caller.
4788</p>
4789</div>
4790
4791<!-- _______________________________________________________________________ -->
4792<div class="doc_subsubsection">
4793 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4794</div>
4795
4796<div class="doc_text">
4797
4798<h5>Syntax:</h5>
4799<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4800 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4801</pre>
4802
4803<h5>Overview:</h5>
4804
4805<p>
4806The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4807the function stack, for use with <a href="#int_stackrestore">
4808<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4809features like scoped automatic variable sized arrays in C99.
4810</p>
4811
4812<h5>Semantics:</h5>
4813
4814<p>
4815This intrinsic returns a opaque pointer value that can be passed to <a
4816href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4817<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4818<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4819state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4820practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4821that were allocated after the <tt>llvm.stacksave</tt> was executed.
4822</p>
4823
4824</div>
4825
4826<!-- _______________________________________________________________________ -->
4827<div class="doc_subsubsection">
4828 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4829</div>
4830
4831<div class="doc_text">
4832
4833<h5>Syntax:</h5>
4834<pre>
4835 declare void @llvm.stackrestore(i8 * %ptr)
4836</pre>
4837
4838<h5>Overview:</h5>
4839
4840<p>
4841The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4842the function stack to the state it was in when the corresponding <a
4843href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4844useful for implementing language features like scoped automatic variable sized
4845arrays in C99.
4846</p>
4847
4848<h5>Semantics:</h5>
4849
4850<p>
4851See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4852</p>
4853
4854</div>
4855
4856
4857<!-- _______________________________________________________________________ -->
4858<div class="doc_subsubsection">
4859 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4860</div>
4861
4862<div class="doc_text">
4863
4864<h5>Syntax:</h5>
4865<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4866 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]4867</pre>
4868
4869<h5>Overview:</h5>
4870
4871
4872<p>
4873The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4874a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4875no
4876effect on the behavior of the program but can change its performance
4877characteristics.
4878</p>
4879
4880<h5>Arguments:</h5>
4881
4882<p>
4883<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4884determining if the fetch should be for a read (0) or write (1), and
4885<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4886locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4887<tt>locality</tt> arguments must be constant integers.
4888</p>
4889
4890<h5>Semantics:</h5>
4891
4892<p>
4893This intrinsic does not modify the behavior of the program. In particular,
4894prefetches cannot trap and do not produce a value. On targets that support this
4895intrinsic, the prefetch can provide hints to the processor cache for better
4896performance.
4897</p>
4898
4899</div>
4900
4901<!-- _______________________________________________________________________ -->
4902<div class="doc_subsubsection">
4903 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4904</div>
4905
4906<div class="doc_text">
4907
4908<h5>Syntax:</h5>
4909<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4910 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4911</pre>
4912
4913<h5>Overview:</h5>
4914
4915
4916<p>
4917The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4918(PC) in a region of
4919code to simulators and other tools. The method is target specific, but it is
4920expected that the marker will use exported symbols to transmit the PC of the
4921marker.
4922The marker makes no guarantees that it will remain with any specific instruction
4923after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4924optimizations. The intended use is to be inserted after optimizations to allow
4925correlations of simulation runs.
4926</p>
4927
4928<h5>Arguments:</h5>
4929
4930<p>
4931<tt>id</tt> is a numerical id identifying the marker.
4932</p>
4933
4934<h5>Semantics:</h5>
4935
4936<p>
4937This intrinsic does not modify the behavior of the program. Backends that do not
4938support this intrinisic may ignore it.
4939</p>
4940
4941</div>
4942
4943<!-- _______________________________________________________________________ -->
4944<div class="doc_subsubsection">
4945 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4946</div>
4947
4948<div class="doc_text">
4949
4950<h5>Syntax:</h5>
4951<pre>
4952 declare i64 @llvm.readcyclecounter( )
4953</pre>
4954
4955<h5>Overview:</h5>
4956
4957
4958<p>
4959The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4960counter register (or similar low latency, high accuracy clocks) on those targets
4961that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4962As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4963should only be used for small timings.
4964</p>
4965
4966<h5>Semantics:</h5>
4967
4968<p>
4969When directly supported, reading the cycle counter should not modify any memory.
4970Implementations are allowed to either return a application specific value or a
4971system wide value. On backends without support, this is lowered to a constant 0.
4972</p>
4973
4974</div>
4975
4976<!-- ======================================================================= -->
4977<div class="doc_subsection">
4978 <a name="int_libc">Standard C Library Intrinsics</a>
4979</div>
4980
4981<div class="doc_text">
4982<p>
4983LLVM provides intrinsics for a few important standard C library functions.
4984These intrinsics allow source-language front-ends to pass information about the
4985alignment of the pointer arguments to the code generator, providing opportunity
4986for more efficient code generation.
4987</p>
4988
4989</div>
4990
4991<!-- _______________________________________________________________________ -->
4992<div class="doc_subsubsection">
4993 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4994</div>
4995
4996<div class="doc_text">
4997
4998<h5>Syntax:</h5>
4999<pre>
5000 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5001 i32 &lt;len&gt;, i32 &lt;align&gt;)
5002 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5003 i64 &lt;len&gt;, i32 &lt;align&gt;)
5004</pre>
5005
5006<h5>Overview:</h5>
5007
5008<p>
5009The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5010location to the destination location.
5011</p>
5012
5013<p>
5014Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5015intrinsics do not return a value, and takes an extra alignment argument.
5016</p>
5017
5018<h5>Arguments:</h5>
5019
5020<p>
5021The first argument is a pointer to the destination, the second is a pointer to
5022the source. The third argument is an integer argument
5023specifying the number of bytes to copy, and the fourth argument is the alignment
5024of the source and destination locations.
5025</p>
5026
5027<p>
5028If the call to this intrinisic has an alignment value that is not 0 or 1, then
5029the caller guarantees that both the source and destination pointers are aligned
5030to that boundary.
5031</p>
5032
5033<h5>Semantics:</h5>
5034
5035<p>
5036The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5037location to the destination location, which are not allowed to overlap. It
5038copies "len" bytes of memory over. If the argument is known to be aligned to
5039some boundary, this can be specified as the fourth argument, otherwise it should
5040be set to 0 or 1.
5041</p>
5042</div>
5043
5044
5045<!-- _______________________________________________________________________ -->
5046<div class="doc_subsubsection">
5047 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5048</div>
5049
5050<div class="doc_text">
5051
5052<h5>Syntax:</h5>
5053<pre>
5054 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5055 i32 &lt;len&gt;, i32 &lt;align&gt;)
5056 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5057 i64 &lt;len&gt;, i32 &lt;align&gt;)
5058</pre>
5059
5060<h5>Overview:</h5>
5061
5062<p>
5063The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5064location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5065'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5066</p>
5067
5068<p>
5069Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5070intrinsics do not return a value, and takes an extra alignment argument.
5071</p>
5072
5073<h5>Arguments:</h5>
5074
5075<p>
5076The first argument is a pointer to the destination, the second is a pointer to
5077the source. The third argument is an integer argument
5078specifying the number of bytes to copy, and the fourth argument is the alignment
5079of the source and destination locations.
5080</p>
5081
5082<p>
5083If the call to this intrinisic has an alignment value that is not 0 or 1, then
5084the caller guarantees that the source and destination pointers are aligned to
5085that boundary.
5086</p>
5087
5088<h5>Semantics:</h5>
5089
5090<p>
5091The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5092location to the destination location, which may overlap. It
5093copies "len" bytes of memory over. If the argument is known to be aligned to
5094some boundary, this can be specified as the fourth argument, otherwise it should
5095be set to 0 or 1.
5096</p>
5097</div>
5098
5099
5100<!-- _______________________________________________________________________ -->
5101<div class="doc_subsubsection">
5102 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5103</div>
5104
5105<div class="doc_text">
5106
5107<h5>Syntax:</h5>
5108<pre>
5109 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5110 i32 &lt;len&gt;, i32 &lt;align&gt;)
5111 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5112 i64 &lt;len&gt;, i32 &lt;align&gt;)
5113</pre>
5114
5115<h5>Overview:</h5>
5116
5117<p>
5118The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5119byte value.
5120</p>
5121
5122<p>
5123Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5124does not return a value, and takes an extra alignment argument.
5125</p>
5126
5127<h5>Arguments:</h5>
5128
5129<p>
5130The first argument is a pointer to the destination to fill, the second is the
5131byte value to fill it with, the third argument is an integer
5132argument specifying the number of bytes to fill, and the fourth argument is the
5133known alignment of destination location.
5134</p>
5135
5136<p>
5137If the call to this intrinisic has an alignment value that is not 0 or 1, then
5138the caller guarantees that the destination pointer is aligned to that boundary.
5139</p>
5140
5141<h5>Semantics:</h5>
5142
5143<p>
5144The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5145the
5146destination location. If the argument is known to be aligned to some boundary,
5147this can be specified as the fourth argument, otherwise it should be set to 0 or
51481.
5149</p>
5150</div>
5151
5152
5153<!-- _______________________________________________________________________ -->
5154<div class="doc_subsubsection">
5155 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5156</div>
5157
5158<div class="doc_text">
5159
5160<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5161<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]5162floating point or vector of floating point type. Not all targets support all
5163types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5164<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5165 declare float @llvm.sqrt.f32(float %Val)
5166 declare double @llvm.sqrt.f64(double %Val)
5167 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5168 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5169 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5170</pre>
5171
5172<h5>Overview:</h5>
5173
5174<p>
5175The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5176returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5177<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5178negative numbers other than -0.0 (which allows for better optimization, because
5179there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5180defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5181</p>
5182
5183<h5>Arguments:</h5>
5184
5185<p>
5186The argument and return value are floating point numbers of the same type.
5187</p>
5188
5189<h5>Semantics:</h5>
5190
5191<p>
5192This function returns the sqrt of the specified operand if it is a nonnegative
5193floating point number.
5194</p>
5195</div>
5196
5197<!-- _______________________________________________________________________ -->
5198<div class="doc_subsubsection">
5199 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5200</div>
5201
5202<div class="doc_text">
5203
5204<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5205<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]5206floating point or vector of floating point type. Not all targets support all
5207types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5208<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5209 declare float @llvm.powi.f32(float %Val, i32 %power)
5210 declare double @llvm.powi.f64(double %Val, i32 %power)
5211 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5212 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5213 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5214</pre>
5215
5216<h5>Overview:</h5>
5217
5218<p>
5219The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5220specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5221multiplications is not defined. When a vector of floating point type is
5222used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5223</p>
5224
5225<h5>Arguments:</h5>
5226
5227<p>
5228The second argument is an integer power, and the first is a value to raise to
5229that power.
5230</p>
5231
5232<h5>Semantics:</h5>
5233
5234<p>
5235This function returns the first value raised to the second power with an
5236unspecified sequence of rounding operations.</p>
5237</div>
5238
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5239<!-- _______________________________________________________________________ -->
5240<div class="doc_subsubsection">
5241 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5242</div>
5243
5244<div class="doc_text">
5245
5246<h5>Syntax:</h5>
5247<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5248floating point or vector of floating point type. Not all targets support all
5249types however.
5250<pre>
5251 declare float @llvm.sin.f32(float %Val)
5252 declare double @llvm.sin.f64(double %Val)
5253 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5254 declare fp128 @llvm.sin.f128(fp128 %Val)
5255 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5256</pre>
5257
5258<h5>Overview:</h5>
5259
5260<p>
5261The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5262</p>
5263
5264<h5>Arguments:</h5>
5265
5266<p>
5267The argument and return value are floating point numbers of the same type.
5268</p>
5269
5270<h5>Semantics:</h5>
5271
5272<p>
5273This function returns the sine of the specified operand, returning the
5274same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5275conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5276</div>
5277
5278<!-- _______________________________________________________________________ -->
5279<div class="doc_subsubsection">
5280 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5281</div>
5282
5283<div class="doc_text">
5284
5285<h5>Syntax:</h5>
5286<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5287floating point or vector of floating point type. Not all targets support all
5288types however.
5289<pre>
5290 declare float @llvm.cos.f32(float %Val)
5291 declare double @llvm.cos.f64(double %Val)
5292 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5293 declare fp128 @llvm.cos.f128(fp128 %Val)
5294 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5295</pre>
5296
5297<h5>Overview:</h5>
5298
5299<p>
5300The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5301</p>
5302
5303<h5>Arguments:</h5>
5304
5305<p>
5306The argument and return value are floating point numbers of the same type.
5307</p>
5308
5309<h5>Semantics:</h5>
5310
5311<p>
5312This function returns the cosine of the specified operand, returning the
5313same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5314conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5315</div>
5316
5317<!-- _______________________________________________________________________ -->
5318<div class="doc_subsubsection">
5319 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5320</div>
5321
5322<div class="doc_text">
5323
5324<h5>Syntax:</h5>
5325<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5326floating point or vector of floating point type. Not all targets support all
5327types however.
5328<pre>
5329 declare float @llvm.pow.f32(float %Val, float %Power)
5330 declare double @llvm.pow.f64(double %Val, double %Power)
5331 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5332 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5333 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5334</pre>
5335
5336<h5>Overview:</h5>
5337
5338<p>
5339The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5340specified (positive or negative) power.
5341</p>
5342
5343<h5>Arguments:</h5>
5344
5345<p>
5346The second argument is a floating point power, and the first is a value to
5347raise to that power.
5348</p>
5349
5350<h5>Semantics:</h5>
5351
5352<p>
5353This function returns the first value raised to the second power,
5354returning the
5355same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5356conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5357</div>
5358
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5359
5360<!-- ======================================================================= -->
5361<div class="doc_subsection">
5362 <a name="int_manip">Bit Manipulation Intrinsics</a>
5363</div>
5364
5365<div class="doc_text">
5366<p>
5367LLVM provides intrinsics for a few important bit manipulation operations.
5368These allow efficient code generation for some algorithms.
5369</p>
5370
5371</div>
5372
5373<!-- _______________________________________________________________________ -->
5374<div class="doc_subsubsection">
5375 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5376</div>
5377
5378<div class="doc_text">
5379
5380<h5>Syntax:</h5>
5381<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5382type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5383<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5384 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5385 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5386 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5387</pre>
5388
5389<h5>Overview:</h5>
5390
5391<p>
5392The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5393values with an even number of bytes (positive multiple of 16 bits). These are
5394useful for performing operations on data that is not in the target's native
5395byte order.
5396</p>
5397
5398<h5>Semantics:</h5>
5399
5400<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5401The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5402and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5403intrinsic returns an i32 value that has the four bytes of the input i32
5404swapped, 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]5405i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5406<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5407additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5408</p>
5409
5410</div>
5411
5412<!-- _______________________________________________________________________ -->
5413<div class="doc_subsubsection">
5414 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5415</div>
5416
5417<div class="doc_text">
5418
5419<h5>Syntax:</h5>
5420<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5421width. Not all targets support all bit widths however.
5422<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5423 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
5424 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5425 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5426 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5427 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5428</pre>
5429
5430<h5>Overview:</h5>
5431
5432<p>
5433The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5434value.
5435</p>
5436
5437<h5>Arguments:</h5>
5438
5439<p>
5440The only argument is the value to be counted. The argument may be of any
5441integer type. The return type must match the argument type.
5442</p>
5443
5444<h5>Semantics:</h5>
5445
5446<p>
5447The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5448</p>
5449</div>
5450
5451<!-- _______________________________________________________________________ -->
5452<div class="doc_subsubsection">
5453 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5454</div>
5455
5456<div class="doc_text">
5457
5458<h5>Syntax:</h5>
5459<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5460integer bit width. Not all targets support all bit widths however.
5461<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5462 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5463 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5464 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5465 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5466 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5467</pre>
5468
5469<h5>Overview:</h5>
5470
5471<p>
5472The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5473leading zeros in a variable.
5474</p>
5475
5476<h5>Arguments:</h5>
5477
5478<p>
5479The only argument is the value to be counted. The argument may be of any
5480integer type. The return type must match the argument type.
5481</p>
5482
5483<h5>Semantics:</h5>
5484
5485<p>
5486The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5487in a variable. If the src == 0 then the result is the size in bits of the type
5488of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5489</p>
5490</div>
5491
5492
5493
5494<!-- _______________________________________________________________________ -->
5495<div class="doc_subsubsection">
5496 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5497</div>
5498
5499<div class="doc_text">
5500
5501<h5>Syntax:</h5>
5502<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5503integer bit width. Not all targets support all bit widths however.
5504<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5505 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5506 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5507 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5508 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5509 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5510</pre>
5511
5512<h5>Overview:</h5>
5513
5514<p>
5515The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5516trailing zeros.
5517</p>
5518
5519<h5>Arguments:</h5>
5520
5521<p>
5522The only argument is the value to be counted. The argument may be of any
5523integer type. The return type must match the argument type.
5524</p>
5525
5526<h5>Semantics:</h5>
5527
5528<p>
5529The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5530in a variable. If the src == 0 then the result is the size in bits of the type
5531of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5532</p>
5533</div>
5534
5535<!-- _______________________________________________________________________ -->
5536<div class="doc_subsubsection">
5537 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5538</div>
5539
5540<div class="doc_text">
5541
5542<h5>Syntax:</h5>
5543<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5544on any integer bit width.
5545<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5546 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5547 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5548</pre>
5549
5550<h5>Overview:</h5>
5551<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5552range of bits from an integer value and returns them in the same bit width as
5553the original value.</p>
5554
5555<h5>Arguments:</h5>
5556<p>The first argument, <tt>%val</tt> and the result may be integer types of
5557any bit width but they must have the same bit width. The second and third
5558arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5559
5560<h5>Semantics:</h5>
5561<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5562of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5563<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5564operates in forward mode.</p>
5565<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5566right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5567only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5568<ol>
5569 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5570 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5571 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5572 to determine the number of bits to retain.</li>
5573 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5574 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5575</ol>
5576<p>In reverse mode, a similar computation is made except that the bits are
5577returned in the reverse order. So, for example, if <tt>X</tt> has the value
5578<tt>i16 0x0ACF (101011001111)</tt> and we apply
5579<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5580<tt>i16 0x0026 (000000100110)</tt>.</p>
5581</div>
5582
5583<div class="doc_subsubsection">
5584 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5585</div>
5586
5587<div class="doc_text">
5588
5589<h5>Syntax:</h5>
5590<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5591on any integer bit width.
5592<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5593 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5594 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]5595</pre>
5596
5597<h5>Overview:</h5>
5598<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5599of bits in an integer value with another integer value. It returns the integer
5600with the replaced bits.</p>
5601
5602<h5>Arguments:</h5>
5603<p>The first argument, <tt>%val</tt> and the result may be integer types of
5604any bit width but they must have the same bit width. <tt>%val</tt> is the value
5605whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5606integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5607type since they specify only a bit index.</p>
5608
5609<h5>Semantics:</h5>
5610<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5611of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5612<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5613operates in forward mode.</p>
5614<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5615truncating it down to the size of the replacement area or zero extending it
5616up to that size.</p>
5617<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5618are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5619in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5620to the <tt>%hi</tt>th bit.
5621<p>In reverse mode, a similar computation is made except that the bits are
5622reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5623<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5624<h5>Examples:</h5>
5625<pre>
5626 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5627 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5628 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5629 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5630 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5631</pre>
5632</div>
5633
5634<!-- ======================================================================= -->
5635<div class="doc_subsection">
5636 <a name="int_debugger">Debugger Intrinsics</a>
5637</div>
5638
5639<div class="doc_text">
5640<p>
5641The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5642are described in the <a
5643href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5644Debugging</a> document.
5645</p>
5646</div>
5647
5648
5649<!-- ======================================================================= -->
5650<div class="doc_subsection">
5651 <a name="int_eh">Exception Handling Intrinsics</a>
5652</div>
5653
5654<div class="doc_text">
5655<p> The LLVM exception handling intrinsics (which all start with
5656<tt>llvm.eh.</tt> prefix), are described in the <a
5657href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5658Handling</a> document. </p>
5659</div>
5660
5661<!-- ======================================================================= -->
5662<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5663 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5664</div>
5665
5666<div class="doc_text">
5667<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5668 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5669 the <tt>nest</tt> attribute, from a function. The result is a callable
5670 function pointer lacking the nest parameter - the caller does not need
5671 to provide a value for it. Instead, the value to use is stored in
5672 advance in a "trampoline", a block of memory usually allocated
5673 on the stack, which also contains code to splice the nest value into the
5674 argument list. This is used to implement the GCC nested function address
5675 extension.
5676</p>
5677<p>
5678 For example, if the function is
5679 <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]5680 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]5681<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5682 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5683 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5684 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5685 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5686</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5687 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5688 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5689</div>
5690
5691<!-- _______________________________________________________________________ -->
5692<div class="doc_subsubsection">
5693 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5694</div>
5695<div class="doc_text">
5696<h5>Syntax:</h5>
5697<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5698declare 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]5699</pre>
5700<h5>Overview:</h5>
5701<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5702 This fills the memory pointed to by <tt>tramp</tt> with code
5703 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5704</p>
5705<h5>Arguments:</h5>
5706<p>
5707 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5708 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5709 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5710 intrinsic. Note that the size and the alignment are target-specific - LLVM
5711 currently provides no portable way of determining them, so a front-end that
5712 generates this intrinsic needs to have some target-specific knowledge.
5713 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]5714</p>
5715<h5>Semantics:</h5>
5716<p>
5717 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5718 dependent code, turning it into a function. A pointer to this function is
5719 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5720 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5721 before being called. The new function's signature is the same as that of
5722 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5723 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5724 of pointer type. Calling the new function is equivalent to calling
5725 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5726 missing <tt>nest</tt> argument. If, after calling
5727 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5728 modified, then the effect of any later call to the returned function pointer is
5729 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5730</p>
5731</div>
5732
5733<!-- ======================================================================= -->
5734<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5735 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5736</div>
5737
5738<div class="doc_text">
5739<p>
5740 These intrinsic functions expand the "universal IR" of LLVM to represent
5741 hardware constructs for atomic operations and memory synchronization. This
5742 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5743 is aimed at a low enough level to allow any programming models or APIs
5744 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5745 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5746 hardware behavior. Just as hardware provides a "universal IR" for source
5747 languages, it also provides a starting point for developing a "universal"
5748 atomic operation and synchronization IR.
5749</p>
5750<p>
5751 These do <em>not</em> form an API such as high-level threading libraries,
5752 software transaction memory systems, atomic primitives, and intrinsic
5753 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5754 application libraries. The hardware interface provided by LLVM should allow
5755 a clean implementation of all of these APIs and parallel programming models.
5756 No one model or paradigm should be selected above others unless the hardware
5757 itself ubiquitously does so.
5758
5759</p>
5760</div>
5761
5762<!-- _______________________________________________________________________ -->
5763<div class="doc_subsubsection">
5764 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5765</div>
5766<div class="doc_text">
5767<h5>Syntax:</h5>
5768<pre>
5769declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5770i1 &lt;device&gt; )
5771
5772</pre>
5773<h5>Overview:</h5>
5774<p>
5775 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5776 specific pairs of memory access types.
5777</p>
5778<h5>Arguments:</h5>
5779<p>
5780 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5781 The first four arguments enables a specific barrier as listed below. The fith
5782 argument specifies that the barrier applies to io or device or uncached memory.
5783
5784</p>
5785 <ul>
5786 <li><tt>ll</tt>: load-load barrier</li>
5787 <li><tt>ls</tt>: load-store barrier</li>
5788 <li><tt>sl</tt>: store-load barrier</li>
5789 <li><tt>ss</tt>: store-store barrier</li>
5790 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5791 </ul>
5792<h5>Semantics:</h5>
5793<p>
5794 This intrinsic causes the system to enforce some ordering constraints upon
5795 the loads and stores of the program. This barrier does not indicate
5796 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5797 which they occur. For any of the specified pairs of load and store operations
5798 (f.ex. load-load, or store-load), all of the first operations preceding the
5799 barrier will complete before any of the second operations succeeding the
5800 barrier begin. Specifically the semantics for each pairing is as follows:
5801</p>
5802 <ul>
5803 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5804 after the barrier begins.</li>
5805
5806 <li><tt>ls</tt>: All loads before the barrier must complete before any
5807 store after the barrier begins.</li>
5808 <li><tt>ss</tt>: All stores before the barrier must complete before any
5809 store after the barrier begins.</li>
5810 <li><tt>sl</tt>: All stores before the barrier must complete before any
5811 load after the barrier begins.</li>
5812 </ul>
5813<p>
5814 These semantics are applied with a logical "and" behavior when more than one
5815 is enabled in a single memory barrier intrinsic.
5816</p>
5817<p>
5818 Backends may implement stronger barriers than those requested when they do not
5819 support as fine grained a barrier as requested. Some architectures do not
5820 need all types of barriers and on such architectures, these become noops.
5821</p>
5822<h5>Example:</h5>
5823<pre>
5824%ptr = malloc i32
5825 store i32 4, %ptr
5826
5827%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5828 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5829 <i>; guarantee the above finishes</i>
5830 store i32 8, %ptr <i>; before this begins</i>
5831</pre>
5832</div>
5833
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5834<!-- _______________________________________________________________________ -->
5835<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5836 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5837</div>
5838<div class="doc_text">
5839<h5>Syntax:</h5>
5840<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5841 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
5842 any integer bit width and for different address spaces. Not all targets
5843 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5844
5845<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5846declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
5847declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
5848declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
5849declare 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]5850
5851</pre>
5852<h5>Overview:</h5>
5853<p>
5854 This loads a value in memory and compares it to a given value. If they are
5855 equal, it stores a new value into the memory.
5856</p>
5857<h5>Arguments:</h5>
5858<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5859 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5860 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5861 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5862 this integer type. While any bit width integer may be used, targets may only
5863 lower representations they support in hardware.
5864
5865</p>
5866<h5>Semantics:</h5>
5867<p>
5868 This entire intrinsic must be executed atomically. It first loads the value
5869 in memory pointed to by <tt>ptr</tt> and compares it with the value
5870 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5871 loaded value is yielded in all cases. This provides the equivalent of an
5872 atomic compare-and-swap operation within the SSA framework.
5873</p>
5874<h5>Examples:</h5>
5875
5876<pre>
5877%ptr = malloc i32
5878 store i32 4, %ptr
5879
5880%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5881%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5882 <i>; yields {i32}:result1 = 4</i>
5883%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5884%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5885
5886%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5887%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5888 <i>; yields {i32}:result2 = 8</i>
5889%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5890
5891%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5892</pre>
5893</div>
5894
5895<!-- _______________________________________________________________________ -->
5896<div class="doc_subsubsection">
5897 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5898</div>
5899<div class="doc_text">
5900<h5>Syntax:</h5>
5901
5902<p>
5903 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5904 integer bit width. Not all targets support all bit widths however.</p>
5905<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5906declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
5907declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
5908declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
5909declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5910
5911</pre>
5912<h5>Overview:</h5>
5913<p>
5914 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
5915 the value from memory. It then stores the value in <tt>val</tt> in the memory
5916 at <tt>ptr</tt>.
5917</p>
5918<h5>Arguments:</h5>
5919
5920<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5921 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5922 <tt>val</tt> argument and the result must be integers of the same bit width.
5923 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
5924 integer type. The targets may only lower integer representations they
5925 support.
5926</p>
5927<h5>Semantics:</h5>
5928<p>
5929 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
5930 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
5931 equivalent of an atomic swap operation within the SSA framework.
5932
5933</p>
5934<h5>Examples:</h5>
5935<pre>
5936%ptr = malloc i32
5937 store i32 4, %ptr
5938
5939%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5940%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5941 <i>; yields {i32}:result1 = 4</i>
5942%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5943%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5944
5945%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5946%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5947 <i>; yields {i32}:result2 = 8</i>
5948
5949%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
5950%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
5951</pre>
5952</div>
5953
5954<!-- _______________________________________________________________________ -->
5955<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5956 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5957
5958</div>
5959<div class="doc_text">
5960<h5>Syntax:</h5>
5961<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5962 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]5963 integer bit width. Not all targets support all bit widths however.</p>
5964<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5965declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
5966declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
5967declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
5968declare 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]5969
5970</pre>
5971<h5>Overview:</h5>
5972<p>
5973 This intrinsic adds <tt>delta</tt> to the value stored in memory at
5974 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
5975</p>
5976<h5>Arguments:</h5>
5977<p>
5978
5979 The intrinsic takes two arguments, the first a pointer to an integer value
5980 and the second an integer value. The result is also an integer value. These
5981 integer types can have any bit width, but they must all have the same bit
5982 width. The targets may only lower integer representations they support.
5983</p>
5984<h5>Semantics:</h5>
5985<p>
5986 This intrinsic does a series of operations atomically. It first loads the
5987 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
5988 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
5989</p>
5990
5991<h5>Examples:</h5>
5992<pre>
5993%ptr = malloc i32
5994 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5995%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5996 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5997%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5998 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5999%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6000 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6001%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6002</pre>
6003</div>
6004
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6005<!-- _______________________________________________________________________ -->
6006<div class="doc_subsubsection">
6007 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6008
6009</div>
6010<div class="doc_text">
6011<h5>Syntax:</h5>
6012<p>
6013 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]6014 any integer bit width and for different address spaces. Not all targets
6015 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6016<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6017declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6018declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6019declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6020declare 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]6021
6022</pre>
6023<h5>Overview:</h5>
6024<p>
6025 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6026 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6027</p>
6028<h5>Arguments:</h5>
6029<p>
6030
6031 The intrinsic takes two arguments, the first a pointer to an integer value
6032 and the second an integer value. The result is also an integer value. These
6033 integer types can have any bit width, but they must all have the same bit
6034 width. The targets may only lower integer representations they support.
6035</p>
6036<h5>Semantics:</h5>
6037<p>
6038 This intrinsic does a series of operations atomically. It first loads the
6039 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6040 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6041</p>
6042
6043<h5>Examples:</h5>
6044<pre>
6045%ptr = malloc i32
6046 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6047%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6048 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6049%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6050 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6051%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6052 <i>; yields {i32}:result3 = 2</i>
6053%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6054</pre>
6055</div>
6056
6057<!-- _______________________________________________________________________ -->
6058<div class="doc_subsubsection">
6059 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6060 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6061 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6062 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6063
6064</div>
6065<div class="doc_text">
6066<h5>Syntax:</h5>
6067<p>
6068 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6069 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6070 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6071 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6072<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6073declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6074declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6075declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6076declare 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]6077
6078</pre>
6079
6080<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6081declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6082declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6083declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6084declare 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]6085
6086</pre>
6087
6088<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6089declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6090declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6091declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6092declare 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]6093
6094</pre>
6095
6096<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6097declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6098declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6099declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6100declare 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]6101
6102</pre>
6103<h5>Overview:</h5>
6104<p>
6105 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6106 the value stored in memory at <tt>ptr</tt>. It yields the original value
6107 at <tt>ptr</tt>.
6108</p>
6109<h5>Arguments:</h5>
6110<p>
6111
6112 These intrinsics take two arguments, the first a pointer to an integer value
6113 and the second an integer value. The result is also an integer value. These
6114 integer types can have any bit width, but they must all have the same bit
6115 width. The targets may only lower integer representations they support.
6116</p>
6117<h5>Semantics:</h5>
6118<p>
6119 These intrinsics does a series of operations atomically. They first load the
6120 value stored at <tt>ptr</tt>. They then do the bitwise operation
6121 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6122 value stored at <tt>ptr</tt>.
6123</p>
6124
6125<h5>Examples:</h5>
6126<pre>
6127%ptr = malloc i32
6128 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6129%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6130 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6131%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6132 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6133%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6134 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6135%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6136 <i>; yields {i32}:result3 = FF</i>
6137%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6138</pre>
6139</div>
6140
6141
6142<!-- _______________________________________________________________________ -->
6143<div class="doc_subsubsection">
6144 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6145 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6146 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6147 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6148
6149</div>
6150<div class="doc_text">
6151<h5>Syntax:</h5>
6152<p>
6153 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6154 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6155 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6156 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6157 support all bit widths however.</p>
6158<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6159declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6160declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6161declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6162declare 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]6163
6164</pre>
6165
6166<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6167declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6168declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6169declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6170declare 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]6171
6172</pre>
6173
6174<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6175declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6176declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6177declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6178declare 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]6179
6180</pre>
6181
6182<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6183declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6184declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6185declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6186declare 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]6187
6188</pre>
6189<h5>Overview:</h5>
6190<p>
6191 These intrinsics takes the signed or unsigned minimum or maximum of
6192 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6193 original value at <tt>ptr</tt>.
6194</p>
6195<h5>Arguments:</h5>
6196<p>
6197
6198 These intrinsics take two arguments, the first a pointer to an integer value
6199 and the second an integer value. The result is also an integer value. These
6200 integer types can have any bit width, but they must all have the same bit
6201 width. The targets may only lower integer representations they support.
6202</p>
6203<h5>Semantics:</h5>
6204<p>
6205 These intrinsics does a series of operations atomically. They first load the
6206 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6207 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6208 the original value stored at <tt>ptr</tt>.
6209</p>
6210
6211<h5>Examples:</h5>
6212<pre>
6213%ptr = malloc i32
6214 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6215%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6216 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6217%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6218 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6219%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6220 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6221%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6222 <i>; yields {i32}:result3 = 8</i>
6223%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6224</pre>
6225</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6226
6227<!-- ======================================================================= -->
6228<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6229 <a name="int_general">General Intrinsics</a>
6230</div>
6231
6232<div class="doc_text">
6233<p> This class of intrinsics is designed to be generic and has
6234no specific purpose. </p>
6235</div>
6236
6237<!-- _______________________________________________________________________ -->
6238<div class="doc_subsubsection">
6239 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6240</div>
6241
6242<div class="doc_text">
6243
6244<h5>Syntax:</h5>
6245<pre>
6246 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6247</pre>
6248
6249<h5>Overview:</h5>
6250
6251<p>
6252The '<tt>llvm.var.annotation</tt>' intrinsic
6253</p>
6254
6255<h5>Arguments:</h5>
6256
6257<p>
6258The first argument is a pointer to a value, the second is a pointer to a
6259global string, the third is a pointer to a global string which is the source
6260file name, and the last argument is the line number.
6261</p>
6262
6263<h5>Semantics:</h5>
6264
6265<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6266This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6267This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6268annotations. These have no other defined use, they are ignored by code
6269generation and optimization.
6270</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6271</div>
6272
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6273<!-- _______________________________________________________________________ -->
6274<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6275 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6276</div>
6277
6278<div class="doc_text">
6279
6280<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6281<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6282any integer bit width.
6283</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6284<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6285 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6286 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6287 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6288 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6289 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]6290</pre>
6291
6292<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6293
6294<p>
6295The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6296</p>
6297
6298<h5>Arguments:</h5>
6299
6300<p>
6301The first argument is an integer value (result of some expression),
6302the second is a pointer to a global string, the third is a pointer to a global
6303string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6304It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6305</p>
6306
6307<h5>Semantics:</h5>
6308
6309<p>
6310This intrinsic allows annotations to be put on arbitrary expressions
6311with arbitrary strings. This can be useful for special purpose optimizations
6312that want to look for these annotations. These have no other defined use, they
6313are ignored by code generation and optimization.
6314</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6315
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6316<!-- _______________________________________________________________________ -->
6317<div class="doc_subsubsection">
6318 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6319</div>
6320
6321<div class="doc_text">
6322
6323<h5>Syntax:</h5>
6324<pre>
6325 declare void @llvm.trap()
6326</pre>
6327
6328<h5>Overview:</h5>
6329
6330<p>
6331The '<tt>llvm.trap</tt>' intrinsic
6332</p>
6333
6334<h5>Arguments:</h5>
6335
6336<p>
6337None
6338</p>
6339
6340<h5>Semantics:</h5>
6341
6342<p>
6343This intrinsics is lowered to the target dependent trap instruction. If the
6344target does not have a trap instruction, this intrinsic will be lowered to the
6345call of the abort() function.
6346</p>
6347</div>
6348
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6349<!-- *********************************************************************** -->
6350<hr>
6351<address>
6352 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6353 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
6354 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6355 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6356
6357 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6358 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6359 Last modified: $Date$
6360</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6361
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6362</body>
6363</html>