Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 4dcf810bd693551137f97e79a155bf470ae08bf0 / . / docs / LangRef.html
blob: 457a56ab0cba6e3321c5e1b266508ef6c1ee7186 [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
12
13<body>
14
15<div class="doc_title"> LLVM Language Reference Manual </div>
16<ol>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]25 <li><a href="#namedtypes">Named Types</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]26 <li><a href="#globalvars">Global Variables</a></li>
27 <li><a href="#functionstructure">Functions</a></li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]28 <li><a href="#aliasstructure">Aliases</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]29 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]30 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]31 <li><a href="#gc">Garbage Collector Names</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]32 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
33 <li><a href="#datalayout">Data Layout</a></li>
34 </ol>
35 </li>
36 <li><a href="#typesystem">Type System</a>
37 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]38 <li><a href="#t_classifications">Type Classifications</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]39 <li><a href="#t_primitive">Primitive Types</a>
40 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]41 <li><a href="#t_floating">Floating Point Types</a></li>
42 <li><a href="#t_void">Void Type</a></li>
43 <li><a href="#t_label">Label Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]44 </ol>
45 </li>
46 <li><a href="#t_derived">Derived Types</a>
47 <ol>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]48 <li><a href="#t_integer">Integer Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]49 <li><a href="#t_array">Array Type</a></li>
50 <li><a href="#t_function">Function Type</a></li>
51 <li><a href="#t_pointer">Pointer Type</a></li>
52 <li><a href="#t_struct">Structure Type</a></li>
53 <li><a href="#t_pstruct">Packed Structure Type</a></li>
54 <li><a href="#t_vector">Vector Type</a></li>
55 <li><a href="#t_opaque">Opaque Type</a></li>
56 </ol>
57 </li>
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]58 <li><a href="#t_uprefs">Type Up-references</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]59 </ol>
60 </li>
61 <li><a href="#constants">Constants</a>
62 <ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]63 <li><a href="#simpleconstants">Simple Constants</a></li>
Chris Lattner97063852009-02-28 18:32:25 +0000[diff] [blame]64 <li><a href="#complexconstants">Complex Constants</a></li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]65 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
66 <li><a href="#undefvalues">Undefined Values</a></li>
67 <li><a href="#constantexprs">Constant Expressions</a></li>
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame^]68 <li><a href="#metadata">Embedded Metadata</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]69 </ol>
70 </li>
71 <li><a href="#othervalues">Other Values</a>
72 <ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]73 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]74 </ol>
75 </li>
76 <li><a href="#instref">Instruction Reference</a>
77 <ol>
78 <li><a href="#terminators">Terminator Instructions</a>
79 <ol>
80 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
81 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
82 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
83 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
84 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
85 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
86 </ol>
87 </li>
88 <li><a href="#binaryops">Binary Operations</a>
89 <ol>
90 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
91 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
92 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
93 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
94 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
95 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
96 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
97 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
98 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
99 </ol>
100 </li>
101 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
102 <ol>
103 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
104 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
105 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
106 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
107 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
108 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
109 </ol>
110 </li>
111 <li><a href="#vectorops">Vector Operations</a>
112 <ol>
113 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
114 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
115 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
116 </ol>
117 </li>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]118 <li><a href="#aggregateops">Aggregate Operations</a>
119 <ol>
120 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
121 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
122 </ol>
123 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]124 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
125 <ol>
126 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
127 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
128 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
129 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
130 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
131 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
132 </ol>
133 </li>
134 <li><a href="#convertops">Conversion Operations</a>
135 <ol>
136 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
137 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
138 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
140 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
141 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
142 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
143 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
144 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
145 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
146 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
147 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
148 </ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]149 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]150 <li><a href="#otherops">Other Operations</a>
151 <ol>
152 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
153 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]154 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
155 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]156 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
157 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
158 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
159 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
160 </ol>
161 </li>
162 </ol>
163 </li>
164 <li><a href="#intrinsics">Intrinsic Functions</a>
165 <ol>
166 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
167 <ol>
168 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
169 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
170 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
171 </ol>
172 </li>
173 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
174 <ol>
175 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
176 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
177 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
178 </ol>
179 </li>
180 <li><a href="#int_codegen">Code Generator Intrinsics</a>
181 <ol>
182 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
183 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
184 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
185 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
186 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
187 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
188 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
189 </ol>
190 </li>
191 <li><a href="#int_libc">Standard C Library Intrinsics</a>
192 <ol>
193 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
194 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]198 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
199 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]201 </ol>
202 </li>
203 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
204 <ol>
205 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
206 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
207 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
208 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
209 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
210 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
211 </ol>
212 </li>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]213 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
214 <ol>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]215 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
216 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
217 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
218 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
219 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingbda98b62009-02-08 23:00:09 +0000[diff] [blame]220 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]221 </ol>
222 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]223 <li><a href="#int_debugger">Debugger intrinsics</a></li>
224 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]225 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]226 <ol>
227 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]228 </ol>
229 </li>
Bill Wendling9127adb2008-11-18 22:10:53 +0000[diff] [blame]230 <li><a href="#int_atomics">Atomic intrinsics</a>
231 <ol>
232 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
233 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
234 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
235 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
236 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
237 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
238 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
239 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
240 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
241 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
242 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
243 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
244 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
245 </ol>
246 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]247 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]248 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]249 <li><a href="#int_var_annotation">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]250 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]251 <li><a href="#int_annotation">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]252 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]253 <li><a href="#int_trap">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]254 '<tt>llvm.trap</tt>' Intrinsic</a></li>
255 <li><a href="#int_stackprotector">
256 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]257 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]258 </li>
259 </ol>
260 </li>
261</ol>
262
263<div class="doc_author">
264 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
265 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
266</div>
267
268<!-- *********************************************************************** -->
269<div class="doc_section"> <a name="abstract">Abstract </a></div>
270<!-- *********************************************************************** -->
271
272<div class="doc_text">
273<p>This document is a reference manual for the LLVM assembly language.
Bill Wendlinge7846a52008-08-05 22:29:16 +0000[diff] [blame]274LLVM is a Static Single Assignment (SSA) based representation that provides
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]275type safety, low-level operations, flexibility, and the capability of
276representing 'all' high-level languages cleanly. It is the common code
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]277representation used throughout all phases of the LLVM compilation
278strategy.</p>
279</div>
280
281<!-- *********************************************************************** -->
282<div class="doc_section"> <a name="introduction">Introduction</a> </div>
283<!-- *********************************************************************** -->
284
285<div class="doc_text">
286
287<p>The LLVM code representation is designed to be used in three
288different forms: as an in-memory compiler IR, as an on-disk bitcode
289representation (suitable for fast loading by a Just-In-Time compiler),
290and as a human readable assembly language representation. This allows
291LLVM to provide a powerful intermediate representation for efficient
292compiler transformations and analysis, while providing a natural means
293to debug and visualize the transformations. The three different forms
294of LLVM are all equivalent. This document describes the human readable
295representation and notation.</p>
296
297<p>The LLVM representation aims to be light-weight and low-level
298while being expressive, typed, and extensible at the same time. It
299aims to be a "universal IR" of sorts, by being at a low enough level
300that high-level ideas may be cleanly mapped to it (similar to how
301microprocessors are "universal IR's", allowing many source languages to
302be mapped to them). By providing type information, LLVM can be used as
303the target of optimizations: for example, through pointer analysis, it
304can be proven that a C automatic variable is never accessed outside of
305the current function... allowing it to be promoted to a simple SSA
306value instead of a memory location.</p>
307
308</div>
309
310<!-- _______________________________________________________________________ -->
311<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
312
313<div class="doc_text">
314
315<p>It is important to note that this document describes 'well formed'
316LLVM assembly language. There is a difference between what the parser
317accepts and what is considered 'well formed'. For example, the
318following instruction is syntactically okay, but not well formed:</p>
319
320<div class="doc_code">
321<pre>
322%x = <a href="#i_add">add</a> i32 1, %x
323</pre>
324</div>
325
326<p>...because the definition of <tt>%x</tt> does not dominate all of
327its uses. The LLVM infrastructure provides a verification pass that may
328be used to verify that an LLVM module is well formed. This pass is
329automatically run by the parser after parsing input assembly and by
330the optimizer before it outputs bitcode. The violations pointed out
331by the verifier pass indicate bugs in transformation passes or input to
332the parser.</p>
333</div>
334
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]335<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]336
337<!-- *********************************************************************** -->
338<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
339<!-- *********************************************************************** -->
340
341<div class="doc_text">
342
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]343 <p>LLVM identifiers come in two basic types: global and local. Global
344 identifiers (functions, global variables) begin with the @ character. Local
345 identifiers (register names, types) begin with the % character. Additionally,
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]346 there are three different formats for identifiers, for different purposes:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]347
348<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]349 <li>Named values are represented as a string of characters with their prefix.
350 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
351 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]352 Identifiers which require other characters in their names can be surrounded
Daniel Dunbar73f9a772008-10-14 23:51:43 +0000[diff] [blame]353 with quotes. Special characters may be escaped using "\xx" where xx is the
354 ASCII code for the character in hexadecimal. In this way, any character can
355 be used in a name value, even quotes themselves.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]356
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]357 <li>Unnamed values are represented as an unsigned numeric value with their
358 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]359
360 <li>Constants, which are described in a <a href="#constants">section about
361 constants</a>, below.</li>
362</ol>
363
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]364<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]365don't need to worry about name clashes with reserved words, and the set of
366reserved words may be expanded in the future without penalty. Additionally,
367unnamed identifiers allow a compiler to quickly come up with a temporary
368variable without having to avoid symbol table conflicts.</p>
369
370<p>Reserved words in LLVM are very similar to reserved words in other
371languages. There are keywords for different opcodes
372('<tt><a href="#i_add">add</a></tt>',
373 '<tt><a href="#i_bitcast">bitcast</a></tt>',
374 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
375href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
376and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]377none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]378
379<p>Here is an example of LLVM code to multiply the integer variable
380'<tt>%X</tt>' by 8:</p>
381
382<p>The easy way:</p>
383
384<div class="doc_code">
385<pre>
386%result = <a href="#i_mul">mul</a> i32 %X, 8
387</pre>
388</div>
389
390<p>After strength reduction:</p>
391
392<div class="doc_code">
393<pre>
394%result = <a href="#i_shl">shl</a> i32 %X, i8 3
395</pre>
396</div>
397
398<p>And the hard way:</p>
399
400<div class="doc_code">
401<pre>
402<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
403<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
404%result = <a href="#i_add">add</a> i32 %1, %1
405</pre>
406</div>
407
408<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
409important lexical features of LLVM:</p>
410
411<ol>
412
413 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
414 line.</li>
415
416 <li>Unnamed temporaries are created when the result of a computation is not
417 assigned to a named value.</li>
418
419 <li>Unnamed temporaries are numbered sequentially</li>
420
421</ol>
422
423<p>...and it also shows a convention that we follow in this document. When
424demonstrating instructions, we will follow an instruction with a comment that
425defines the type and name of value produced. Comments are shown in italic
426text.</p>
427
428</div>
429
430<!-- *********************************************************************** -->
431<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
432<!-- *********************************************************************** -->
433
434<!-- ======================================================================= -->
435<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
436</div>
437
438<div class="doc_text">
439
440<p>LLVM programs are composed of "Module"s, each of which is a
441translation unit of the input programs. Each module consists of
442functions, global variables, and symbol table entries. Modules may be
443combined together with the LLVM linker, which merges function (and
444global variable) definitions, resolves forward declarations, and merges
445symbol table entries. Here is an example of the "hello world" module:</p>
446
447<div class="doc_code">
448<pre><i>; Declare the string constant as a global constant...</i>
449<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
450 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
451
452<i>; External declaration of the puts function</i>
453<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
454
455<i>; Definition of main function</i>
456define i32 @main() { <i>; i32()* </i>
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]457 <i>; Convert [13 x i8]* to i8 *...</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]458 %cast210 = <a
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]459 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]460
461 <i>; Call puts function to write out the string to stdout...</i>
462 <a
463 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
464 <a
465 href="#i_ret">ret</a> i32 0<br>}<br>
466</pre>
467</div>
468
469<p>This example is made up of a <a href="#globalvars">global variable</a>
470named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
471function, and a <a href="#functionstructure">function definition</a>
472for "<tt>main</tt>".</p>
473
474<p>In general, a module is made up of a list of global values,
475where both functions and global variables are global values. Global values are
476represented by a pointer to a memory location (in this case, a pointer to an
477array of char, and a pointer to a function), and have one of the following <a
478href="#linkage">linkage types</a>.</p>
479
480</div>
481
482<!-- ======================================================================= -->
483<div class="doc_subsection">
484 <a name="linkage">Linkage Types</a>
485</div>
486
487<div class="doc_text">
488
489<p>
490All Global Variables and Functions have one of the following types of linkage:
491</p>
492
493<dl>
494
Rafael Espindolaa168fc92009-01-15 20:18:42 +0000[diff] [blame]495 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
496
497 <dd>Global values with private linkage are only directly accessible by
498 objects in the current module. In particular, linking code into a module with
499 an private global value may cause the private to be renamed as necessary to
500 avoid collisions. Because the symbol is private to the module, all
501 references can be updated. This doesn't show up in any symbol table in the
502 object file.
503 </dd>
504
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]505 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]506
Duncan Sandsa75223a2009-01-16 09:29:46 +0000[diff] [blame]507 <dd> Similar to private, but the value shows as a local symbol (STB_LOCAL in
Rafael Espindolaa168fc92009-01-15 20:18:42 +0000[diff] [blame]508 the case of ELF) in the object file. This corresponds to the notion of the
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]509 '<tt>static</tt>' keyword in C.
510 </dd>
511
512 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
513
514 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
515 the same name when linkage occurs. This is typically used to implement
516 inline functions, templates, or other code which must be generated in each
517 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
518 allowed to be discarded.
519 </dd>
520
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]521 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
522
523 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
524 linkage, except that unreferenced <tt>common</tt> globals may not be
525 discarded. This is used for globals that may be emitted in multiple
526 translation units, but that are not guaranteed to be emitted into every
527 translation unit that uses them. One example of this is tentative
528 definitions in C, such as "<tt>int X;</tt>" at global scope.
529 </dd>
530
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]531 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
532
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]533 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
534 that some targets may choose to emit different assembly sequences for them
535 for target-dependent reasons. This is used for globals that are declared
536 "weak" in C source code.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]537 </dd>
538
539 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
540
541 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
542 pointer to array type. When two global variables with appending linkage are
543 linked together, the two global arrays are appended together. This is the
544 LLVM, typesafe, equivalent of having the system linker append together
545 "sections" with identical names when .o files are linked.
546 </dd>
547
548 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]549
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]550 <dd>The semantics of this linkage follow the ELF object file model: the
551 symbol is weak until linked, if not linked, the symbol becomes null instead
552 of being an undefined reference.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]553 </dd>
554
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]555 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]556 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]557 <dd>Some languages allow inequivalent globals to be merged, such as two
558 functions with different semantics. Other languages, such as <tt>C++</tt>,
559 ensure that only equivalent globals are ever merged (the "one definition
Duncan Sandsb95df792009-03-11 20:14:15 +0000[diff] [blame]560 rule" - <tt>odr</tt>). Such languages can use the <tt>linkonce_odr</tt>
561 and <tt>weak_odr</tt> linkage types to indicate that the global will only
562 be merged with equivalent globals. These linkage types are otherwise the
563 same as their non-<tt>odr</tt> versions.
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]564 </dd>
565
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]566 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
567
568 <dd>If none of the above identifiers are used, the global is externally
569 visible, meaning that it participates in linkage and can be used to resolve
570 external symbol references.
571 </dd>
572</dl>
573
574 <p>
575 The next two types of linkage are targeted for Microsoft Windows platform
576 only. They are designed to support importing (exporting) symbols from (to)
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]577 DLLs (Dynamic Link Libraries).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]578 </p>
579
580 <dl>
581 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
582
583 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
584 or variable via a global pointer to a pointer that is set up by the DLL
585 exporting the symbol. On Microsoft Windows targets, the pointer name is
Dan Gohman5ec99832009-01-12 21:35:55 +0000[diff] [blame]586 formed by combining <code>__imp_</code> and the function or variable name.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]587 </dd>
588
589 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
590
591 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
592 pointer to a pointer in a DLL, so that it can be referenced with the
593 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
Dan Gohman5ec99832009-01-12 21:35:55 +0000[diff] [blame]594 name is formed by combining <code>__imp_</code> and the function or variable
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]595 name.
596 </dd>
597
598</dl>
599
Dan Gohman4dfac702008-11-24 17:18:39 +0000[diff] [blame]600<p>For example, since the "<tt>.LC0</tt>"
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]601variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
602variable and was linked with this one, one of the two would be renamed,
603preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
604external (i.e., lacking any linkage declarations), they are accessible
605outside of the current module.</p>
606<p>It is illegal for a function <i>declaration</i>
Duncan Sands565f65d2009-03-11 08:08:06 +0000[diff] [blame]607to have any linkage type other than "externally visible", <tt>dllimport</tt>
608or <tt>extern_weak</tt>.</p>
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]609<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
610or <tt>weak_odr</tt> linkages.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]611</div>
612
613<!-- ======================================================================= -->
614<div class="doc_subsection">
615 <a name="callingconv">Calling Conventions</a>
616</div>
617
618<div class="doc_text">
619
620<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
621and <a href="#i_invoke">invokes</a> can all have an optional calling convention
622specified for the call. The calling convention of any pair of dynamic
623caller/callee must match, or the behavior of the program is undefined. The
624following calling conventions are supported by LLVM, and more may be added in
625the future:</p>
626
627<dl>
628 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
629
630 <dd>This calling convention (the default if no other calling convention is
631 specified) matches the target C calling conventions. This calling convention
632 supports varargs function calls and tolerates some mismatch in the declared
633 prototype and implemented declaration of the function (as does normal C).
634 </dd>
635
636 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
637
638 <dd>This calling convention attempts to make calls as fast as possible
639 (e.g. by passing things in registers). This calling convention allows the
640 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]641 without having to conform to an externally specified ABI (Application Binary
642 Interface). Implementations of this convention should allow arbitrary
Arnold Schwaighofer07444922008-05-14 09:17:12 +0000[diff] [blame]643 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
644 supported. This calling convention does not support varargs and requires the
645 prototype of all callees to exactly match the prototype of the function
646 definition.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]647 </dd>
648
649 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
650
651 <dd>This calling convention attempts to make code in the caller as efficient
652 as possible under the assumption that the call is not commonly executed. As
653 such, these calls often preserve all registers so that the call does not break
654 any live ranges in the caller side. This calling convention does not support
655 varargs and requires the prototype of all callees to exactly match the
656 prototype of the function definition.
657 </dd>
658
659 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
660
661 <dd>Any calling convention may be specified by number, allowing
662 target-specific calling conventions to be used. Target specific calling
663 conventions start at 64.
664 </dd>
665</dl>
666
667<p>More calling conventions can be added/defined on an as-needed basis, to
668support pascal conventions or any other well-known target-independent
669convention.</p>
670
671</div>
672
673<!-- ======================================================================= -->
674<div class="doc_subsection">
675 <a name="visibility">Visibility Styles</a>
676</div>
677
678<div class="doc_text">
679
680<p>
681All Global Variables and Functions have one of the following visibility styles:
682</p>
683
684<dl>
685 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
686
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]687 <dd>On targets that use the ELF object file format, default visibility means
688 that the declaration is visible to other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]689 modules and, in shared libraries, means that the declared entity may be
690 overridden. On Darwin, default visibility means that the declaration is
691 visible to other modules. Default visibility corresponds to "external
692 linkage" in the language.
693 </dd>
694
695 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
696
697 <dd>Two declarations of an object with hidden visibility refer to the same
698 object if they are in the same shared object. Usually, hidden visibility
699 indicates that the symbol will not be placed into the dynamic symbol table,
700 so no other module (executable or shared library) can reference it
701 directly.
702 </dd>
703
704 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
705
706 <dd>On ELF, protected visibility indicates that the symbol will be placed in
707 the dynamic symbol table, but that references within the defining module will
708 bind to the local symbol. That is, the symbol cannot be overridden by another
709 module.
710 </dd>
711</dl>
712
713</div>
714
715<!-- ======================================================================= -->
716<div class="doc_subsection">
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]717 <a name="namedtypes">Named Types</a>
718</div>
719
720<div class="doc_text">
721
722<p>LLVM IR allows you to specify name aliases for certain types. This can make
723it easier to read the IR and make the IR more condensed (particularly when
724recursive types are involved). An example of a name specification is:
725</p>
726
727<div class="doc_code">
728<pre>
729%mytype = type { %mytype*, i32 }
730</pre>
731</div>
732
733<p>You may give a name to any <a href="#typesystem">type</a> except "<a
734href="t_void">void</a>". Type name aliases may be used anywhere a type is
735expected with the syntax "%mytype".</p>
736
737<p>Note that type names are aliases for the structural type that they indicate,
738and that you can therefore specify multiple names for the same type. This often
739leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
740structural typing, the name is not part of the type. When printing out LLVM IR,
741the printer will pick <em>one name</em> to render all types of a particular
742shape. This means that if you have code where two different source types end up
743having the same LLVM type, that the dumper will sometimes print the "wrong" or
744unexpected type. This is an important design point and isn't going to
745change.</p>
746
747</div>
748
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]749<!-- ======================================================================= -->
750<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]751 <a name="globalvars">Global Variables</a>
752</div>
753
754<div class="doc_text">
755
756<p>Global variables define regions of memory allocated at compilation time
757instead of run-time. Global variables may optionally be initialized, may have
758an explicit section to be placed in, and may have an optional explicit alignment
759specified. A variable may be defined as "thread_local", which means that it
760will not be shared by threads (each thread will have a separated copy of the
761variable). A variable may be defined as a global "constant," which indicates
762that the contents of the variable will <b>never</b> be modified (enabling better
763optimization, allowing the global data to be placed in the read-only section of
764an executable, etc). Note that variables that need runtime initialization
765cannot be marked "constant" as there is a store to the variable.</p>
766
767<p>
768LLVM explicitly allows <em>declarations</em> of global variables to be marked
769constant, even if the final definition of the global is not. This capability
770can be used to enable slightly better optimization of the program, but requires
771the language definition to guarantee that optimizations based on the
772'constantness' are valid for the translation units that do not include the
773definition.
774</p>
775
776<p>As SSA values, global variables define pointer values that are in
777scope (i.e. they dominate) all basic blocks in the program. Global
778variables always define a pointer to their "content" type because they
779describe a region of memory, and all memory objects in LLVM are
780accessed through pointers.</p>
781
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]782<p>A global variable may be declared to reside in a target-specifc numbered
783address space. For targets that support them, address spaces may affect how
784optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]785the variable. The default address space is zero. The address space qualifier
786must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]787
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]788<p>LLVM allows an explicit section to be specified for globals. If the target
789supports it, it will emit globals to the section specified.</p>
790
791<p>An explicit alignment may be specified for a global. If not present, or if
792the alignment is set to zero, the alignment of the global is set by the target
793to whatever it feels convenient. If an explicit alignment is specified, the
794global is forced to have at least that much alignment. All alignments must be
795a power of 2.</p>
796
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]797<p>For example, the following defines a global in a numbered address space with
798an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]799
800<div class="doc_code">
801<pre>
Dan Gohman21ef02c2009-01-11 00:40:00 +0000[diff] [blame]802@G = addrspace(5) constant float 1.0, section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]803</pre>
804</div>
805
806</div>
807
808
809<!-- ======================================================================= -->
810<div class="doc_subsection">
811 <a name="functionstructure">Functions</a>
812</div>
813
814<div class="doc_text">
815
816<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
817an optional <a href="#linkage">linkage type</a>, an optional
818<a href="#visibility">visibility style</a>, an optional
819<a href="#callingconv">calling convention</a>, a return type, an optional
820<a href="#paramattrs">parameter attribute</a> for the return type, a function
821name, a (possibly empty) argument list (each with optional
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]822<a href="#paramattrs">parameter attributes</a>), optional
823<a href="#fnattrs">function attributes</a>, an optional section,
824an optional alignment, an optional <a href="#gc">garbage collector name</a>,
Chris Lattneree202542008-10-04 18:10:21 +0000[diff] [blame]825an opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]826
827LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
828optional <a href="#linkage">linkage type</a>, an optional
829<a href="#visibility">visibility style</a>, an optional
830<a href="#callingconv">calling convention</a>, a return type, an optional
831<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]832name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]833<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]834
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]835<p>A function definition contains a list of basic blocks, forming the CFG
836(Control Flow Graph) for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]837the function. Each basic block may optionally start with a label (giving the
838basic block a symbol table entry), contains a list of instructions, and ends
839with a <a href="#terminators">terminator</a> instruction (such as a branch or
840function return).</p>
841
842<p>The first basic block in a function is special in two ways: it is immediately
843executed on entrance to the function, and it is not allowed to have predecessor
844basic blocks (i.e. there can not be any branches to the entry block of a
845function). Because the block can have no predecessors, it also cannot have any
846<a href="#i_phi">PHI nodes</a>.</p>
847
848<p>LLVM allows an explicit section to be specified for functions. If the target
849supports it, it will emit functions to the section specified.</p>
850
851<p>An explicit alignment may be specified for a function. If not present, or if
852the alignment is set to zero, the alignment of the function is set by the target
853to whatever it feels convenient. If an explicit alignment is specified, the
854function is forced to have at least that much alignment. All alignments must be
855a power of 2.</p>
856
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]857 <h5>Syntax:</h5>
858
859<div class="doc_code">
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000[diff] [blame]860<tt>
861define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
862 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
863 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
864 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
865 [<a href="#gc">gc</a>] { ... }
866</tt>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]867</div>
868
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]869</div>
870
871
872<!-- ======================================================================= -->
873<div class="doc_subsection">
874 <a name="aliasstructure">Aliases</a>
875</div>
876<div class="doc_text">
877 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]878 function, global variable, another alias or bitcast of global value). Aliases
879 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]880 optional <a href="#visibility">visibility style</a>.</p>
881
882 <h5>Syntax:</h5>
883
884<div class="doc_code">
885<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000[diff] [blame]886@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]887</pre>
888</div>
889
890</div>
891
892
893
894<!-- ======================================================================= -->
895<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
896<div class="doc_text">
897 <p>The return type and each parameter of a function type may have a set of
898 <i>parameter attributes</i> associated with them. Parameter attributes are
899 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]900 a function. Parameter attributes are considered to be part of the function,
901 not of the function type, so functions with different parameter attributes
902 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]903
904 <p>Parameter attributes are simple keywords that follow the type specified. If
905 multiple parameter attributes are needed, they are space separated. For
906 example:</p>
907
908<div class="doc_code">
909<pre>
Nick Lewycky3022a742009-02-15 23:06:14 +0000[diff] [blame]910declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]911declare i32 @atoi(i8 zeroext)
912declare signext i8 @returns_signed_char()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]913</pre>
914</div>
915
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]916 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
917 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]918
919 <p>Currently, only the following parameter attributes are defined:</p>
920 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]921 <dt><tt>zeroext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]922 <dd>This indicates to the code generator that the parameter or return value
923 should be zero-extended to a 32-bit value by the caller (for a parameter)
924 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]925
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]926 <dt><tt>signext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]927 <dd>This indicates to the code generator that the parameter or return value
928 should be sign-extended to a 32-bit value by the caller (for a parameter)
929 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]930
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]931 <dt><tt>inreg</tt></dt>
Dale Johannesenc08a0e22008-09-25 20:47:45 +0000[diff] [blame]932 <dd>This indicates that this parameter or return value should be treated
933 in a special target-dependent fashion during while emitting code for a
934 function call or return (usually, by putting it in a register as opposed
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]935 to memory, though some targets use it to distinguish between two different
936 kinds of registers). Use of this attribute is target-specific.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]937
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]938 <dt><tt><a name="byval">byval</a></tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]939 <dd>This indicates that the pointer parameter should really be passed by
940 value to the function. The attribute implies that a hidden copy of the
941 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]942 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]943 pointer arguments. It is generally used to pass structs and arrays by
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]944 value, but is also valid on pointers to scalars. The copy is considered to
945 belong to the caller not the callee (for example,
946 <tt><a href="#readonly">readonly</a></tt> functions should not write to
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]947 <tt>byval</tt> parameters). This is not a valid attribute for return
Chris Lattnere39c5bf2009-02-05 05:42:28 +0000[diff] [blame]948 values. The byval attribute also supports specifying an alignment with the
949 align attribute. This has a target-specific effect on the code generator
950 that usually indicates a desired alignment for the synthesized stack
951 slot.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]952
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]953 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]954 <dd>This indicates that the pointer parameter specifies the address of a
955 structure that is the return value of the function in the source program.
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]956 This pointer must be guaranteed by the caller to be valid: loads and stores
957 to the structure may be assumed by the callee to not to trap. This may only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]958 be applied to the first parameter. This is not a valid attribute for
959 return values. </dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]960
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]961 <dt><tt>noalias</tt></dt>
Nick Lewyckyff384472008-11-24 03:41:24 +0000[diff] [blame]962 <dd>This indicates that the pointer does not alias any global or any other
963 parameter. The caller is responsible for ensuring that this is the
Nick Lewyckya604a942008-11-24 05:00:44 +0000[diff] [blame]964 case. On a function return value, <tt>noalias</tt> additionally indicates
965 that the pointer does not alias any other pointers visible to the
Nick Lewycky40fb6f22008-12-19 06:39:12 +0000[diff] [blame]966 caller. For further details, please see the discussion of the NoAlias
967 response in
968 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
969 analysis</a>.</dd>
970
971 <dt><tt>nocapture</tt></dt>
972 <dd>This indicates that the callee does not make any copies of the pointer
973 that outlive the callee itself. This is not a valid attribute for return
974 values.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]975
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]976 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]977 <dd>This indicates that the pointer parameter can be excised using the
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]978 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
979 attribute for return values.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]980 </dl>
981
982</div>
983
984<!-- ======================================================================= -->
985<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]986 <a name="gc">Garbage Collector Names</a>
987</div>
988
989<div class="doc_text">
990<p>Each function may specify a garbage collector name, which is simply a
991string.</p>
992
993<div class="doc_code"><pre
994>define void @f() gc "name" { ...</pre></div>
995
996<p>The compiler declares the supported values of <i>name</i>. Specifying a
997collector which will cause the compiler to alter its output in order to support
998the named garbage collection algorithm.</p>
999</div>
1000
1001<!-- ======================================================================= -->
1002<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1003 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1004</div>
1005
1006<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1007
1008<p>Function attributes are set to communicate additional information about
1009 a function. Function attributes are considered to be part of the function,
1010 not of the function type, so functions with different parameter attributes
1011 can have the same function type.</p>
1012
1013 <p>Function attributes are simple keywords that follow the type specified. If
1014 multiple attributes are needed, they are space separated. For
1015 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1016
1017<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1018<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1019define void @f() noinline { ... }
1020define void @f() alwaysinline { ... }
1021define void @f() alwaysinline optsize { ... }
1022define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1023</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1024</div>
1025
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1026<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1027<dt><tt>alwaysinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1028<dd>This attribute indicates that the inliner should attempt to inline this
1029function into callers whenever possible, ignoring any active inlining size
1030threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1031
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1032<dt><tt>noinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1033<dd>This attribute indicates that the inliner should never inline this function
Chris Lattner377f3ae2008-10-05 17:14:59 +0000[diff] [blame]1034in any situation. This attribute may not be used together with the
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1035<tt>alwaysinline</tt> attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1036
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1037<dt><tt>optsize</tt></dt>
Devang Patelc29099b2008-09-29 18:34:44 +0000[diff] [blame]1038<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1039make choices that keep the code size of this function low, and otherwise do
1040optimizations specifically to reduce code size.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1041
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1042<dt><tt>noreturn</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1043<dd>This function attribute indicates that the function never returns normally.
1044This produces undefined behavior at runtime if the function ever does
1045dynamically return.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1046
1047<dt><tt>nounwind</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1048<dd>This function attribute indicates that the function never returns with an
1049unwind or exceptional control flow. If the function does unwind, its runtime
1050behavior is undefined.</dd>
1051
1052<dt><tt>readnone</tt></dt>
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]1053<dd>This attribute indicates that the function computes its result (or the
1054exception it throws) based strictly on its arguments, without dereferencing any
1055pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1056registers, etc) visible to caller functions. It does not write through any
1057pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
1058never changes any state visible to callers.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1059
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]1060<dt><tt><a name="readonly">readonly</a></tt></dt>
1061<dd>This attribute indicates that the function does not write through any
1062pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1063or otherwise modify any state (e.g. memory, control registers, etc) visible to
1064caller functions. It may dereference pointer arguments and read state that may
1065be set in the caller. A readonly function always returns the same value (or
1066throws the same exception) when called with the same set of arguments and global
1067state.</dd>
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1068
1069<dt><tt><a name="ssp">ssp</a></tt></dt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1070<dd>This attribute indicates that the function should emit a stack smashing
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1071protector. It is in the form of a "canary"&mdash;a random value placed on the
1072stack before the local variables that's checked upon return from the function to
1073see if it has been overwritten. A heuristic is used to determine if a function
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1074needs stack protectors or not.
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1075
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +0000[diff] [blame]1076<p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1077that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1078have an <tt>ssp</tt> attribute.</p></dd>
1079
1080<dt><tt>sspreq</tt></dt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1081<dd>This attribute indicates that the function should <em>always</em> emit a
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1082stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1083function attribute.
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +0000[diff] [blame]1084
1085<p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1086function that doesn't have an <tt>sspreq</tt> attribute or which has
1087an <tt>ssp</tt> attribute, then the resulting function will have
1088an <tt>sspreq</tt> attribute.</p></dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1089</dl>
1090
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1091</div>
1092
1093<!-- ======================================================================= -->
1094<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1095 <a name="moduleasm">Module-Level Inline Assembly</a>
1096</div>
1097
1098<div class="doc_text">
1099<p>
1100Modules may contain "module-level inline asm" blocks, which corresponds to the
1101GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1102LLVM and treated as a single unit, but may be separated in the .ll file if
1103desired. The syntax is very simple:
1104</p>
1105
1106<div class="doc_code">
1107<pre>
1108module asm "inline asm code goes here"
1109module asm "more can go here"
1110</pre>
1111</div>
1112
1113<p>The strings can contain any character by escaping non-printable characters.
1114 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1115 for the number.
1116</p>
1117
1118<p>
1119 The inline asm code is simply printed to the machine code .s file when
1120 assembly code is generated.
1121</p>
1122</div>
1123
1124<!-- ======================================================================= -->
1125<div class="doc_subsection">
1126 <a name="datalayout">Data Layout</a>
1127</div>
1128
1129<div class="doc_text">
1130<p>A module may specify a target specific data layout string that specifies how
1131data is to be laid out in memory. The syntax for the data layout is simply:</p>
1132<pre> target datalayout = "<i>layout specification</i>"</pre>
1133<p>The <i>layout specification</i> consists of a list of specifications
1134separated by the minus sign character ('-'). Each specification starts with a
1135letter and may include other information after the letter to define some
1136aspect of the data layout. The specifications accepted are as follows: </p>
1137<dl>
1138 <dt><tt>E</tt></dt>
1139 <dd>Specifies that the target lays out data in big-endian form. That is, the
1140 bits with the most significance have the lowest address location.</dd>
1141 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1142 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1143 the bits with the least significance have the lowest address location.</dd>
1144 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1145 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1146 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1147 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1148 too.</dd>
1149 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1150 <dd>This specifies the alignment for an integer type of a given bit
1151 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1152 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1153 <dd>This specifies the alignment for a vector type of a given bit
1154 <i>size</i>.</dd>
1155 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1156 <dd>This specifies the alignment for a floating point type of a given bit
1157 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1158 (double).</dd>
1159 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1160 <dd>This specifies the alignment for an aggregate type of a given bit
1161 <i>size</i>.</dd>
1162</dl>
1163<p>When constructing the data layout for a given target, LLVM starts with a
1164default set of specifications which are then (possibly) overriden by the
1165specifications in the <tt>datalayout</tt> keyword. The default specifications
1166are given in this list:</p>
1167<ul>
1168 <li><tt>E</tt> - big endian</li>
1169 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1170 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1171 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1172 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1173 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1174 <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]1175 alignment of 64-bits</li>
1176 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1177 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1178 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1179 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1180 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1181</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1182<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1183following rules:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1184<ol>
1185 <li>If the type sought is an exact match for one of the specifications, that
1186 specification is used.</li>
1187 <li>If no match is found, and the type sought is an integer type, then the
1188 smallest integer type that is larger than the bitwidth of the sought type is
1189 used. If none of the specifications are larger than the bitwidth then the the
1190 largest integer type is used. For example, given the default specifications
1191 above, the i7 type will use the alignment of i8 (next largest) while both
1192 i65 and i256 will use the alignment of i64 (largest specified).</li>
1193 <li>If no match is found, and the type sought is a vector type, then the
1194 largest vector type that is smaller than the sought vector type will be used
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1195 as a fall back. This happens because &lt;128 x double&gt; can be implemented
1196 in terms of 64 &lt;2 x double&gt;, for example.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1197</ol>
1198</div>
1199
1200<!-- *********************************************************************** -->
1201<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1202<!-- *********************************************************************** -->
1203
1204<div class="doc_text">
1205
1206<p>The LLVM type system is one of the most important features of the
1207intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1208optimizations to be performed on the intermediate representation directly,
1209without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1210extra analyses on the side before the transformation. A strong type
1211system makes it easier to read the generated code and enables novel
1212analyses and transformations that are not feasible to perform on normal
1213three address code representations.</p>
1214
1215</div>
1216
1217<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1218<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1219Classifications</a> </div>
1220<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1221<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1222classifications:</p>
1223
1224<table border="1" cellspacing="0" cellpadding="4">
1225 <tbody>
1226 <tr><th>Classification</th><th>Types</th></tr>
1227 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1228 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1229 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1230 </tr>
1231 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1232 <td><a href="#t_floating">floating point</a></td>
1233 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1234 </tr>
1235 <tr>
1236 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1237 <td><a href="#t_integer">integer</a>,
1238 <a href="#t_floating">floating point</a>,
1239 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1240 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1241 <a href="#t_struct">structure</a>,
1242 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1243 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1244 </td>
1245 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1246 <tr>
1247 <td><a href="#t_primitive">primitive</a></td>
1248 <td><a href="#t_label">label</a>,
1249 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1250 <a href="#t_floating">floating point</a>.</td>
1251 </tr>
1252 <tr>
1253 <td><a href="#t_derived">derived</a></td>
1254 <td><a href="#t_integer">integer</a>,
1255 <a href="#t_array">array</a>,
1256 <a href="#t_function">function</a>,
1257 <a href="#t_pointer">pointer</a>,
1258 <a href="#t_struct">structure</a>,
1259 <a href="#t_pstruct">packed structure</a>,
1260 <a href="#t_vector">vector</a>,
1261 <a href="#t_opaque">opaque</a>.
Dan Gohman032ba852008-10-14 16:32:04 +0000[diff] [blame]1262 </td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1263 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1264 </tbody>
1265</table>
1266
1267<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1268most important. Values of these types are the only ones which can be
1269produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1270instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1271</div>
1272
1273<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1274<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1275
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1276<div class="doc_text">
1277<p>The primitive types are the fundamental building blocks of the LLVM
1278system.</p>
1279
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1280</div>
1281
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1282<!-- _______________________________________________________________________ -->
1283<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1284
1285<div class="doc_text">
1286 <table>
1287 <tbody>
1288 <tr><th>Type</th><th>Description</th></tr>
1289 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1290 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1291 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1292 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1293 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1294 </tbody>
1295 </table>
1296</div>
1297
1298<!-- _______________________________________________________________________ -->
1299<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1300
1301<div class="doc_text">
1302<h5>Overview:</h5>
1303<p>The void type does not represent any value and has no size.</p>
1304
1305<h5>Syntax:</h5>
1306
1307<pre>
1308 void
1309</pre>
1310</div>
1311
1312<!-- _______________________________________________________________________ -->
1313<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1314
1315<div class="doc_text">
1316<h5>Overview:</h5>
1317<p>The label type represents code labels.</p>
1318
1319<h5>Syntax:</h5>
1320
1321<pre>
1322 label
1323</pre>
1324</div>
1325
1326
1327<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1328<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1329
1330<div class="doc_text">
1331
1332<p>The real power in LLVM comes from the derived types in the system.
1333This is what allows a programmer to represent arrays, functions,
1334pointers, and other useful types. Note that these derived types may be
1335recursive: For example, it is possible to have a two dimensional array.</p>
1336
1337</div>
1338
1339<!-- _______________________________________________________________________ -->
1340<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1341
1342<div class="doc_text">
1343
1344<h5>Overview:</h5>
1345<p>The integer type is a very simple derived type that simply specifies an
1346arbitrary bit width for the integer type desired. Any bit width from 1 bit to
13472^23-1 (about 8 million) can be specified.</p>
1348
1349<h5>Syntax:</h5>
1350
1351<pre>
1352 iN
1353</pre>
1354
1355<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1356value.</p>
1357
1358<h5>Examples:</h5>
1359<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1360 <tbody>
1361 <tr>
1362 <td><tt>i1</tt></td>
1363 <td>a single-bit integer.</td>
1364 </tr><tr>
1365 <td><tt>i32</tt></td>
1366 <td>a 32-bit integer.</td>
1367 </tr><tr>
1368 <td><tt>i1942652</tt></td>
1369 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1370 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1371 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1372</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1373
1374<p>Note that the code generator does not yet support large integer types
1375to be used as function return types. The specific limit on how large a
1376return type the code generator can currently handle is target-dependent;
1377currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1378targets.</p>
1379
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1380</div>
1381
1382<!-- _______________________________________________________________________ -->
1383<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1384
1385<div class="doc_text">
1386
1387<h5>Overview:</h5>
1388
1389<p>The array type is a very simple derived type that arranges elements
1390sequentially in memory. The array type requires a size (number of
1391elements) and an underlying data type.</p>
1392
1393<h5>Syntax:</h5>
1394
1395<pre>
1396 [&lt;# elements&gt; x &lt;elementtype&gt;]
1397</pre>
1398
1399<p>The number of elements is a constant integer value; elementtype may
1400be any type with a size.</p>
1401
1402<h5>Examples:</h5>
1403<table class="layout">
1404 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1405 <td class="left"><tt>[40 x i32]</tt></td>
1406 <td class="left">Array of 40 32-bit integer values.</td>
1407 </tr>
1408 <tr class="layout">
1409 <td class="left"><tt>[41 x i32]</tt></td>
1410 <td class="left">Array of 41 32-bit integer values.</td>
1411 </tr>
1412 <tr class="layout">
1413 <td class="left"><tt>[4 x i8]</tt></td>
1414 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1415 </tr>
1416</table>
1417<p>Here are some examples of multidimensional arrays:</p>
1418<table class="layout">
1419 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1420 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1421 <td class="left">3x4 array of 32-bit integer values.</td>
1422 </tr>
1423 <tr class="layout">
1424 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1425 <td class="left">12x10 array of single precision floating point values.</td>
1426 </tr>
1427 <tr class="layout">
1428 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1429 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1430 </tr>
1431</table>
1432
1433<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1434length array. Normally, accesses past the end of an array are undefined in
1435LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1436As a special case, however, zero length arrays are recognized to be variable
1437length. This allows implementation of 'pascal style arrays' with the LLVM
1438type "{ i32, [0 x float]}", for example.</p>
1439
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1440<p>Note that the code generator does not yet support large aggregate types
1441to be used as function return types. The specific limit on how large an
1442aggregate return type the code generator can currently handle is
1443target-dependent, and also dependent on the aggregate element types.</p>
1444
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1445</div>
1446
1447<!-- _______________________________________________________________________ -->
1448<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1449<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1450
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1451<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1452
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1453<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1454consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1455return 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]1456If 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]1457class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1458
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1459<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1460
1461<pre>
1462 &lt;returntype list&gt; (&lt;parameter list&gt;)
1463</pre>
1464
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1465<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1466specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1467which indicates that the function takes a variable number of arguments.
1468Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1469 href="#int_varargs">variable argument handling intrinsic</a> functions.
1470'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1471<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1472
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1473<h5>Examples:</h5>
1474<table class="layout">
1475 <tr class="layout">
1476 <td class="left"><tt>i32 (i32)</tt></td>
1477 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1478 </td>
1479 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1480 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1481 </tt></td>
1482 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1483 an <tt>i16</tt> that should be sign extended and a
1484 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1485 <tt>float</tt>.
1486 </td>
1487 </tr><tr class="layout">
1488 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1489 <td class="left">A vararg function that takes at least one
1490 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1491 which returns an integer. This is the signature for <tt>printf</tt> in
1492 LLVM.
1493 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1494 </tr><tr class="layout">
1495 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Misha Brukmanafc88b02008-11-27 06:41:20 +0000[diff] [blame]1496 <td class="left">A function taking an <tt>i32</tt>, returning two
1497 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1498 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1499 </tr>
1500</table>
1501
1502</div>
1503<!-- _______________________________________________________________________ -->
1504<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1505<div class="doc_text">
1506<h5>Overview:</h5>
1507<p>The structure type is used to represent a collection of data members
1508together in memory. The packing of the field types is defined to match
1509the ABI of the underlying processor. The elements of a structure may
1510be any type that has a size.</p>
1511<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1512and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1513field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1514instruction.</p>
1515<h5>Syntax:</h5>
1516<pre> { &lt;type list&gt; }<br></pre>
1517<h5>Examples:</h5>
1518<table class="layout">
1519 <tr class="layout">
1520 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1521 <td class="left">A triple of three <tt>i32</tt> values</td>
1522 </tr><tr class="layout">
1523 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1524 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1525 second element is a <a href="#t_pointer">pointer</a> to a
1526 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1527 an <tt>i32</tt>.</td>
1528 </tr>
1529</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1530
1531<p>Note that the code generator does not yet support large aggregate types
1532to be used as function return types. The specific limit on how large an
1533aggregate return type the code generator can currently handle is
1534target-dependent, and also dependent on the aggregate element types.</p>
1535
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1536</div>
1537
1538<!-- _______________________________________________________________________ -->
1539<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1540</div>
1541<div class="doc_text">
1542<h5>Overview:</h5>
1543<p>The packed structure type is used to represent a collection of data members
1544together in memory. There is no padding between fields. Further, the alignment
1545of a packed structure is 1 byte. The elements of a packed structure may
1546be any type that has a size.</p>
1547<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1548and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1549field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1550instruction.</p>
1551<h5>Syntax:</h5>
1552<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1553<h5>Examples:</h5>
1554<table class="layout">
1555 <tr class="layout">
1556 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1557 <td class="left">A triple of three <tt>i32</tt> values</td>
1558 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1559 <td class="left">
1560<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1561 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1562 second element is a <a href="#t_pointer">pointer</a> to a
1563 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1564 an <tt>i32</tt>.</td>
1565 </tr>
1566</table>
1567</div>
1568
1569<!-- _______________________________________________________________________ -->
1570<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1571<div class="doc_text">
1572<h5>Overview:</h5>
1573<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1574reference to another object, which must live in memory. Pointer types may have
1575an optional address space attribute defining the target-specific numbered
1576address space where the pointed-to object resides. The default address space is
1577zero.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +0000[diff] [blame]1578
1579<p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does
Chris Lattner0d879e02009-02-08 22:21:28 +0000[diff] [blame]1580it permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +0000[diff] [blame]1581
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1582<h5>Syntax:</h5>
1583<pre> &lt;type&gt; *<br></pre>
1584<h5>Examples:</h5>
1585<table class="layout">
1586 <tr class="layout">
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]1587 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1588 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1589 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1590 </tr>
1591 <tr class="layout">
1592 <td class="left"><tt>i32 (i32 *) *</tt></td>
1593 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1594 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1595 <tt>i32</tt>.</td>
1596 </tr>
1597 <tr class="layout">
1598 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1599 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1600 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1601 </tr>
1602</table>
1603</div>
1604
1605<!-- _______________________________________________________________________ -->
1606<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1607<div class="doc_text">
1608
1609<h5>Overview:</h5>
1610
1611<p>A vector type is a simple derived type that represents a vector
1612of elements. Vector types are used when multiple primitive data
1613are operated in parallel using a single instruction (SIMD).
1614A vector type requires a size (number of
1615elements) and an underlying primitive data type. Vectors must have a power
1616of two length (1, 2, 4, 8, 16 ...). Vector types are
1617considered <a href="#t_firstclass">first class</a>.</p>
1618
1619<h5>Syntax:</h5>
1620
1621<pre>
1622 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1623</pre>
1624
1625<p>The number of elements is a constant integer value; elementtype may
1626be any integer or floating point type.</p>
1627
1628<h5>Examples:</h5>
1629
1630<table class="layout">
1631 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1632 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1633 <td class="left">Vector of 4 32-bit integer values.</td>
1634 </tr>
1635 <tr class="layout">
1636 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1637 <td class="left">Vector of 8 32-bit floating-point values.</td>
1638 </tr>
1639 <tr class="layout">
1640 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1641 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1642 </tr>
1643</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1644
1645<p>Note that the code generator does not yet support large vector types
1646to be used as function return types. The specific limit on how large a
1647vector return type codegen can currently handle is target-dependent;
1648currently it's often a few times longer than a hardware vector register.</p>
1649
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1650</div>
1651
1652<!-- _______________________________________________________________________ -->
1653<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1654<div class="doc_text">
1655
1656<h5>Overview:</h5>
1657
1658<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1659corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1660In LLVM, opaque types can eventually be resolved to any type (not just a
1661structure type).</p>
1662
1663<h5>Syntax:</h5>
1664
1665<pre>
1666 opaque
1667</pre>
1668
1669<h5>Examples:</h5>
1670
1671<table class="layout">
1672 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1673 <td class="left"><tt>opaque</tt></td>
1674 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1675 </tr>
1676</table>
1677</div>
1678
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]1679<!-- ======================================================================= -->
1680<div class="doc_subsection">
1681 <a name="t_uprefs">Type Up-references</a>
1682</div>
1683
1684<div class="doc_text">
1685<h5>Overview:</h5>
1686<p>
1687An "up reference" allows you to refer to a lexically enclosing type without
1688requiring it to have a name. For instance, a structure declaration may contain a
1689pointer to any of the types it is lexically a member of. Example of up
1690references (with their equivalent as named type declarations) include:</p>
1691
1692<pre>
Chris Lattner5ad632d2009-02-09 10:00:56 +0000[diff] [blame]1693 { \2 * } %x = type { %x* }
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]1694 { \2 }* %y = type { %y }*
1695 \1* %z = type %z*
1696</pre>
1697
1698<p>
1699An up reference is needed by the asmprinter for printing out cyclic types when
1700there is no declared name for a type in the cycle. Because the asmprinter does
1701not want to print out an infinite type string, it needs a syntax to handle
1702recursive types that have no names (all names are optional in llvm IR).
1703</p>
1704
1705<h5>Syntax:</h5>
1706<pre>
1707 \&lt;level&gt;
1708</pre>
1709
1710<p>
1711The level is the count of the lexical type that is being referred to.
1712</p>
1713
1714<h5>Examples:</h5>
1715
1716<table class="layout">
1717 <tr class="layout">
1718 <td class="left"><tt>\1*</tt></td>
1719 <td class="left">Self-referential pointer.</td>
1720 </tr>
1721 <tr class="layout">
1722 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1723 <td class="left">Recursive structure where the upref refers to the out-most
1724 structure.</td>
1725 </tr>
1726</table>
1727</div>
1728
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1729
1730<!-- *********************************************************************** -->
1731<div class="doc_section"> <a name="constants">Constants</a> </div>
1732<!-- *********************************************************************** -->
1733
1734<div class="doc_text">
1735
1736<p>LLVM has several different basic types of constants. This section describes
1737them all and their syntax.</p>
1738
1739</div>
1740
1741<!-- ======================================================================= -->
1742<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1743
1744<div class="doc_text">
1745
1746<dl>
1747 <dt><b>Boolean constants</b></dt>
1748
1749 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1750 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1751 </dd>
1752
1753 <dt><b>Integer constants</b></dt>
1754
1755 <dd>Standard integers (such as '4') are constants of the <a
1756 href="#t_integer">integer</a> type. Negative numbers may be used with
1757 integer types.
1758 </dd>
1759
1760 <dt><b>Floating point constants</b></dt>
1761
1762 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1763 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1764 notation (see below). The assembler requires the exact decimal value of
1765 a floating-point constant. For example, the assembler accepts 1.25 but
1766 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1767 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1768
1769 <dt><b>Null pointer constants</b></dt>
1770
1771 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1772 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1773
1774</dl>
1775
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1776<p>The one non-intuitive notation for constants is the hexadecimal form
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1777of floating point constants. For example, the form '<tt>double
17780x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
17794.5e+15</tt>'. The only time hexadecimal floating point constants are required
1780(and the only time that they are generated by the disassembler) is when a
1781floating point constant must be emitted but it cannot be represented as a
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1782decimal floating point number in a reasonable number of digits. For example,
1783NaN's, infinities, and other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1784special values are represented in their IEEE hexadecimal format so that
1785assembly and disassembly do not cause any bits to change in the constants.</p>
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1786<p>When using the hexadecimal form, constants of types float and double are
1787represented using the 16-digit form shown above (which matches the IEEE754
1788representation for double); float values must, however, be exactly representable
1789as IEE754 single precision.
1790Hexadecimal format is always used for long
1791double, and there are three forms of long double. The 80-bit
1792format used by x86 is represented as <tt>0xK</tt>
1793followed by 20 hexadecimal digits.
1794The 128-bit format used by PowerPC (two adjacent doubles) is represented
1795by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit
1796format is represented
1797by <tt>0xL</tt> followed by 32 hexadecimal digits; no currently supported
1798target uses this format. Long doubles will only work if they match
1799the long double format on your target. All hexadecimal formats are big-endian
1800(sign bit at the left).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1801</div>
1802
1803<!-- ======================================================================= -->
Chris Lattner97063852009-02-28 18:32:25 +0000[diff] [blame]1804<div class="doc_subsection">
1805<a name="aggregateconstants"> <!-- old anchor -->
1806<a name="complexconstants">Complex Constants</a></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1807</div>
1808
1809<div class="doc_text">
Chris Lattner97063852009-02-28 18:32:25 +0000[diff] [blame]1810<p>Complex constants are a (potentially recursive) combination of simple
1811constants and smaller complex constants.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1812
1813<dl>
1814 <dt><b>Structure constants</b></dt>
1815
1816 <dd>Structure constants are represented with notation similar to structure
1817 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1818 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1819 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]1820 must have <a href="#t_struct">structure type</a>, and the number and
1821 types of elements must match those specified by the type.
1822 </dd>
1823
1824 <dt><b>Array constants</b></dt>
1825
1826 <dd>Array constants are represented with notation similar to array type
1827 definitions (a comma separated list of elements, surrounded by square brackets
1828 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1829 constants must have <a href="#t_array">array type</a>, and the number and
1830 types of elements must match those specified by the type.
1831 </dd>
1832
1833 <dt><b>Vector constants</b></dt>
1834
1835 <dd>Vector constants are represented with notation similar to vector type
1836 definitions (a comma separated list of elements, surrounded by
1837 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1838 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1839 href="#t_vector">vector type</a>, and the number and types of elements must
1840 match those specified by the type.
1841 </dd>
1842
1843 <dt><b>Zero initialization</b></dt>
1844
1845 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1846 value to zero of <em>any</em> type, including scalar and aggregate types.
1847 This is often used to avoid having to print large zero initializers (e.g. for
1848 large arrays) and is always exactly equivalent to using explicit zero
1849 initializers.
1850 </dd>
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame^]1851
1852 <dt><b>Metadata node</b></dt>
1853
1854 <dd>A metadata node is a structure-like constant with the type of an empty
1855 struct. For example: "<tt>{ } !{ i32 0, { } !"test" }</tt>". Unlike other
1856 constants that are meant to be interpreted as part of the instruction stream,
1857 metadata is a place to attach additional information such as debug info.
1858 </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1859</dl>
1860
1861</div>
1862
1863<!-- ======================================================================= -->
1864<div class="doc_subsection">
1865 <a name="globalconstants">Global Variable and Function Addresses</a>
1866</div>
1867
1868<div class="doc_text">
1869
1870<p>The addresses of <a href="#globalvars">global variables</a> and <a
1871href="#functionstructure">functions</a> are always implicitly valid (link-time)
1872constants. These constants are explicitly referenced when the <a
1873href="#identifiers">identifier for the global</a> is used and always have <a
1874href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1875file:</p>
1876
1877<div class="doc_code">
1878<pre>
1879@X = global i32 17
1880@Y = global i32 42
1881@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1882</pre>
1883</div>
1884
1885</div>
1886
1887<!-- ======================================================================= -->
1888<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1889<div class="doc_text">
1890 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1891 no specific value. Undefined values may be of any type and be used anywhere
1892 a constant is permitted.</p>
1893
1894 <p>Undefined values indicate to the compiler that the program is well defined
1895 no matter what value is used, giving the compiler more freedom to optimize.
1896 </p>
1897</div>
1898
1899<!-- ======================================================================= -->
1900<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1901</div>
1902
1903<div class="doc_text">
1904
1905<p>Constant expressions are used to allow expressions involving other constants
1906to be used as constants. Constant expressions may be of any <a
1907href="#t_firstclass">first class</a> type and may involve any LLVM operation
1908that does not have side effects (e.g. load and call are not supported). The
1909following is the syntax for constant expressions:</p>
1910
1911<dl>
1912 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1913 <dd>Truncate a constant to another type. The bit size of CST must be larger
1914 than the bit size of TYPE. Both types must be integers.</dd>
1915
1916 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1917 <dd>Zero extend a constant to another type. The bit size of CST must be
1918 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1919
1920 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1921 <dd>Sign extend a constant to another type. The bit size of CST must be
1922 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1923
1924 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1925 <dd>Truncate a floating point constant to another floating point type. The
1926 size of CST must be larger than the size of TYPE. Both types must be
1927 floating point.</dd>
1928
1929 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1930 <dd>Floating point extend a constant to another type. The size of CST must be
1931 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1932
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1933 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1934 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1935 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1936 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1937 of the same number of elements. If the value won't fit in the integer type,
1938 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1939
1940 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1941 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1942 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1943 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1944 of the same number of elements. If the value won't fit in the integer type,
1945 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1946
1947 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1948 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1949 constant. TYPE must be a scalar or vector floating point type. CST must be of
1950 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1951 of the same number of elements. If the value won't fit in the floating point
1952 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1953
1954 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1955 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1956 constant. TYPE must be a scalar or vector floating point type. CST must be of
1957 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1958 of the same number of elements. If the value won't fit in the floating point
1959 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1960
1961 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1962 <dd>Convert a pointer typed constant to the corresponding integer constant
1963 TYPE must be an integer type. CST must be of pointer type. The CST value is
1964 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1965
1966 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1967 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1968 pointer type. CST must be of integer type. The CST value is zero extended,
1969 truncated, or unchanged to make it fit in a pointer size. This one is
1970 <i>really</i> dangerous!</dd>
1971
1972 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
Chris Lattner557bc5d2009-02-28 18:27:03 +0000[diff] [blame]1973 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
1974 are the same as those for the <a href="#i_bitcast">bitcast
1975 instruction</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1976
1977 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1978
1979 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1980 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1981 instruction, the index list may have zero or more indexes, which are required
1982 to make sense for the type of "CSTPTR".</dd>
1983
1984 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1985
1986 <dd>Perform the <a href="#i_select">select operation</a> on
1987 constants.</dd>
1988
1989 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1990 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1991
1992 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1993 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1994
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1995 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1996 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1997
1998 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1999 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
2000
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2001 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2002
2003 <dd>Perform the <a href="#i_extractelement">extractelement
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2004 operation</a> on constants.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2005
2006 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2007
2008 <dd>Perform the <a href="#i_insertelement">insertelement
2009 operation</a> on constants.</dd>
2010
2011
2012 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2013
2014 <dd>Perform the <a href="#i_shufflevector">shufflevector
2015 operation</a> on constants.</dd>
2016
2017 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2018
2019 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2020 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
2021 binary</a> operations. The constraints on operands are the same as those for
2022 the corresponding instruction (e.g. no bitwise operations on floating point
2023 values are allowed).</dd>
2024</dl>
2025</div>
2026
Nick Lewycky4dcf8102009-04-04 07:22:01 +0000[diff] [blame^]2027<!-- ======================================================================= -->
2028<div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2029</div>
2030
2031<div class="doc_text">
2032
2033<p>Embedded metadata provides a way to attach arbitrary data to the
2034instruction stream without affecting the behaviour of the program. There are
2035two metadata primitives, strings and nodes. All metadata has the type of an
2036empty struct and is identified in syntax by a preceding exclamation point
2037('<tt>!</tt>').
2038</p>
2039
2040<p>A metadata string is a string surrounded by double quotes. It can contain
2041any character by escaping non-printable characters with "\xx" where "xx" is
2042the two digit hex code. For example: "<tt>!"test\00"</tt>".
2043</p>
2044
2045<p>Metadata nodes are represented with notation similar to structure constants
2046(a comma separated list of elements, surrounded by braces and preceeded by an
2047exclamation point). For example: "<tt>!{ { } !"test\00", i32 10}</tt>".
2048</p>
2049
2050<p>Optimizations may rely on metadata to provide additional information about
2051the program that isn't available in the instructions, or that isn't easily
2052computable. Similarly, the code generator may expect a certain metadata format
2053to be used to express debugging information.</p>
2054</div>
2055
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2056<!-- *********************************************************************** -->
2057<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2058<!-- *********************************************************************** -->
2059
2060<!-- ======================================================================= -->
2061<div class="doc_subsection">
2062<a name="inlineasm">Inline Assembler Expressions</a>
2063</div>
2064
2065<div class="doc_text">
2066
2067<p>
2068LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
2069Module-Level Inline Assembly</a>) through the use of a special value. This
2070value represents the inline assembler as a string (containing the instructions
2071to emit), a list of operand constraints (stored as a string), and a flag that
2072indicates whether or not the inline asm expression has side effects. An example
2073inline assembler expression is:
2074</p>
2075
2076<div class="doc_code">
2077<pre>
2078i32 (i32) asm "bswap $0", "=r,r"
2079</pre>
2080</div>
2081
2082<p>
2083Inline assembler expressions may <b>only</b> be used as the callee operand of
2084a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2085</p>
2086
2087<div class="doc_code">
2088<pre>
2089%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2090</pre>
2091</div>
2092
2093<p>
2094Inline asms with side effects not visible in the constraint list must be marked
2095as having side effects. This is done through the use of the
2096'<tt>sideeffect</tt>' keyword, like so:
2097</p>
2098
2099<div class="doc_code">
2100<pre>
2101call void asm sideeffect "eieio", ""()
2102</pre>
2103</div>
2104
2105<p>TODO: The format of the asm and constraints string still need to be
2106documented here. Constraints on what can be done (e.g. duplication, moving, etc
Chris Lattner1cdd64e2008-10-04 18:36:02 +0000[diff] [blame]2107need to be documented). This is probably best done by reference to another
2108document that covers inline asm from a holistic perspective.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2109</p>
2110
2111</div>
2112
2113<!-- *********************************************************************** -->
2114<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2115<!-- *********************************************************************** -->
2116
2117<div class="doc_text">
2118
2119<p>The LLVM instruction set consists of several different
2120classifications of instructions: <a href="#terminators">terminator
2121instructions</a>, <a href="#binaryops">binary instructions</a>,
2122<a href="#bitwiseops">bitwise binary instructions</a>, <a
2123 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2124instructions</a>.</p>
2125
2126</div>
2127
2128<!-- ======================================================================= -->
2129<div class="doc_subsection"> <a name="terminators">Terminator
2130Instructions</a> </div>
2131
2132<div class="doc_text">
2133
2134<p>As mentioned <a href="#functionstructure">previously</a>, every
2135basic block in a program ends with a "Terminator" instruction, which
2136indicates which block should be executed after the current block is
2137finished. These terminator instructions typically yield a '<tt>void</tt>'
2138value: they produce control flow, not values (the one exception being
2139the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2140<p>There are six different terminator instructions: the '<a
2141 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2142instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
2143the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2144 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2145 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2146
2147</div>
2148
2149<!-- _______________________________________________________________________ -->
2150<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2151Instruction</a> </div>
2152<div class="doc_text">
2153<h5>Syntax:</h5>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2154<pre>
2155 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]2156 ret void <i>; Return from void function</i>
2157</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2158
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2159<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2160
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2161<p>The '<tt>ret</tt>' instruction is used to return control flow (and
2162optionally a value) from a function back to the caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2163<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2164returns a value and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2165control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2166
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2167<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2168
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2169<p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2170the return value. The type of the return value must be a
2171'<a href="#t_firstclass">first class</a>' type.</p>
2172
2173<p>A function is not <a href="#wellformed">well formed</a> if
2174it it has a non-void return type and contains a '<tt>ret</tt>'
2175instruction with no return value or a return value with a type that
2176does not match its type, or if it has a void return type and contains
2177a '<tt>ret</tt>' instruction with a return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2178
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2179<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2180
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2181<p>When the '<tt>ret</tt>' instruction is executed, control flow
2182returns back to the calling function's context. If the caller is a "<a
2183 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2184the instruction after the call. If the caller was an "<a
2185 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2186at the beginning of the "normal" destination block. If the instruction
2187returns a value, that value shall set the call or invoke instruction's
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2188return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2189
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2190<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2191
2192<pre>
2193 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2194 ret void <i>; Return from a void function</i>
Bill Wendlingd163e2d2009-02-28 22:12:54 +0000[diff] [blame]2195 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2196</pre>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]2197
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]2198<p>Note that the code generator does not yet fully support large
2199 return values. The specific sizes that are currently supported are
2200 dependent on the target. For integers, on 32-bit targets the limit
2201 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2202 For aggregate types, the current limits are dependent on the element
2203 types; for example targets are often limited to 2 total integer
2204 elements and 2 total floating-point elements.</p>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]2205
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2206</div>
2207<!-- _______________________________________________________________________ -->
2208<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2209<div class="doc_text">
2210<h5>Syntax:</h5>
2211<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
2212</pre>
2213<h5>Overview:</h5>
2214<p>The '<tt>br</tt>' instruction is used to cause control flow to
2215transfer to a different basic block in the current function. There are
2216two forms of this instruction, corresponding to a conditional branch
2217and an unconditional branch.</p>
2218<h5>Arguments:</h5>
2219<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2220single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2221unconditional form of the '<tt>br</tt>' instruction takes a single
2222'<tt>label</tt>' value as a target.</p>
2223<h5>Semantics:</h5>
2224<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2225argument is evaluated. If the value is <tt>true</tt>, control flows
2226to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2227control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2228<h5>Example:</h5>
2229<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
2230 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2231</div>
2232<!-- _______________________________________________________________________ -->
2233<div class="doc_subsubsection">
2234 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2235</div>
2236
2237<div class="doc_text">
2238<h5>Syntax:</h5>
2239
2240<pre>
2241 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
2242</pre>
2243
2244<h5>Overview:</h5>
2245
2246<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2247several different places. It is a generalization of the '<tt>br</tt>'
2248instruction, allowing a branch to occur to one of many possible
2249destinations.</p>
2250
2251
2252<h5>Arguments:</h5>
2253
2254<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2255comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2256an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2257table is not allowed to contain duplicate constant entries.</p>
2258
2259<h5>Semantics:</h5>
2260
2261<p>The <tt>switch</tt> instruction specifies a table of values and
2262destinations. When the '<tt>switch</tt>' instruction is executed, this
2263table is searched for the given value. If the value is found, control flow is
2264transfered to the corresponding destination; otherwise, control flow is
2265transfered to the default destination.</p>
2266
2267<h5>Implementation:</h5>
2268
2269<p>Depending on properties of the target machine and the particular
2270<tt>switch</tt> instruction, this instruction may be code generated in different
2271ways. For example, it could be generated as a series of chained conditional
2272branches or with a lookup table.</p>
2273
2274<h5>Example:</h5>
2275
2276<pre>
2277 <i>; Emulate a conditional br instruction</i>
2278 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]2279 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2280
2281 <i>; Emulate an unconditional br instruction</i>
2282 switch i32 0, label %dest [ ]
2283
2284 <i>; Implement a jump table:</i>
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]2285 switch i32 %val, label %otherwise [ i32 0, label %onzero
2286 i32 1, label %onone
2287 i32 2, label %ontwo ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2288</pre>
2289</div>
2290
2291<!-- _______________________________________________________________________ -->
2292<div class="doc_subsubsection">
2293 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2294</div>
2295
2296<div class="doc_text">
2297
2298<h5>Syntax:</h5>
2299
2300<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2301 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] &lt;ptr to function ty&gt; &lt;function ptr val&gt;(&lt;function args&gt;) [<a href="#fnattrs">fn attrs</a>]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2302 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2303</pre>
2304
2305<h5>Overview:</h5>
2306
2307<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2308function, with the possibility of control flow transfer to either the
2309'<tt>normal</tt>' label or the
2310'<tt>exception</tt>' label. If the callee function returns with the
2311"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2312"normal" label. If the callee (or any indirect callees) returns with the "<a
2313href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2314continued at the dynamically nearest "exception" label.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2315
2316<h5>Arguments:</h5>
2317
2318<p>This instruction requires several arguments:</p>
2319
2320<ol>
2321 <li>
2322 The optional "cconv" marker indicates which <a href="#callingconv">calling
2323 convention</a> the call should use. If none is specified, the call defaults
2324 to using C calling conventions.
2325 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2326
2327 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2328 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2329 and '<tt>inreg</tt>' attributes are valid here.</li>
2330
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2331 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2332 function value being invoked. In most cases, this is a direct function
2333 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2334 an arbitrary pointer to function value.
2335 </li>
2336
2337 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2338 function to be invoked. </li>
2339
2340 <li>'<tt>function args</tt>': argument list whose types match the function
2341 signature argument types. If the function signature indicates the function
2342 accepts a variable number of arguments, the extra arguments can be
2343 specified. </li>
2344
2345 <li>'<tt>normal label</tt>': the label reached when the called function
2346 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2347
2348 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2349 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2350
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2351 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2352 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2353 '<tt>readnone</tt>' attributes are valid here.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2354</ol>
2355
2356<h5>Semantics:</h5>
2357
2358<p>This instruction is designed to operate as a standard '<tt><a
2359href="#i_call">call</a></tt>' instruction in most regards. The primary
2360difference is that it establishes an association with a label, which is used by
2361the runtime library to unwind the stack.</p>
2362
2363<p>This instruction is used in languages with destructors to ensure that proper
2364cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2365exception. Additionally, this is important for implementation of
2366'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2367
2368<h5>Example:</h5>
2369<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2370 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2371 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2372 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2373 unwind label %TestCleanup <i>; {i32}:retval set</i>
2374</pre>
2375</div>
2376
2377
2378<!-- _______________________________________________________________________ -->
2379
2380<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2381Instruction</a> </div>
2382
2383<div class="doc_text">
2384
2385<h5>Syntax:</h5>
2386<pre>
2387 unwind
2388</pre>
2389
2390<h5>Overview:</h5>
2391
2392<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2393at the first callee in the dynamic call stack which used an <a
2394href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2395primarily used to implement exception handling.</p>
2396
2397<h5>Semantics:</h5>
2398
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2399<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2400immediately halt. The dynamic call stack is then searched for the first <a
2401href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2402execution continues at the "exceptional" destination block specified by the
2403<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2404dynamic call chain, undefined behavior results.</p>
2405</div>
2406
2407<!-- _______________________________________________________________________ -->
2408
2409<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2410Instruction</a> </div>
2411
2412<div class="doc_text">
2413
2414<h5>Syntax:</h5>
2415<pre>
2416 unreachable
2417</pre>
2418
2419<h5>Overview:</h5>
2420
2421<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2422instruction is used to inform the optimizer that a particular portion of the
2423code is not reachable. This can be used to indicate that the code after a
2424no-return function cannot be reached, and other facts.</p>
2425
2426<h5>Semantics:</h5>
2427
2428<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2429</div>
2430
2431
2432
2433<!-- ======================================================================= -->
2434<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2435<div class="doc_text">
2436<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2437program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2438produce a single value. The operands might represent
2439multiple 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]2440The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2441<p>There are several different binary operators:</p>
2442</div>
2443<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2444<div class="doc_subsubsection">
2445 <a name="i_add">'<tt>add</tt>' Instruction</a>
2446</div>
2447
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2448<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2449
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2450<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2451
2452<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2453 &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]2454</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2455
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2456<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2457
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2458<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2459
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2460<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2461
2462<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2463 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2464 <a href="#t_vector">vector</a> values. Both arguments must have identical
2465 types.</p>
2466
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2467<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2468
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2469<p>The value produced is the integer or floating point sum of the two
2470operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2471
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2472<p>If an integer sum has unsigned overflow, the result returned is the
2473mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2474the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2475
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2476<p>Because LLVM integers use a two's complement representation, this
2477instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2478
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2479<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2480
2481<pre>
2482 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2483</pre>
2484</div>
2485<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2486<div class="doc_subsubsection">
2487 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2488</div>
2489
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2490<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2491
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2492<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2493
2494<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2495 &lt;result&gt; = 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]2496</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2497
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2498<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2499
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2500<p>The '<tt>sub</tt>' instruction returns the difference of its two
2501operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2502
2503<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2504'<tt>neg</tt>' instruction present in most other intermediate
2505representations.</p>
2506
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2507<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2508
2509<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2510 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2511 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2512 types.</p>
2513
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2514<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2515
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2516<p>The value produced is the integer or floating point difference of
2517the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2518
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2519<p>If an integer difference has unsigned overflow, the result returned is the
2520mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2521the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2522
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2523<p>Because LLVM integers use a two's complement representation, this
2524instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2525
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2526<h5>Example:</h5>
2527<pre>
2528 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2529 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2530</pre>
2531</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2532
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2533<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2534<div class="doc_subsubsection">
2535 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2536</div>
2537
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2538<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2539
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2540<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2541<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]2542</pre>
2543<h5>Overview:</h5>
2544<p>The '<tt>mul</tt>' instruction returns the product of its two
2545operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2546
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2547<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2548
2549<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2550href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2551or <a href="#t_vector">vector</a> values. Both arguments must have identical
2552types.</p>
2553
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2554<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2555
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2556<p>The value produced is the integer or floating point product of the
2557two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2558
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2559<p>If the result of an integer multiplication has unsigned overflow,
2560the result returned is the mathematical result modulo
25612<sup>n</sup>, where n is the bit width of the result.</p>
2562<p>Because LLVM integers use a two's complement representation, and the
2563result is the same width as the operands, this instruction returns the
2564correct result for both signed and unsigned integers. If a full product
2565(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2566should be sign-extended or zero-extended as appropriate to the
2567width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2568<h5>Example:</h5>
2569<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2570</pre>
2571</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2572
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2573<!-- _______________________________________________________________________ -->
2574<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2575</a></div>
2576<div class="doc_text">
2577<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2578<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]2579</pre>
2580<h5>Overview:</h5>
2581<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2582operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2583
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2584<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2585
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2586<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2587<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2588values. Both arguments must have identical types.</p>
2589
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2590<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2591
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2592<p>The value produced is the unsigned integer quotient of the two operands.</p>
2593<p>Note that unsigned integer division and signed integer division are distinct
2594operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2595<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2596<h5>Example:</h5>
2597<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2598</pre>
2599</div>
2600<!-- _______________________________________________________________________ -->
2601<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2602</a> </div>
2603<div class="doc_text">
2604<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2605<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2606 &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]2607</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2608
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2609<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2610
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2611<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2612operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2613
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2614<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2615
2616<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2617<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2618values. Both arguments must have identical types.</p>
2619
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2620<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2621<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]2622<p>Note that signed integer division and unsigned integer division are distinct
2623operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2624<p>Division by zero leads to undefined behavior. Overflow also leads to
2625undefined behavior; this is a rare case, but can occur, for example,
2626by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2627<h5>Example:</h5>
2628<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2629</pre>
2630</div>
2631<!-- _______________________________________________________________________ -->
2632<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2633Instruction</a> </div>
2634<div class="doc_text">
2635<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2636<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2637 &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]2638</pre>
2639<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2640
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2641<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2642operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2643
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2644<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2645
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2646<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2647<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2648of floating point values. Both arguments must have identical types.</p>
2649
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2650<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2651
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2652<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2653
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2654<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2655
2656<pre>
2657 &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]2658</pre>
2659</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2660
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2661<!-- _______________________________________________________________________ -->
2662<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2663</div>
2664<div class="doc_text">
2665<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2666<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]2667</pre>
2668<h5>Overview:</h5>
2669<p>The '<tt>urem</tt>' instruction returns the remainder from the
2670unsigned division of its two arguments.</p>
2671<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2672<p>The two arguments to the '<tt>urem</tt>' instruction must be
2673<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2674values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2675<h5>Semantics:</h5>
2676<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2677This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2678<p>Note that unsigned integer remainder and signed integer remainder are
2679distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2680<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2681<h5>Example:</h5>
2682<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2683</pre>
2684
2685</div>
2686<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2687<div class="doc_subsubsection">
2688 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2689</div>
2690
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2691<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2692
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2693<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2694
2695<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2696 &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]2697</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2698
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2699<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2700
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2701<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2702signed division of its two operands. This instruction can also take
2703<a href="#t_vector">vector</a> versions of the values in which case
2704the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2705
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2706<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2707
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2708<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2709<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2710values. Both arguments must have identical types.</p>
2711
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2712<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2713
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2714<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2715has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2716operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2717a value. For more information about the difference, see <a
2718 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2719Math Forum</a>. For a table of how this is implemented in various languages,
2720please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2721Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2722<p>Note that signed integer remainder and unsigned integer remainder are
2723distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2724<p>Taking the remainder of a division by zero leads to undefined behavior.
2725Overflow also leads to undefined behavior; this is a rare case, but can occur,
2726for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2727(The remainder doesn't actually overflow, but this rule lets srem be
2728implemented using instructions that return both the result of the division
2729and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2730<h5>Example:</h5>
2731<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2732</pre>
2733
2734</div>
2735<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2736<div class="doc_subsubsection">
2737 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2738
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2739<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2740
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2741<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2742<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]2743</pre>
2744<h5>Overview:</h5>
2745<p>The '<tt>frem</tt>' instruction returns the remainder from the
2746division of its two operands.</p>
2747<h5>Arguments:</h5>
2748<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2749<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2750of floating point values. Both arguments must have identical types.</p>
2751
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2752<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2753
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2754<p>This instruction returns the <i>remainder</i> of a division.
2755The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2756
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2757<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2758
2759<pre>
2760 &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]2761</pre>
2762</div>
2763
2764<!-- ======================================================================= -->
2765<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2766Operations</a> </div>
2767<div class="doc_text">
2768<p>Bitwise binary operators are used to do various forms of
2769bit-twiddling in a program. They are generally very efficient
2770instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2771instructions. They require two operands of the same type, execute an operation on them,
2772and 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]2773</div>
2774
2775<!-- _______________________________________________________________________ -->
2776<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2777Instruction</a> </div>
2778<div class="doc_text">
2779<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2780<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]2781</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2782
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2783<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2784
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2785<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2786the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2787
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2788<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2789
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2790<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2791 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2792type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2793
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2794<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2795
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2796<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2797where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
Mon P Wangb0f51822008-12-10 08:55:09 +0000[diff] [blame]2798equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2799If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2800corresponding shift amount in <tt>op2</tt>.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2801
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2802<h5>Example:</h5><pre>
2803 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2804 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2805 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2806 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2807 &lt;result&gt; = shl &lt;2 x i32&gt; &lt; i32 1, i32 1&gt;, &lt; i32 1, i32 2&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 2, i32 4&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2808</pre>
2809</div>
2810<!-- _______________________________________________________________________ -->
2811<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2812Instruction</a> </div>
2813<div class="doc_text">
2814<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2815<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]2816</pre>
2817
2818<h5>Overview:</h5>
2819<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2820operand shifted to the right a specified number of bits with zero fill.</p>
2821
2822<h5>Arguments:</h5>
2823<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2824<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]2825type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2826
2827<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2828
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2829<p>This instruction always performs a logical shift right operation. The most
2830significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2831shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
Mon P Wangb0f51822008-12-10 08:55:09 +0000[diff] [blame]2832the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2833vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2834amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2835
2836<h5>Example:</h5>
2837<pre>
2838 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2839 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2840 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2841 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2842 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2843 &lt;result&gt; = lshr &lt;2 x i32&gt; &lt; i32 -2, i32 4&gt;, &lt; i32 1, i32 2&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0x7FFFFFFF, i32 1&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2844</pre>
2845</div>
2846
2847<!-- _______________________________________________________________________ -->
2848<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2849Instruction</a> </div>
2850<div class="doc_text">
2851
2852<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2853<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]2854</pre>
2855
2856<h5>Overview:</h5>
2857<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2858operand shifted to the right a specified number of bits with sign extension.</p>
2859
2860<h5>Arguments:</h5>
2861<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2862<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]2863type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2864
2865<h5>Semantics:</h5>
2866<p>This instruction always performs an arithmetic shift right operation,
2867The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2868of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
Mon P Wangb0f51822008-12-10 08:55:09 +0000[diff] [blame]2869larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2870arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2871corresponding shift amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2872
2873<h5>Example:</h5>
2874<pre>
2875 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2876 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2877 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2878 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2879 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2880 &lt;result&gt; = ashr &lt;2 x i32&gt; &lt; i32 -2, i32 4&gt;, &lt; i32 1, i32 3&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 -1, i32 0&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2881</pre>
2882</div>
2883
2884<!-- _______________________________________________________________________ -->
2885<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2886Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2887
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2888<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2889
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2890<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2891
2892<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2893 &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]2894</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2895
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2896<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2897
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2898<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2899its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2900
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2901<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2902
2903<p>The two arguments to the '<tt>and</tt>' instruction must be
2904<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2905values. Both arguments must have identical types.</p>
2906
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2907<h5>Semantics:</h5>
2908<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2909<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2910<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2911<table border="1" cellspacing="0" cellpadding="4">
2912 <tbody>
2913 <tr>
2914 <td>In0</td>
2915 <td>In1</td>
2916 <td>Out</td>
2917 </tr>
2918 <tr>
2919 <td>0</td>
2920 <td>0</td>
2921 <td>0</td>
2922 </tr>
2923 <tr>
2924 <td>0</td>
2925 <td>1</td>
2926 <td>0</td>
2927 </tr>
2928 <tr>
2929 <td>1</td>
2930 <td>0</td>
2931 <td>0</td>
2932 </tr>
2933 <tr>
2934 <td>1</td>
2935 <td>1</td>
2936 <td>1</td>
2937 </tr>
2938 </tbody>
2939</table>
2940</div>
2941<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2942<pre>
2943 &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]2944 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2945 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2946</pre>
2947</div>
2948<!-- _______________________________________________________________________ -->
2949<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2950<div class="doc_text">
2951<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2952<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]2953</pre>
2954<h5>Overview:</h5>
2955<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2956or of its two operands.</p>
2957<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2958
2959<p>The two arguments to the '<tt>or</tt>' instruction must be
2960<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2961values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2962<h5>Semantics:</h5>
2963<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2964<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2965<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2966<table border="1" cellspacing="0" cellpadding="4">
2967 <tbody>
2968 <tr>
2969 <td>In0</td>
2970 <td>In1</td>
2971 <td>Out</td>
2972 </tr>
2973 <tr>
2974 <td>0</td>
2975 <td>0</td>
2976 <td>0</td>
2977 </tr>
2978 <tr>
2979 <td>0</td>
2980 <td>1</td>
2981 <td>1</td>
2982 </tr>
2983 <tr>
2984 <td>1</td>
2985 <td>0</td>
2986 <td>1</td>
2987 </tr>
2988 <tr>
2989 <td>1</td>
2990 <td>1</td>
2991 <td>1</td>
2992 </tr>
2993 </tbody>
2994</table>
2995</div>
2996<h5>Example:</h5>
2997<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2998 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2999 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3000</pre>
3001</div>
3002<!-- _______________________________________________________________________ -->
3003<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3004Instruction</a> </div>
3005<div class="doc_text">
3006<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3007<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]3008</pre>
3009<h5>Overview:</h5>
3010<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
3011or of its two operands. The <tt>xor</tt> is used to implement the
3012"one's complement" operation, which is the "~" operator in C.</p>
3013<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3014<p>The two arguments to the '<tt>xor</tt>' instruction must be
3015<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3016values. Both arguments must have identical types.</p>
3017
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3018<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3019
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3020<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3021<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]3022<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3023<table border="1" cellspacing="0" cellpadding="4">
3024 <tbody>
3025 <tr>
3026 <td>In0</td>
3027 <td>In1</td>
3028 <td>Out</td>
3029 </tr>
3030 <tr>
3031 <td>0</td>
3032 <td>0</td>
3033 <td>0</td>
3034 </tr>
3035 <tr>
3036 <td>0</td>
3037 <td>1</td>
3038 <td>1</td>
3039 </tr>
3040 <tr>
3041 <td>1</td>
3042 <td>0</td>
3043 <td>1</td>
3044 </tr>
3045 <tr>
3046 <td>1</td>
3047 <td>1</td>
3048 <td>0</td>
3049 </tr>
3050 </tbody>
3051</table>
3052</div>
3053<p> </p>
3054<h5>Example:</h5>
3055<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3056 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3057 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3058 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3059</pre>
3060</div>
3061
3062<!-- ======================================================================= -->
3063<div class="doc_subsection">
3064 <a name="vectorops">Vector Operations</a>
3065</div>
3066
3067<div class="doc_text">
3068
3069<p>LLVM supports several instructions to represent vector operations in a
3070target-independent manner. These instructions cover the element-access and
3071vector-specific operations needed to process vectors effectively. While LLVM
3072does directly support these vector operations, many sophisticated algorithms
3073will want to use target-specific intrinsics to take full advantage of a specific
3074target.</p>
3075
3076</div>
3077
3078<!-- _______________________________________________________________________ -->
3079<div class="doc_subsubsection">
3080 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3081</div>
3082
3083<div class="doc_text">
3084
3085<h5>Syntax:</h5>
3086
3087<pre>
3088 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
3089</pre>
3090
3091<h5>Overview:</h5>
3092
3093<p>
3094The '<tt>extractelement</tt>' instruction extracts a single scalar
3095element from a vector at a specified index.
3096</p>
3097
3098
3099<h5>Arguments:</h5>
3100
3101<p>
3102The first operand of an '<tt>extractelement</tt>' instruction is a
3103value of <a href="#t_vector">vector</a> type. The second operand is
3104an index indicating the position from which to extract the element.
3105The index may be a variable.</p>
3106
3107<h5>Semantics:</h5>
3108
3109<p>
3110The result is a scalar of the same type as the element type of
3111<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3112<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3113results are undefined.
3114</p>
3115
3116<h5>Example:</h5>
3117
3118<pre>
3119 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
3120</pre>
3121</div>
3122
3123
3124<!-- _______________________________________________________________________ -->
3125<div class="doc_subsubsection">
3126 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3127</div>
3128
3129<div class="doc_text">
3130
3131<h5>Syntax:</h5>
3132
3133<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]3134 &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]3135</pre>
3136
3137<h5>Overview:</h5>
3138
3139<p>
3140The '<tt>insertelement</tt>' instruction inserts a scalar
3141element into a vector at a specified index.
3142</p>
3143
3144
3145<h5>Arguments:</h5>
3146
3147<p>
3148The first operand of an '<tt>insertelement</tt>' instruction is a
3149value of <a href="#t_vector">vector</a> type. The second operand is a
3150scalar value whose type must equal the element type of the first
3151operand. The third operand is an index indicating the position at
3152which to insert the value. The index may be a variable.</p>
3153
3154<h5>Semantics:</h5>
3155
3156<p>
3157The result is a vector of the same type as <tt>val</tt>. Its
3158element values are those of <tt>val</tt> except at position
3159<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3160exceeds the length of <tt>val</tt>, the results are undefined.
3161</p>
3162
3163<h5>Example:</h5>
3164
3165<pre>
3166 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
3167</pre>
3168</div>
3169
3170<!-- _______________________________________________________________________ -->
3171<div class="doc_subsubsection">
3172 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3173</div>
3174
3175<div class="doc_text">
3176
3177<h5>Syntax:</h5>
3178
3179<pre>
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3180 &lt;result&gt; = shufflevector &lt;n x &lt;ty&gt;&gt; &lt;v1&gt;, &lt;n x &lt;ty&gt;&gt; &lt;v2&gt;, &lt;m x i32&gt; &lt;mask&gt; <i>; yields &lt;m x &lt;ty&gt;&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3181</pre>
3182
3183<h5>Overview:</h5>
3184
3185<p>
3186The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3187from two input vectors, returning a vector with the same element type as
3188the input and length that is the same as the shuffle mask.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3189</p>
3190
3191<h5>Arguments:</h5>
3192
3193<p>
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3194The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3195with types that match each other. The third argument is a shuffle mask whose
3196element type is always 'i32'. The result of the instruction is a vector whose
3197length is the same as the shuffle mask and whose element type is the same as
3198the element type of the first two operands.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3199</p>
3200
3201<p>
3202The shuffle mask operand is required to be a constant vector with either
3203constant integer or undef values.
3204</p>
3205
3206<h5>Semantics:</h5>
3207
3208<p>
3209The elements of the two input vectors are numbered from left to right across
3210both of the vectors. The shuffle mask operand specifies, for each element of
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3211the result vector, which element of the two input vectors the result element
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3212gets. The element selector may be undef (meaning "don't care") and the second
3213operand may be undef if performing a shuffle from only one vector.
3214</p>
3215
3216<h5>Example:</h5>
3217
3218<pre>
3219 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3220 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
3221 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
3222 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i> - Identity shuffle.
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3223 %result = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
3224 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i>
3225 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3226 &lt;8 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 &gt; <i>; yields &lt;8 x i32&gt;</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3227</pre>
3228</div>
3229
3230
3231<!-- ======================================================================= -->
3232<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3233 <a name="aggregateops">Aggregate Operations</a>
3234</div>
3235
3236<div class="doc_text">
3237
3238<p>LLVM supports several instructions for working with aggregate values.
3239</p>
3240
3241</div>
3242
3243<!-- _______________________________________________________________________ -->
3244<div class="doc_subsubsection">
3245 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3246</div>
3247
3248<div class="doc_text">
3249
3250<h5>Syntax:</h5>
3251
3252<pre>
3253 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
3254</pre>
3255
3256<h5>Overview:</h5>
3257
3258<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3259The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3260or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3261</p>
3262
3263
3264<h5>Arguments:</h5>
3265
3266<p>
3267The first operand of an '<tt>extractvalue</tt>' instruction is a
3268value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3269type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3270in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3271'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3272</p>
3273
3274<h5>Semantics:</h5>
3275
3276<p>
3277The result is the value at the position in the aggregate specified by
3278the index operands.
3279</p>
3280
3281<h5>Example:</h5>
3282
3283<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3284 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3285</pre>
3286</div>
3287
3288
3289<!-- _______________________________________________________________________ -->
3290<div class="doc_subsubsection">
3291 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3292</div>
3293
3294<div class="doc_text">
3295
3296<h5>Syntax:</h5>
3297
3298<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3299 &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]3300</pre>
3301
3302<h5>Overview:</h5>
3303
3304<p>
3305The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3306into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3307</p>
3308
3309
3310<h5>Arguments:</h5>
3311
3312<p>
3313The first operand of an '<tt>insertvalue</tt>' instruction is a
3314value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3315The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3316The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3317indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3318indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3319'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3320The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3321by the indices.
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3322</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3323
3324<h5>Semantics:</h5>
3325
3326<p>
3327The result is an aggregate of the same type as <tt>val</tt>. Its
3328value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3329specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3330</p>
3331
3332<h5>Example:</h5>
3333
3334<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3335 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3336</pre>
3337</div>
3338
3339
3340<!-- ======================================================================= -->
3341<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3342 <a name="memoryops">Memory Access and Addressing Operations</a>
3343</div>
3344
3345<div class="doc_text">
3346
3347<p>A key design point of an SSA-based representation is how it
3348represents memory. In LLVM, no memory locations are in SSA form, which
3349makes things very simple. This section describes how to read, write,
3350allocate, and free memory in LLVM.</p>
3351
3352</div>
3353
3354<!-- _______________________________________________________________________ -->
3355<div class="doc_subsubsection">
3356 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3357</div>
3358
3359<div class="doc_text">
3360
3361<h5>Syntax:</h5>
3362
3363<pre>
3364 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3365</pre>
3366
3367<h5>Overview:</h5>
3368
3369<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3370heap and returns a pointer to it. The object is always allocated in the generic
3371address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3372
3373<h5>Arguments:</h5>
3374
3375<p>The '<tt>malloc</tt>' instruction allocates
3376<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3377bytes of memory from the operating system and returns a pointer of the
3378appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3379number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3380If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3381be aligned to at least that boundary. If not specified, or if zero, the target can
3382choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3383
3384<p>'<tt>type</tt>' must be a sized type.</p>
3385
3386<h5>Semantics:</h5>
3387
3388<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Nick Lewyckyff384472008-11-24 03:41:24 +0000[diff] [blame]3389a pointer is returned. The result of a zero byte allocation is undefined. The
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3390result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3391
3392<h5>Example:</h5>
3393
3394<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3395 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3396
3397 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3398 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3399 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3400 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3401 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3402</pre>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]3403
3404<p>Note that the code generator does not yet respect the
3405 alignment value.</p>
3406
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3407</div>
3408
3409<!-- _______________________________________________________________________ -->
3410<div class="doc_subsubsection">
3411 <a name="i_free">'<tt>free</tt>' Instruction</a>
3412</div>
3413
3414<div class="doc_text">
3415
3416<h5>Syntax:</h5>
3417
3418<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3419 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3420</pre>
3421
3422<h5>Overview:</h5>
3423
3424<p>The '<tt>free</tt>' instruction returns memory back to the unused
3425memory heap to be reallocated in the future.</p>
3426
3427<h5>Arguments:</h5>
3428
3429<p>'<tt>value</tt>' shall be a pointer value that points to a value
3430that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3431instruction.</p>
3432
3433<h5>Semantics:</h5>
3434
3435<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3436after this instruction executes. If the pointer is null, the operation
3437is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3438
3439<h5>Example:</h5>
3440
3441<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3442 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3443 free [4 x i8]* %array
3444</pre>
3445</div>
3446
3447<!-- _______________________________________________________________________ -->
3448<div class="doc_subsubsection">
3449 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3450</div>
3451
3452<div class="doc_text">
3453
3454<h5>Syntax:</h5>
3455
3456<pre>
3457 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3458</pre>
3459
3460<h5>Overview:</h5>
3461
3462<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3463currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3464returns to its caller. The object is always allocated in the generic address
3465space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3466
3467<h5>Arguments:</h5>
3468
3469<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3470bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3471appropriate type to the program. If "NumElements" is specified, it is the
3472number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3473If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3474to be aligned to at least that boundary. If not specified, or if zero, the target
3475can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3476
3477<p>'<tt>type</tt>' may be any sized type.</p>
3478
3479<h5>Semantics:</h5>
3480
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3481<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3482there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3483memory is automatically released when the function returns. The '<tt>alloca</tt>'
3484instruction is commonly used to represent automatic variables that must
3485have an address available. When the function returns (either with the <tt><a
3486 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]3487instructions), the memory is reclaimed. Allocating zero bytes
3488is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3489
3490<h5>Example:</h5>
3491
3492<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3493 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3494 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3495 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3496 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3497</pre>
3498</div>
3499
3500<!-- _______________________________________________________________________ -->
3501<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3502Instruction</a> </div>
3503<div class="doc_text">
3504<h5>Syntax:</h5>
3505<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>
3506<h5>Overview:</h5>
3507<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3508<h5>Arguments:</h5>
3509<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3510address from which to load. The pointer must point to a <a
3511 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3512marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3513the number or order of execution of this <tt>load</tt> with other
3514volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3515instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3516<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3517The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3518(that is, the alignment of the memory address). A value of 0 or an
3519omitted "align" argument means that the operation has the preferential
3520alignment for the target. It is the responsibility of the code emitter
3521to ensure that the alignment information is correct. Overestimating
3522the alignment results in an undefined behavior. Underestimating the
3523alignment may produce less efficient code. An alignment of 1 is always
3524safe.
3525</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3526<h5>Semantics:</h5>
Duncan Sandsfe87fff2009-03-22 11:33:16 +0000[diff] [blame]3527<p>The location of memory pointed to is loaded. If the value being loaded
3528is of scalar type then the number of bytes read does not exceed the minimum
3529number of bytes needed to hold all bits of the type. For example, loading an
3530<tt>i24</tt> reads at most three bytes. When loading a value of a type like
3531<tt>i20</tt> with a size that is not an integral number of bytes, the result
3532is undefined if the value was not originally written using a store of the
3533same type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3534<h5>Examples:</h5>
3535<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3536 <a
3537 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3538 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3539</pre>
3540</div>
3541<!-- _______________________________________________________________________ -->
3542<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3543Instruction</a> </div>
3544<div class="doc_text">
3545<h5>Syntax:</h5>
3546<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3547 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3548</pre>
3549<h5>Overview:</h5>
3550<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3551<h5>Arguments:</h5>
3552<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3553to 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]3554operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3555of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3556operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3557optimizer is not allowed to modify the number or order of execution of
3558this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3559 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3560<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3561The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3562(that is, the alignment of the memory address). A value of 0 or an
3563omitted "align" argument means that the operation has the preferential
3564alignment for the target. It is the responsibility of the code emitter
3565to ensure that the alignment information is correct. Overestimating
3566the alignment results in an undefined behavior. Underestimating the
3567alignment may produce less efficient code. An alignment of 1 is always
3568safe.
3569</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3570<h5>Semantics:</h5>
3571<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
Duncan Sandsfe87fff2009-03-22 11:33:16 +0000[diff] [blame]3572at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.
3573If '<tt>&lt;value&gt;</tt>' is of scalar type then the number of bytes
3574written does not exceed the minimum number of bytes needed to hold all
3575bits of the type. For example, storing an <tt>i24</tt> writes at most
3576three bytes. When writing a value of a type like <tt>i20</tt> with a
3577size that is not an integral number of bytes, it is unspecified what
3578happens to the extra bits that do not belong to the type, but they will
3579typically be overwritten.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3580<h5>Example:</h5>
3581<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3582 store i32 3, i32* %ptr <i>; yields {void}</i>
3583 %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]3584</pre>
3585</div>
3586
3587<!-- _______________________________________________________________________ -->
3588<div class="doc_subsubsection">
3589 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3590</div>
3591
3592<div class="doc_text">
3593<h5>Syntax:</h5>
3594<pre>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3595 &lt;result&gt; = getelementptr &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3596</pre>
3597
3598<h5>Overview:</h5>
3599
3600<p>
3601The '<tt>getelementptr</tt>' instruction is used to get the address of a
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3602subelement of an aggregate data structure. It performs address calculation only
3603and does not access memory.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3604
3605<h5>Arguments:</h5>
3606
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3607<p>The first argument is always a pointer, and forms the basis of the
3608calculation. The remaining arguments are indices, that indicate which of the
3609elements of the aggregate object are indexed. The interpretation of each index
3610is dependent on the type being indexed into. The first index always indexes the
3611pointer value given as the first argument, the second index indexes a value of
3612the type pointed to (not necessarily the value directly pointed to, since the
3613first index can be non-zero), etc. The first type indexed into must be a pointer
3614value, subsequent types can be arrays, vectors and structs. Note that subsequent
3615types being indexed into can never be pointers, since that would require loading
3616the pointer before continuing calculation.</p>
3617
3618<p>The type of each index argument depends on the type it is indexing into.
3619When indexing into a (packed) structure, only <tt>i32</tt> integer
3620<b>constants</b> are allowed. When indexing into an array, pointer or vector,
3621only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3622will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3623
3624<p>For example, let's consider a C code fragment and how it gets
3625compiled to LLVM:</p>
3626
3627<div class="doc_code">
3628<pre>
3629struct RT {
3630 char A;
3631 int B[10][20];
3632 char C;
3633};
3634struct ST {
3635 int X;
3636 double Y;
3637 struct RT Z;
3638};
3639
3640int *foo(struct ST *s) {
3641 return &amp;s[1].Z.B[5][13];
3642}
3643</pre>
3644</div>
3645
3646<p>The LLVM code generated by the GCC frontend is:</p>
3647
3648<div class="doc_code">
3649<pre>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]3650%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3651%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3652
3653define i32* %foo(%ST* %s) {
3654entry:
3655 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3656 ret i32* %reg
3657}
3658</pre>
3659</div>
3660
3661<h5>Semantics:</h5>
3662
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3663<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3664type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3665}</tt>' type, a structure. The second index indexes into the third element of
3666the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3667i8 }</tt>' type, another structure. The third index indexes into the second
3668element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3669array. The two dimensions of the array are subscripted into, yielding an
3670'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3671to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3672
3673<p>Note that it is perfectly legal to index partially through a
3674structure, returning a pointer to an inner element. Because of this,
3675the LLVM code for the given testcase is equivalent to:</p>
3676
3677<pre>
3678 define i32* %foo(%ST* %s) {
3679 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3680 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3681 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3682 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3683 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3684 ret i32* %t5
3685 }
3686</pre>
3687
Chris Lattner50609942009-03-09 20:55:18 +0000[diff] [blame]3688<p>Note that it is undefined to access an array out of bounds: array
3689and pointer indexes must always be within the defined bounds of the
3690array type when accessed with an instruction that dereferences the
3691pointer (e.g. a load or store instruction). The one exception for
3692this rule is zero length arrays. These arrays are defined to be
3693accessible as variable length arrays, which requires access beyond the
3694zero'th element.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3695
3696<p>The getelementptr instruction is often confusing. For some more insight
3697into how it works, see <a href="GetElementPtr.html">the getelementptr
3698FAQ</a>.</p>
3699
3700<h5>Example:</h5>
3701
3702<pre>
3703 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3704 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3705 <i>; yields i8*:vptr</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3706 %vptr = getelementptr {i32, &lt;2 x i8&gt;}* %svptr, i64 0, i32 1, i32 1
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3707 <i>; yields i8*:eptr</i>
3708 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3709</pre>
3710</div>
3711
3712<!-- ======================================================================= -->
3713<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3714</div>
3715<div class="doc_text">
3716<p>The instructions in this category are the conversion instructions (casting)
3717which all take a single operand and a type. They perform various bit conversions
3718on the operand.</p>
3719</div>
3720
3721<!-- _______________________________________________________________________ -->
3722<div class="doc_subsubsection">
3723 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3724</div>
3725<div class="doc_text">
3726
3727<h5>Syntax:</h5>
3728<pre>
3729 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3730</pre>
3731
3732<h5>Overview:</h5>
3733<p>
3734The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3735</p>
3736
3737<h5>Arguments:</h5>
3738<p>
3739The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3740be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3741and type of the result, which must be an <a href="#t_integer">integer</a>
3742type. The bit size of <tt>value</tt> must be larger than the bit size of
3743<tt>ty2</tt>. Equal sized types are not allowed.</p>
3744
3745<h5>Semantics:</h5>
3746<p>
3747The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3748and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3749larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3750It will always truncate bits.</p>
3751
3752<h5>Example:</h5>
3753<pre>
3754 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3755 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3756 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3757</pre>
3758</div>
3759
3760<!-- _______________________________________________________________________ -->
3761<div class="doc_subsubsection">
3762 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3763</div>
3764<div class="doc_text">
3765
3766<h5>Syntax:</h5>
3767<pre>
3768 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3769</pre>
3770
3771<h5>Overview:</h5>
3772<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3773<tt>ty2</tt>.</p>
3774
3775
3776<h5>Arguments:</h5>
3777<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3778<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3779also be of <a href="#t_integer">integer</a> type. The bit size of the
3780<tt>value</tt> must be smaller than the bit size of the destination type,
3781<tt>ty2</tt>.</p>
3782
3783<h5>Semantics:</h5>
3784<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3785bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3786
3787<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3788
3789<h5>Example:</h5>
3790<pre>
3791 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3792 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3793</pre>
3794</div>
3795
3796<!-- _______________________________________________________________________ -->
3797<div class="doc_subsubsection">
3798 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3799</div>
3800<div class="doc_text">
3801
3802<h5>Syntax:</h5>
3803<pre>
3804 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3805</pre>
3806
3807<h5>Overview:</h5>
3808<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3809
3810<h5>Arguments:</h5>
3811<p>
3812The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3813<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3814also be of <a href="#t_integer">integer</a> type. The bit size of the
3815<tt>value</tt> must be smaller than the bit size of the destination type,
3816<tt>ty2</tt>.</p>
3817
3818<h5>Semantics:</h5>
3819<p>
3820The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3821bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3822the type <tt>ty2</tt>.</p>
3823
3824<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3825
3826<h5>Example:</h5>
3827<pre>
3828 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3829 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3830</pre>
3831</div>
3832
3833<!-- _______________________________________________________________________ -->
3834<div class="doc_subsubsection">
3835 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3836</div>
3837
3838<div class="doc_text">
3839
3840<h5>Syntax:</h5>
3841
3842<pre>
3843 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3844</pre>
3845
3846<h5>Overview:</h5>
3847<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3848<tt>ty2</tt>.</p>
3849
3850
3851<h5>Arguments:</h5>
3852<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3853 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3854cast it to. The size of <tt>value</tt> must be larger than the size of
3855<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3856<i>no-op cast</i>.</p>
3857
3858<h5>Semantics:</h5>
3859<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3860<a href="#t_floating">floating point</a> type to a smaller
3861<a href="#t_floating">floating point</a> type. If the value cannot fit within
3862the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3863
3864<h5>Example:</h5>
3865<pre>
3866 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3867 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3868</pre>
3869</div>
3870
3871<!-- _______________________________________________________________________ -->
3872<div class="doc_subsubsection">
3873 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3874</div>
3875<div class="doc_text">
3876
3877<h5>Syntax:</h5>
3878<pre>
3879 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3880</pre>
3881
3882<h5>Overview:</h5>
3883<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3884floating point value.</p>
3885
3886<h5>Arguments:</h5>
3887<p>The '<tt>fpext</tt>' instruction takes a
3888<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3889and a <a href="#t_floating">floating point</a> type to cast it to. The source
3890type must be smaller than the destination type.</p>
3891
3892<h5>Semantics:</h5>
3893<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3894<a href="#t_floating">floating point</a> type to a larger
3895<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3896used to make a <i>no-op cast</i> because it always changes bits. Use
3897<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3898
3899<h5>Example:</h5>
3900<pre>
3901 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3902 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3903</pre>
3904</div>
3905
3906<!-- _______________________________________________________________________ -->
3907<div class="doc_subsubsection">
3908 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3909</div>
3910<div class="doc_text">
3911
3912<h5>Syntax:</h5>
3913<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3914 &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]3915</pre>
3916
3917<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3918<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3919unsigned integer equivalent of type <tt>ty2</tt>.
3920</p>
3921
3922<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3923<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]3924scalar or vector <a href="#t_floating">floating point</a> value, and a type
3925to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3926type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3927vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3928
3929<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3930<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3931<a href="#t_floating">floating point</a> operand into the nearest (rounding
3932towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3933the results are undefined.</p>
3934
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3935<h5>Example:</h5>
3936<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3937 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3938 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3939 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3940</pre>
3941</div>
3942
3943<!-- _______________________________________________________________________ -->
3944<div class="doc_subsubsection">
3945 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3946</div>
3947<div class="doc_text">
3948
3949<h5>Syntax:</h5>
3950<pre>
3951 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3952</pre>
3953
3954<h5>Overview:</h5>
3955<p>The '<tt>fptosi</tt>' instruction converts
3956<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3957</p>
3958
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3959<h5>Arguments:</h5>
3960<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]3961scalar or vector <a href="#t_floating">floating point</a> value, and a type
3962to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3963type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3964vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3965
3966<h5>Semantics:</h5>
3967<p>The '<tt>fptosi</tt>' instruction converts its
3968<a href="#t_floating">floating point</a> operand into the nearest (rounding
3969towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3970the results are undefined.</p>
3971
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3972<h5>Example:</h5>
3973<pre>
3974 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3975 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3976 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3977</pre>
3978</div>
3979
3980<!-- _______________________________________________________________________ -->
3981<div class="doc_subsubsection">
3982 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3983</div>
3984<div class="doc_text">
3985
3986<h5>Syntax:</h5>
3987<pre>
3988 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3989</pre>
3990
3991<h5>Overview:</h5>
3992<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3993integer and converts that value to the <tt>ty2</tt> type.</p>
3994
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3995<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3996<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3997scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3998to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3999type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4000floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4001
4002<h5>Semantics:</h5>
4003<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4004integer quantity and converts it to the corresponding floating point value. If
4005the value cannot fit in the floating point value, the results are undefined.</p>
4006
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4007<h5>Example:</h5>
4008<pre>
4009 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4010 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4011</pre>
4012</div>
4013
4014<!-- _______________________________________________________________________ -->
4015<div class="doc_subsubsection">
4016 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4017</div>
4018<div class="doc_text">
4019
4020<h5>Syntax:</h5>
4021<pre>
4022 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4023</pre>
4024
4025<h5>Overview:</h5>
4026<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
4027integer and converts that value to the <tt>ty2</tt> type.</p>
4028
4029<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]4030<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4031scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4032to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4033type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4034floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4035
4036<h5>Semantics:</h5>
4037<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
4038integer quantity and converts it to the corresponding floating point value. If
4039the value cannot fit in the floating point value, the results are undefined.</p>
4040
4041<h5>Example:</h5>
4042<pre>
4043 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4044 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4045</pre>
4046</div>
4047
4048<!-- _______________________________________________________________________ -->
4049<div class="doc_subsubsection">
4050 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4051</div>
4052<div class="doc_text">
4053
4054<h5>Syntax:</h5>
4055<pre>
4056 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4057</pre>
4058
4059<h5>Overview:</h5>
4060<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4061the integer type <tt>ty2</tt>.</p>
4062
4063<h5>Arguments:</h5>
4064<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4065must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4066<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4067
4068<h5>Semantics:</h5>
4069<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4070<tt>ty2</tt> by interpreting the pointer value as an integer and either
4071truncating or zero extending that value to the size of the integer type. If
4072<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4073<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4074are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4075change.</p>
4076
4077<h5>Example:</h5>
4078<pre>
4079 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4080 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4081</pre>
4082</div>
4083
4084<!-- _______________________________________________________________________ -->
4085<div class="doc_subsubsection">
4086 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4087</div>
4088<div class="doc_text">
4089
4090<h5>Syntax:</h5>
4091<pre>
4092 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4093</pre>
4094
4095<h5>Overview:</h5>
4096<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4097a pointer type, <tt>ty2</tt>.</p>
4098
4099<h5>Arguments:</h5>
4100<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4101value to cast, and a type to cast it to, which must be a
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4102<a href="#t_pointer">pointer</a> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4103
4104<h5>Semantics:</h5>
4105<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4106<tt>ty2</tt> by applying either a zero extension or a truncation depending on
4107the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4108size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4109the size of a pointer then a zero extension is done. If they are the same size,
4110nothing is done (<i>no-op cast</i>).</p>
4111
4112<h5>Example:</h5>
4113<pre>
4114 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4115 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4116 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4117</pre>
4118</div>
4119
4120<!-- _______________________________________________________________________ -->
4121<div class="doc_subsubsection">
4122 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4123</div>
4124<div class="doc_text">
4125
4126<h5>Syntax:</h5>
4127<pre>
4128 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4129</pre>
4130
4131<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4132
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4133<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4134<tt>ty2</tt> without changing any bits.</p>
4135
4136<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4137
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4138<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]4139a non-aggregate first class value, and a type to cast it to, which must also be
4140a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4141<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4142and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4143type is a pointer, the destination type must also be a pointer. This
4144instruction supports bitwise conversion of vectors to integers and to vectors
4145of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4146
4147<h5>Semantics:</h5>
4148<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4149<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4150this conversion. The conversion is done as if the <tt>value</tt> had been
4151stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4152converted to other pointer types with this instruction. To convert pointers to
4153other types, use the <a href="#i_inttoptr">inttoptr</a> or
4154<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4155
4156<h5>Example:</h5>
4157<pre>
4158 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4159 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4160 %Z = bitcast &lt;2 x int&gt; %V to i64; <i>; yields i64: %V</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4161</pre>
4162</div>
4163
4164<!-- ======================================================================= -->
4165<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4166<div class="doc_text">
4167<p>The instructions in this category are the "miscellaneous"
4168instructions, which defy better classification.</p>
4169</div>
4170
4171<!-- _______________________________________________________________________ -->
4172<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4173</div>
4174<div class="doc_text">
4175<h5>Syntax:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4176<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]4177</pre>
4178<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4179<p>The '<tt>icmp</tt>' instruction returns a boolean value or
4180a vector of boolean values based on comparison
4181of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4182<h5>Arguments:</h5>
4183<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4184the condition code indicating the kind of comparison to perform. It is not
4185a value, just a keyword. The possible condition code are:
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4186</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4187<ol>
4188 <li><tt>eq</tt>: equal</li>
4189 <li><tt>ne</tt>: not equal </li>
4190 <li><tt>ugt</tt>: unsigned greater than</li>
4191 <li><tt>uge</tt>: unsigned greater or equal</li>
4192 <li><tt>ult</tt>: unsigned less than</li>
4193 <li><tt>ule</tt>: unsigned less or equal</li>
4194 <li><tt>sgt</tt>: signed greater than</li>
4195 <li><tt>sge</tt>: signed greater or equal</li>
4196 <li><tt>slt</tt>: signed less than</li>
4197 <li><tt>sle</tt>: signed less or equal</li>
4198</ol>
4199<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4200<a href="#t_pointer">pointer</a>
4201or integer <a href="#t_vector">vector</a> typed.
4202They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4203<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4204<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]4205the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4206yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4207</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4208<ol>
4209 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4210 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4211 </li>
4212 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4213 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4214 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4215 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4216 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4217 <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]4218 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4219 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4220 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4221 <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]4222 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4223 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4224 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4225 <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]4226 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4227 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4228 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4229 <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]4230</ol>
4231<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4232values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4233<p>If the operands are integer vectors, then they are compared
4234element by element. The result is an <tt>i1</tt> vector with
4235the same number of elements as the values being compared.
4236Otherwise, the result is an <tt>i1</tt>.
4237</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4238
4239<h5>Example:</h5>
4240<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
4241 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
4242 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
4243 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4244 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4245 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
4246</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4247
4248<p>Note that the code generator does not yet support vector types with
4249 the <tt>icmp</tt> instruction.</p>
4250
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4251</div>
4252
4253<!-- _______________________________________________________________________ -->
4254<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4255</div>
4256<div class="doc_text">
4257<h5>Syntax:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4258<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]4259</pre>
4260<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4261<p>The '<tt>fcmp</tt>' instruction returns a boolean value
4262or vector of boolean values based on comparison
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4263of its operands.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4264<p>
4265If the operands are floating point scalars, then the result
4266type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4267</p>
4268<p>If the operands are floating point vectors, then the result type
4269is a vector of boolean with the same number of elements as the
4270operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4271<h5>Arguments:</h5>
4272<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4273the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4274a value, just a keyword. The possible condition code are:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4275<ol>
4276 <li><tt>false</tt>: no comparison, always returns false</li>
4277 <li><tt>oeq</tt>: ordered and equal</li>
4278 <li><tt>ogt</tt>: ordered and greater than </li>
4279 <li><tt>oge</tt>: ordered and greater than or equal</li>
4280 <li><tt>olt</tt>: ordered and less than </li>
4281 <li><tt>ole</tt>: ordered and less than or equal</li>
4282 <li><tt>one</tt>: ordered and not equal</li>
4283 <li><tt>ord</tt>: ordered (no nans)</li>
4284 <li><tt>ueq</tt>: unordered or equal</li>
4285 <li><tt>ugt</tt>: unordered or greater than </li>
4286 <li><tt>uge</tt>: unordered or greater than or equal</li>
4287 <li><tt>ult</tt>: unordered or less than </li>
4288 <li><tt>ule</tt>: unordered or less than or equal</li>
4289 <li><tt>une</tt>: unordered or not equal</li>
4290 <li><tt>uno</tt>: unordered (either nans)</li>
4291 <li><tt>true</tt>: no comparison, always returns true</li>
4292</ol>
4293<p><i>Ordered</i> means that neither operand is a QNAN while
4294<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4295<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4296either a <a href="#t_floating">floating point</a> type
4297or a <a href="#t_vector">vector</a> of floating point type.
4298They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4299<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4300<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4301according to the condition code given as <tt>cond</tt>.
4302If the operands are vectors, then the vectors are compared
4303element by element.
4304Each comparison performed
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4305always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4306<ol>
4307 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4308 <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]4309 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4310 <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]4311 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4312 <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]4313 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4314 <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]4315 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4316 <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]4317 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4318 <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]4319 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4320 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4321 <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]4322 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4323 <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]4324 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4325 <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]4326 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4327 <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]4328 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4329 <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]4330 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4331 <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]4332 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4333 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4334 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4335</ol>
4336
4337<h5>Example:</h5>
4338<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]4339 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4340 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4341 &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]4342</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4343
4344<p>Note that the code generator does not yet support vector types with
4345 the <tt>fcmp</tt> instruction.</p>
4346
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4347</div>
4348
4349<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4350<div class="doc_subsubsection">
4351 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4352</div>
4353<div class="doc_text">
4354<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4355<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]4356</pre>
4357<h5>Overview:</h5>
4358<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4359element-wise comparison of its two integer vector operands.</p>
4360<h5>Arguments:</h5>
4361<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4362the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4363a value, just a keyword. The possible condition code are:</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4364<ol>
4365 <li><tt>eq</tt>: equal</li>
4366 <li><tt>ne</tt>: not equal </li>
4367 <li><tt>ugt</tt>: unsigned greater than</li>
4368 <li><tt>uge</tt>: unsigned greater or equal</li>
4369 <li><tt>ult</tt>: unsigned less than</li>
4370 <li><tt>ule</tt>: unsigned less or equal</li>
4371 <li><tt>sgt</tt>: signed greater than</li>
4372 <li><tt>sge</tt>: signed greater or equal</li>
4373 <li><tt>slt</tt>: signed less than</li>
4374 <li><tt>sle</tt>: signed less or equal</li>
4375</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4376<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4377<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4378<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4379<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4380according to the condition code given as <tt>cond</tt>. The comparison yields a
4381<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4382identical type as the values being compared. The most significant bit in each
4383element is 1 if the element-wise comparison evaluates to true, and is 0
4384otherwise. All other bits of the result are undefined. The condition codes
4385are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4386instruction</a>.</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4387
4388<h5>Example:</h5>
4389<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4390 &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>
4391 &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]4392</pre>
4393</div>
4394
4395<!-- _______________________________________________________________________ -->
4396<div class="doc_subsubsection">
4397 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4398</div>
4399<div class="doc_text">
4400<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4401<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]4402<h5>Overview:</h5>
4403<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4404element-wise comparison of its two floating point vector operands. The output
4405elements have the same width as the input elements.</p>
4406<h5>Arguments:</h5>
4407<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4408the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4409a value, just a keyword. The possible condition code are:</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4410<ol>
4411 <li><tt>false</tt>: no comparison, always returns false</li>
4412 <li><tt>oeq</tt>: ordered and equal</li>
4413 <li><tt>ogt</tt>: ordered and greater than </li>
4414 <li><tt>oge</tt>: ordered and greater than or equal</li>
4415 <li><tt>olt</tt>: ordered and less than </li>
4416 <li><tt>ole</tt>: ordered and less than or equal</li>
4417 <li><tt>one</tt>: ordered and not equal</li>
4418 <li><tt>ord</tt>: ordered (no nans)</li>
4419 <li><tt>ueq</tt>: unordered or equal</li>
4420 <li><tt>ugt</tt>: unordered or greater than </li>
4421 <li><tt>uge</tt>: unordered or greater than or equal</li>
4422 <li><tt>ult</tt>: unordered or less than </li>
4423 <li><tt>ule</tt>: unordered or less than or equal</li>
4424 <li><tt>une</tt>: unordered or not equal</li>
4425 <li><tt>uno</tt>: unordered (either nans)</li>
4426 <li><tt>true</tt>: no comparison, always returns true</li>
4427</ol>
4428<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4429<a href="#t_floating">floating point</a> typed. They must also be identical
4430types.</p>
4431<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4432<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4433according to the condition code given as <tt>cond</tt>. The comparison yields a
4434<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4435an identical number of elements as the values being compared, and each element
4436having identical with to the width of the floating point elements. The most
4437significant bit in each element is 1 if the element-wise comparison evaluates to
4438true, and is 0 otherwise. All other bits of the result are undefined. The
4439condition codes are evaluated identically to the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4440<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4441
4442<h5>Example:</h5>
4443<pre>
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000[diff] [blame]4444 <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4445 &lt;result&gt; = vfcmp oeq &lt;2 x float&gt; &lt; float 4, float 0 &gt;, &lt; float 5, float 0 &gt;
4446
4447 <i>; yields: result=&lt;2 x i64&gt; &lt; i64 -1, i64 0 &gt;</i>
4448 &lt;result&gt; = vfcmp ult &lt;2 x double&gt; &lt; double 1, double 2 &gt;, &lt; double 2, double 2&gt;
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4449</pre>
4450</div>
4451
4452<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4453<div class="doc_subsubsection">
4454 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4455</div>
4456
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4457<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4458
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4459<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4460
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4461<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4462<h5>Overview:</h5>
4463<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4464the SSA graph representing the function.</p>
4465<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4466
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4467<p>The type of the incoming values is specified with the first type
4468field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4469as arguments, with one pair for each predecessor basic block of the
4470current block. Only values of <a href="#t_firstclass">first class</a>
4471type may be used as the value arguments to the PHI node. Only labels
4472may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4473
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4474<p>There must be no non-phi instructions between the start of a basic
4475block and the PHI instructions: i.e. PHI instructions must be first in
4476a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4477
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4478<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4479
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4480<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4481specified by the pair corresponding to the predecessor basic block that executed
4482just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4483
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4484<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4485<pre>
4486Loop: ; Infinite loop that counts from 0 on up...
4487 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4488 %nextindvar = add i32 %indvar, 1
4489 br label %Loop
4490</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4491</div>
4492
4493<!-- _______________________________________________________________________ -->
4494<div class="doc_subsubsection">
4495 <a name="i_select">'<tt>select</tt>' Instruction</a>
4496</div>
4497
4498<div class="doc_text">
4499
4500<h5>Syntax:</h5>
4501
4502<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4503 &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>
4504
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4505 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4506</pre>
4507
4508<h5>Overview:</h5>
4509
4510<p>
4511The '<tt>select</tt>' instruction is used to choose one value based on a
4512condition, without branching.
4513</p>
4514
4515
4516<h5>Arguments:</h5>
4517
4518<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4519The '<tt>select</tt>' instruction requires an 'i1' value or
4520a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4521condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4522type. If the val1/val2 are vectors and
4523the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4524individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4525</p>
4526
4527<h5>Semantics:</h5>
4528
4529<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4530If 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]4531value argument; otherwise, it returns the second value argument.
4532</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4533<p>
4534If the condition is a vector of i1, then the value arguments must
4535be vectors of the same size, and the selection is done element
4536by element.
4537</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4538
4539<h5>Example:</h5>
4540
4541<pre>
4542 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4543</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4544
4545<p>Note that the code generator does not yet support conditions
4546 with vector type.</p>
4547
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4548</div>
4549
4550
4551<!-- _______________________________________________________________________ -->
4552<div class="doc_subsubsection">
4553 <a name="i_call">'<tt>call</tt>' Instruction</a>
4554</div>
4555
4556<div class="doc_text">
4557
4558<h5>Syntax:</h5>
4559<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4560 &lt;result&gt; = [tail] call [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] &lt;ty&gt; [&lt;fnty&gt;*] &lt;fnptrval&gt;(&lt;function args&gt;) [<a href="#fnattrs">fn attrs</a>]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4561</pre>
4562
4563<h5>Overview:</h5>
4564
4565<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4566
4567<h5>Arguments:</h5>
4568
4569<p>This instruction requires several arguments:</p>
4570
4571<ol>
4572 <li>
4573 <p>The optional "tail" marker indicates whether the callee function accesses
4574 any allocas or varargs in the caller. If the "tail" marker is present, the
4575 function call is eligible for tail call optimization. Note that calls may
4576 be marked "tail" even if they do not occur before a <a
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4577 href="#i_ret"><tt>ret</tt></a> instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4578 </li>
4579 <li>
4580 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4581 convention</a> the call should use. If none is specified, the call defaults
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4582 to using C calling conventions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4583 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4584
4585 <li>
4586 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4587 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4588 and '<tt>inreg</tt>' attributes are valid here.</p>
4589 </li>
4590
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4591 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4592 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4593 the type of the return value. Functions that return no value are marked
4594 <tt><a href="#t_void">void</a></tt>.</p>
4595 </li>
4596 <li>
4597 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4598 value being invoked. The argument types must match the types implied by
4599 this signature. This type can be omitted if the function is not varargs
4600 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4601 </li>
4602 <li>
4603 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4604 be invoked. In most cases, this is a direct function invocation, but
4605 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4606 to function value.</p>
4607 </li>
4608 <li>
4609 <p>'<tt>function args</tt>': argument list whose types match the
4610 function signature argument types. All arguments must be of
4611 <a href="#t_firstclass">first class</a> type. If the function signature
4612 indicates the function accepts a variable number of arguments, the extra
4613 arguments can be specified.</p>
4614 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4615 <li>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4616 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4617 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4618 '<tt>readnone</tt>' attributes are valid here.</p>
4619 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4620</ol>
4621
4622<h5>Semantics:</h5>
4623
4624<p>The '<tt>call</tt>' instruction is used to cause control flow to
4625transfer to a specified function, with its incoming arguments bound to
4626the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4627instruction in the called function, control flow continues with the
4628instruction after the function call, and the return value of the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4629function is bound to the result argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4630
4631<h5>Example:</h5>
4632
4633<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4634 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4635 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4636 %X = tail call i32 @foo() <i>; yields i32</i>
4637 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4638 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4639
4640 %struct.A = type { i32, i8 }
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4641 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4642 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4643 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattnerac454b32008-10-08 06:26:11 +0000[diff] [blame]4644 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijman2c4e05a2008-10-07 10:03:45 +0000[diff] [blame]4645 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4646</pre>
4647
4648</div>
4649
4650<!-- _______________________________________________________________________ -->
4651<div class="doc_subsubsection">
4652 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4653</div>
4654
4655<div class="doc_text">
4656
4657<h5>Syntax:</h5>
4658
4659<pre>
4660 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4661</pre>
4662
4663<h5>Overview:</h5>
4664
4665<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4666the "variable argument" area of a function call. It is used to implement the
4667<tt>va_arg</tt> macro in C.</p>
4668
4669<h5>Arguments:</h5>
4670
4671<p>This instruction takes a <tt>va_list*</tt> value and the type of
4672the argument. It returns a value of the specified argument type and
4673increments the <tt>va_list</tt> to point to the next argument. The
4674actual type of <tt>va_list</tt> is target specific.</p>
4675
4676<h5>Semantics:</h5>
4677
4678<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4679type from the specified <tt>va_list</tt> and causes the
4680<tt>va_list</tt> to point to the next argument. For more information,
4681see the variable argument handling <a href="#int_varargs">Intrinsic
4682Functions</a>.</p>
4683
4684<p>It is legal for this instruction to be called in a function which does not
4685take a variable number of arguments, for example, the <tt>vfprintf</tt>
4686function.</p>
4687
4688<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4689href="#intrinsics">intrinsic function</a> because it takes a type as an
4690argument.</p>
4691
4692<h5>Example:</h5>
4693
4694<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4695
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]4696<p>Note that the code generator does not yet fully support va_arg
4697 on many targets. Also, it does not currently support va_arg with
4698 aggregate types on any target.</p>
4699
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4700</div>
4701
4702<!-- *********************************************************************** -->
4703<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4704<!-- *********************************************************************** -->
4705
4706<div class="doc_text">
4707
4708<p>LLVM supports the notion of an "intrinsic function". These functions have
4709well known names and semantics and are required to follow certain restrictions.
4710Overall, these intrinsics represent an extension mechanism for the LLVM
4711language that does not require changing all of the transformations in LLVM when
4712adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4713
4714<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4715prefix is reserved in LLVM for intrinsic names; thus, function names may not
4716begin with this prefix. Intrinsic functions must always be external functions:
4717you cannot define the body of intrinsic functions. Intrinsic functions may
4718only be used in call or invoke instructions: it is illegal to take the address
4719of an intrinsic function. Additionally, because intrinsic functions are part
4720of the LLVM language, it is required if any are added that they be documented
4721here.</p>
4722
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4723<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4724a family of functions that perform the same operation but on different data
4725types. Because LLVM can represent over 8 million different integer types,
4726overloading is used commonly to allow an intrinsic function to operate on any
4727integer type. One or more of the argument types or the result type can be
4728overloaded to accept any integer type. Argument types may also be defined as
4729exactly matching a previous argument's type or the result type. This allows an
4730intrinsic function which accepts multiple arguments, but needs all of them to
4731be of the same type, to only be overloaded with respect to a single argument or
4732the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4733
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4734<p>Overloaded intrinsics will have the names of its overloaded argument types
4735encoded into its function name, each preceded by a period. Only those types
4736which are overloaded result in a name suffix. Arguments whose type is matched
4737against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4738take an integer of any width and returns an integer of exactly the same integer
4739width. This leads to a family of functions such as
4740<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4741Only one type, the return type, is overloaded, and only one type suffix is
4742required. Because the argument's type is matched against the return type, it
4743does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4744
4745<p>To learn how to add an intrinsic function, please see the
4746<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4747</p>
4748
4749</div>
4750
4751<!-- ======================================================================= -->
4752<div class="doc_subsection">
4753 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4754</div>
4755
4756<div class="doc_text">
4757
4758<p>Variable argument support is defined in LLVM with the <a
4759 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4760intrinsic functions. These functions are related to the similarly
4761named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4762
4763<p>All of these functions operate on arguments that use a
4764target-specific value type "<tt>va_list</tt>". The LLVM assembly
4765language reference manual does not define what this type is, so all
4766transformations should be prepared to handle these functions regardless of
4767the type used.</p>
4768
4769<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4770instruction and the variable argument handling intrinsic functions are
4771used.</p>
4772
4773<div class="doc_code">
4774<pre>
4775define i32 @test(i32 %X, ...) {
4776 ; Initialize variable argument processing
4777 %ap = alloca i8*
4778 %ap2 = bitcast i8** %ap to i8*
4779 call void @llvm.va_start(i8* %ap2)
4780
4781 ; Read a single integer argument
4782 %tmp = va_arg i8** %ap, i32
4783
4784 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4785 %aq = alloca i8*
4786 %aq2 = bitcast i8** %aq to i8*
4787 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4788 call void @llvm.va_end(i8* %aq2)
4789
4790 ; Stop processing of arguments.
4791 call void @llvm.va_end(i8* %ap2)
4792 ret i32 %tmp
4793}
4794
4795declare void @llvm.va_start(i8*)
4796declare void @llvm.va_copy(i8*, i8*)
4797declare void @llvm.va_end(i8*)
4798</pre>
4799</div>
4800
4801</div>
4802
4803<!-- _______________________________________________________________________ -->
4804<div class="doc_subsubsection">
4805 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4806</div>
4807
4808
4809<div class="doc_text">
4810<h5>Syntax:</h5>
4811<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4812<h5>Overview:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4813<p>The '<tt>llvm.va_start</tt>' intrinsic initializes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4814<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4815href="#i_va_arg">va_arg</a></tt>.</p>
4816
4817<h5>Arguments:</h5>
4818
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4819<p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4820
4821<h5>Semantics:</h5>
4822
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4823<p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4824macro available in C. In a target-dependent way, it initializes the
4825<tt>va_list</tt> element to which the argument points, so that the next call to
4826<tt>va_arg</tt> will produce the first variable argument passed to the function.
4827Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4828last argument of the function as the compiler can figure that out.</p>
4829
4830</div>
4831
4832<!-- _______________________________________________________________________ -->
4833<div class="doc_subsubsection">
4834 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4835</div>
4836
4837<div class="doc_text">
4838<h5>Syntax:</h5>
4839<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4840<h5>Overview:</h5>
4841
4842<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4843which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4844or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4845
4846<h5>Arguments:</h5>
4847
4848<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4849
4850<h5>Semantics:</h5>
4851
4852<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4853macro available in C. In a target-dependent way, it destroys the
4854<tt>va_list</tt> element to which the argument points. Calls to <a
4855href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4856<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4857<tt>llvm.va_end</tt>.</p>
4858
4859</div>
4860
4861<!-- _______________________________________________________________________ -->
4862<div class="doc_subsubsection">
4863 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4864</div>
4865
4866<div class="doc_text">
4867
4868<h5>Syntax:</h5>
4869
4870<pre>
4871 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4872</pre>
4873
4874<h5>Overview:</h5>
4875
4876<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4877from the source argument list to the destination argument list.</p>
4878
4879<h5>Arguments:</h5>
4880
4881<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4882The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4883
4884
4885<h5>Semantics:</h5>
4886
4887<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4888macro available in C. In a target-dependent way, it copies the source
4889<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4890intrinsic is necessary because the <tt><a href="#int_va_start">
4891llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4892example, memory allocation.</p>
4893
4894</div>
4895
4896<!-- ======================================================================= -->
4897<div class="doc_subsection">
4898 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4899</div>
4900
4901<div class="doc_text">
4902
4903<p>
4904LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4905Collection</a> (GC) requires the implementation and generation of these
4906intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4907These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4908stack</a>, as well as garbage collector implementations that require <a
4909href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4910Front-ends for type-safe garbage collected languages should generate these
4911intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4912href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4913</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4914
4915<p>The garbage collection intrinsics only operate on objects in the generic
4916 address space (address space zero).</p>
4917
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4918</div>
4919
4920<!-- _______________________________________________________________________ -->
4921<div class="doc_subsubsection">
4922 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4923</div>
4924
4925<div class="doc_text">
4926
4927<h5>Syntax:</h5>
4928
4929<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4930 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4931</pre>
4932
4933<h5>Overview:</h5>
4934
4935<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4936the code generator, and allows some metadata to be associated with it.</p>
4937
4938<h5>Arguments:</h5>
4939
4940<p>The first argument specifies the address of a stack object that contains the
4941root pointer. The second pointer (which must be either a constant or a global
4942value address) contains the meta-data to be associated with the root.</p>
4943
4944<h5>Semantics:</h5>
4945
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4946<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]4947location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4948the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4949intrinsic may only be used in a function which <a href="#gc">specifies a GC
4950algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4951
4952</div>
4953
4954
4955<!-- _______________________________________________________________________ -->
4956<div class="doc_subsubsection">
4957 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4958</div>
4959
4960<div class="doc_text">
4961
4962<h5>Syntax:</h5>
4963
4964<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4965 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4966</pre>
4967
4968<h5>Overview:</h5>
4969
4970<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4971locations, allowing garbage collector implementations that require read
4972barriers.</p>
4973
4974<h5>Arguments:</h5>
4975
4976<p>The second argument is the address to read from, which should be an address
4977allocated from the garbage collector. The first object is a pointer to the
4978start of the referenced object, if needed by the language runtime (otherwise
4979null).</p>
4980
4981<h5>Semantics:</h5>
4982
4983<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4984instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4985garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4986may only be used in a function which <a href="#gc">specifies a GC
4987algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4988
4989</div>
4990
4991
4992<!-- _______________________________________________________________________ -->
4993<div class="doc_subsubsection">
4994 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4995</div>
4996
4997<div class="doc_text">
4998
4999<h5>Syntax:</h5>
5000
5001<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5002 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5003</pre>
5004
5005<h5>Overview:</h5>
5006
5007<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5008locations, allowing garbage collector implementations that require write
5009barriers (such as generational or reference counting collectors).</p>
5010
5011<h5>Arguments:</h5>
5012
5013<p>The first argument is the reference to store, the second is the start of the
5014object to store it to, and the third is the address of the field of Obj to
5015store to. If the runtime does not require a pointer to the object, Obj may be
5016null.</p>
5017
5018<h5>Semantics:</h5>
5019
5020<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5021instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]5022garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5023may only be used in a function which <a href="#gc">specifies a GC
5024algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5025
5026</div>
5027
5028
5029
5030<!-- ======================================================================= -->
5031<div class="doc_subsection">
5032 <a name="int_codegen">Code Generator Intrinsics</a>
5033</div>
5034
5035<div class="doc_text">
5036<p>
5037These intrinsics are provided by LLVM to expose special features that may only
5038be implemented with code generator support.
5039</p>
5040
5041</div>
5042
5043<!-- _______________________________________________________________________ -->
5044<div class="doc_subsubsection">
5045 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5046</div>
5047
5048<div class="doc_text">
5049
5050<h5>Syntax:</h5>
5051<pre>
5052 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
5053</pre>
5054
5055<h5>Overview:</h5>
5056
5057<p>
5058The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5059target-specific value indicating the return address of the current function
5060or one of its callers.
5061</p>
5062
5063<h5>Arguments:</h5>
5064
5065<p>
5066The argument to this intrinsic indicates which function to return the address
5067for. Zero indicates the calling function, one indicates its caller, etc. The
5068argument is <b>required</b> to be a constant integer value.
5069</p>
5070
5071<h5>Semantics:</h5>
5072
5073<p>
5074The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
5075the return address of the specified call frame, or zero if it cannot be
5076identified. The value returned by this intrinsic is likely to be incorrect or 0
5077for arguments other than zero, so it should only be used for debugging purposes.
5078</p>
5079
5080<p>
5081Note that calling this intrinsic does not prevent function inlining or other
5082aggressive transformations, so the value returned may not be that of the obvious
5083source-language caller.
5084</p>
5085</div>
5086
5087
5088<!-- _______________________________________________________________________ -->
5089<div class="doc_subsubsection">
5090 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5091</div>
5092
5093<div class="doc_text">
5094
5095<h5>Syntax:</h5>
5096<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5097 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5098</pre>
5099
5100<h5>Overview:</h5>
5101
5102<p>
5103The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5104target-specific frame pointer value for the specified stack frame.
5105</p>
5106
5107<h5>Arguments:</h5>
5108
5109<p>
5110The argument to this intrinsic indicates which function to return the frame
5111pointer for. Zero indicates the calling function, one indicates its caller,
5112etc. The argument is <b>required</b> to be a constant integer value.
5113</p>
5114
5115<h5>Semantics:</h5>
5116
5117<p>
5118The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5119the frame address of the specified call frame, or zero if it cannot be
5120identified. The value returned by this intrinsic is likely to be incorrect or 0
5121for arguments other than zero, so it should only be used for debugging purposes.
5122</p>
5123
5124<p>
5125Note that calling this intrinsic does not prevent function inlining or other
5126aggressive transformations, so the value returned may not be that of the obvious
5127source-language caller.
5128</p>
5129</div>
5130
5131<!-- _______________________________________________________________________ -->
5132<div class="doc_subsubsection">
5133 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5134</div>
5135
5136<div class="doc_text">
5137
5138<h5>Syntax:</h5>
5139<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5140 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5141</pre>
5142
5143<h5>Overview:</h5>
5144
5145<p>
5146The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
5147the function stack, for use with <a href="#int_stackrestore">
5148<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5149features like scoped automatic variable sized arrays in C99.
5150</p>
5151
5152<h5>Semantics:</h5>
5153
5154<p>
5155This intrinsic returns a opaque pointer value that can be passed to <a
5156href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
5157<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5158<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5159state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5160practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5161that were allocated after the <tt>llvm.stacksave</tt> was executed.
5162</p>
5163
5164</div>
5165
5166<!-- _______________________________________________________________________ -->
5167<div class="doc_subsubsection">
5168 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5169</div>
5170
5171<div class="doc_text">
5172
5173<h5>Syntax:</h5>
5174<pre>
5175 declare void @llvm.stackrestore(i8 * %ptr)
5176</pre>
5177
5178<h5>Overview:</h5>
5179
5180<p>
5181The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5182the function stack to the state it was in when the corresponding <a
5183href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
5184useful for implementing language features like scoped automatic variable sized
5185arrays in C99.
5186</p>
5187
5188<h5>Semantics:</h5>
5189
5190<p>
5191See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
5192</p>
5193
5194</div>
5195
5196
5197<!-- _______________________________________________________________________ -->
5198<div class="doc_subsubsection">
5199 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5200</div>
5201
5202<div class="doc_text">
5203
5204<h5>Syntax:</h5>
5205<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5206 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]5207</pre>
5208
5209<h5>Overview:</h5>
5210
5211
5212<p>
5213The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5214a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5215no
5216effect on the behavior of the program but can change its performance
5217characteristics.
5218</p>
5219
5220<h5>Arguments:</h5>
5221
5222<p>
5223<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5224determining if the fetch should be for a read (0) or write (1), and
5225<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5226locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5227<tt>locality</tt> arguments must be constant integers.
5228</p>
5229
5230<h5>Semantics:</h5>
5231
5232<p>
5233This intrinsic does not modify the behavior of the program. In particular,
5234prefetches cannot trap and do not produce a value. On targets that support this
5235intrinsic, the prefetch can provide hints to the processor cache for better
5236performance.
5237</p>
5238
5239</div>
5240
5241<!-- _______________________________________________________________________ -->
5242<div class="doc_subsubsection">
5243 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5244</div>
5245
5246<div class="doc_text">
5247
5248<h5>Syntax:</h5>
5249<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5250 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5251</pre>
5252
5253<h5>Overview:</h5>
5254
5255
5256<p>
5257The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5258(PC) in a region of
5259code to simulators and other tools. The method is target specific, but it is
5260expected that the marker will use exported symbols to transmit the PC of the
5261marker.
5262The marker makes no guarantees that it will remain with any specific instruction
5263after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5264optimizations. The intended use is to be inserted after optimizations to allow
5265correlations of simulation runs.
5266</p>
5267
5268<h5>Arguments:</h5>
5269
5270<p>
5271<tt>id</tt> is a numerical id identifying the marker.
5272</p>
5273
5274<h5>Semantics:</h5>
5275
5276<p>
5277This intrinsic does not modify the behavior of the program. Backends that do not
5278support this intrinisic may ignore it.
5279</p>
5280
5281</div>
5282
5283<!-- _______________________________________________________________________ -->
5284<div class="doc_subsubsection">
5285 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5286</div>
5287
5288<div class="doc_text">
5289
5290<h5>Syntax:</h5>
5291<pre>
5292 declare i64 @llvm.readcyclecounter( )
5293</pre>
5294
5295<h5>Overview:</h5>
5296
5297
5298<p>
5299The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5300counter register (or similar low latency, high accuracy clocks) on those targets
5301that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5302As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5303should only be used for small timings.
5304</p>
5305
5306<h5>Semantics:</h5>
5307
5308<p>
5309When directly supported, reading the cycle counter should not modify any memory.
5310Implementations are allowed to either return a application specific value or a
5311system wide value. On backends without support, this is lowered to a constant 0.
5312</p>
5313
5314</div>
5315
5316<!-- ======================================================================= -->
5317<div class="doc_subsection">
5318 <a name="int_libc">Standard C Library Intrinsics</a>
5319</div>
5320
5321<div class="doc_text">
5322<p>
5323LLVM provides intrinsics for a few important standard C library functions.
5324These intrinsics allow source-language front-ends to pass information about the
5325alignment of the pointer arguments to the code generator, providing opportunity
5326for more efficient code generation.
5327</p>
5328
5329</div>
5330
5331<!-- _______________________________________________________________________ -->
5332<div class="doc_subsubsection">
5333 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5334</div>
5335
5336<div class="doc_text">
5337
5338<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5339<p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5340width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5341<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5342 declare void @llvm.memcpy.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5343 i8 &lt;len&gt;, i32 &lt;align&gt;)
5344 declare void @llvm.memcpy.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5345 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5346 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5347 i32 &lt;len&gt;, i32 &lt;align&gt;)
5348 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5349 i64 &lt;len&gt;, i32 &lt;align&gt;)
5350</pre>
5351
5352<h5>Overview:</h5>
5353
5354<p>
5355The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5356location to the destination location.
5357</p>
5358
5359<p>
5360Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5361intrinsics do not return a value, and takes an extra alignment argument.
5362</p>
5363
5364<h5>Arguments:</h5>
5365
5366<p>
5367The first argument is a pointer to the destination, the second is a pointer to
5368the source. The third argument is an integer argument
5369specifying the number of bytes to copy, and the fourth argument is the alignment
5370of the source and destination locations.
5371</p>
5372
5373<p>
5374If the call to this intrinisic has an alignment value that is not 0 or 1, then
5375the caller guarantees that both the source and destination pointers are aligned
5376to that boundary.
5377</p>
5378
5379<h5>Semantics:</h5>
5380
5381<p>
5382The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5383location to the destination location, which are not allowed to overlap. It
5384copies "len" bytes of memory over. If the argument is known to be aligned to
5385some boundary, this can be specified as the fourth argument, otherwise it should
5386be set to 0 or 1.
5387</p>
5388</div>
5389
5390
5391<!-- _______________________________________________________________________ -->
5392<div class="doc_subsubsection">
5393 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5394</div>
5395
5396<div class="doc_text">
5397
5398<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5399<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5400width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5401<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5402 declare void @llvm.memmove.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5403 i8 &lt;len&gt;, i32 &lt;align&gt;)
5404 declare void @llvm.memmove.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5405 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5406 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5407 i32 &lt;len&gt;, i32 &lt;align&gt;)
5408 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5409 i64 &lt;len&gt;, i32 &lt;align&gt;)
5410</pre>
5411
5412<h5>Overview:</h5>
5413
5414<p>
5415The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5416location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5417'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5418</p>
5419
5420<p>
5421Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5422intrinsics do not return a value, and takes an extra alignment argument.
5423</p>
5424
5425<h5>Arguments:</h5>
5426
5427<p>
5428The first argument is a pointer to the destination, the second is a pointer to
5429the source. The third argument is an integer argument
5430specifying the number of bytes to copy, and the fourth argument is the alignment
5431of the source and destination locations.
5432</p>
5433
5434<p>
5435If the call to this intrinisic has an alignment value that is not 0 or 1, then
5436the caller guarantees that the source and destination pointers are aligned to
5437that boundary.
5438</p>
5439
5440<h5>Semantics:</h5>
5441
5442<p>
5443The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5444location to the destination location, which may overlap. It
5445copies "len" bytes of memory over. If the argument is known to be aligned to
5446some boundary, this can be specified as the fourth argument, otherwise it should
5447be set to 0 or 1.
5448</p>
5449</div>
5450
5451
5452<!-- _______________________________________________________________________ -->
5453<div class="doc_subsubsection">
5454 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5455</div>
5456
5457<div class="doc_text">
5458
5459<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5460<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5461width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5462<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5463 declare void @llvm.memset.i8(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5464 i8 &lt;len&gt;, i32 &lt;align&gt;)
5465 declare void @llvm.memset.i16(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5466 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5467 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5468 i32 &lt;len&gt;, i32 &lt;align&gt;)
5469 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5470 i64 &lt;len&gt;, i32 &lt;align&gt;)
5471</pre>
5472
5473<h5>Overview:</h5>
5474
5475<p>
5476The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5477byte value.
5478</p>
5479
5480<p>
5481Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5482does not return a value, and takes an extra alignment argument.
5483</p>
5484
5485<h5>Arguments:</h5>
5486
5487<p>
5488The first argument is a pointer to the destination to fill, the second is the
5489byte value to fill it with, the third argument is an integer
5490argument specifying the number of bytes to fill, and the fourth argument is the
5491known alignment of destination location.
5492</p>
5493
5494<p>
5495If the call to this intrinisic has an alignment value that is not 0 or 1, then
5496the caller guarantees that the destination pointer is aligned to that boundary.
5497</p>
5498
5499<h5>Semantics:</h5>
5500
5501<p>
5502The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5503the
5504destination location. If the argument is known to be aligned to some boundary,
5505this can be specified as the fourth argument, otherwise it should be set to 0 or
55061.
5507</p>
5508</div>
5509
5510
5511<!-- _______________________________________________________________________ -->
5512<div class="doc_subsubsection">
5513 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5514</div>
5515
5516<div class="doc_text">
5517
5518<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5519<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]5520floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5521types however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5522<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5523 declare float @llvm.sqrt.f32(float %Val)
5524 declare double @llvm.sqrt.f64(double %Val)
5525 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5526 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5527 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5528</pre>
5529
5530<h5>Overview:</h5>
5531
5532<p>
5533The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5534returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5535<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5536negative numbers other than -0.0 (which allows for better optimization, because
5537there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5538defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5539</p>
5540
5541<h5>Arguments:</h5>
5542
5543<p>
5544The argument and return value are floating point numbers of the same type.
5545</p>
5546
5547<h5>Semantics:</h5>
5548
5549<p>
5550This function returns the sqrt of the specified operand if it is a nonnegative
5551floating point number.
5552</p>
5553</div>
5554
5555<!-- _______________________________________________________________________ -->
5556<div class="doc_subsubsection">
5557 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5558</div>
5559
5560<div class="doc_text">
5561
5562<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5563<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]5564floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5565types however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5566<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5567 declare float @llvm.powi.f32(float %Val, i32 %power)
5568 declare double @llvm.powi.f64(double %Val, i32 %power)
5569 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5570 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5571 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5572</pre>
5573
5574<h5>Overview:</h5>
5575
5576<p>
5577The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5578specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5579multiplications is not defined. When a vector of floating point type is
5580used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5581</p>
5582
5583<h5>Arguments:</h5>
5584
5585<p>
5586The second argument is an integer power, and the first is a value to raise to
5587that power.
5588</p>
5589
5590<h5>Semantics:</h5>
5591
5592<p>
5593This function returns the first value raised to the second power with an
5594unspecified sequence of rounding operations.</p>
5595</div>
5596
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5597<!-- _______________________________________________________________________ -->
5598<div class="doc_subsubsection">
5599 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5600</div>
5601
5602<div class="doc_text">
5603
5604<h5>Syntax:</h5>
5605<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5606floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5607types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5608<pre>
5609 declare float @llvm.sin.f32(float %Val)
5610 declare double @llvm.sin.f64(double %Val)
5611 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5612 declare fp128 @llvm.sin.f128(fp128 %Val)
5613 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5614</pre>
5615
5616<h5>Overview:</h5>
5617
5618<p>
5619The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5620</p>
5621
5622<h5>Arguments:</h5>
5623
5624<p>
5625The argument and return value are floating point numbers of the same type.
5626</p>
5627
5628<h5>Semantics:</h5>
5629
5630<p>
5631This function returns the sine of the specified operand, returning the
5632same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5633conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5634</div>
5635
5636<!-- _______________________________________________________________________ -->
5637<div class="doc_subsubsection">
5638 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5639</div>
5640
5641<div class="doc_text">
5642
5643<h5>Syntax:</h5>
5644<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5645floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5646types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5647<pre>
5648 declare float @llvm.cos.f32(float %Val)
5649 declare double @llvm.cos.f64(double %Val)
5650 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5651 declare fp128 @llvm.cos.f128(fp128 %Val)
5652 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5653</pre>
5654
5655<h5>Overview:</h5>
5656
5657<p>
5658The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5659</p>
5660
5661<h5>Arguments:</h5>
5662
5663<p>
5664The argument and return value are floating point numbers of the same type.
5665</p>
5666
5667<h5>Semantics:</h5>
5668
5669<p>
5670This function returns the cosine of the specified operand, returning the
5671same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5672conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5673</div>
5674
5675<!-- _______________________________________________________________________ -->
5676<div class="doc_subsubsection">
5677 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5678</div>
5679
5680<div class="doc_text">
5681
5682<h5>Syntax:</h5>
5683<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5684floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5685types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5686<pre>
5687 declare float @llvm.pow.f32(float %Val, float %Power)
5688 declare double @llvm.pow.f64(double %Val, double %Power)
5689 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5690 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5691 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5692</pre>
5693
5694<h5>Overview:</h5>
5695
5696<p>
5697The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5698specified (positive or negative) power.
5699</p>
5700
5701<h5>Arguments:</h5>
5702
5703<p>
5704The second argument is a floating point power, and the first is a value to
5705raise to that power.
5706</p>
5707
5708<h5>Semantics:</h5>
5709
5710<p>
5711This function returns the first value raised to the second power,
5712returning the
5713same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5714conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5715</div>
5716
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5717
5718<!-- ======================================================================= -->
5719<div class="doc_subsection">
5720 <a name="int_manip">Bit Manipulation Intrinsics</a>
5721</div>
5722
5723<div class="doc_text">
5724<p>
5725LLVM provides intrinsics for a few important bit manipulation operations.
5726These allow efficient code generation for some algorithms.
5727</p>
5728
5729</div>
5730
5731<!-- _______________________________________________________________________ -->
5732<div class="doc_subsubsection">
5733 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5734</div>
5735
5736<div class="doc_text">
5737
5738<h5>Syntax:</h5>
5739<p>This is an overloaded intrinsic function. You can use bswap on any integer
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5740type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5741<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5742 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5743 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5744 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5745</pre>
5746
5747<h5>Overview:</h5>
5748
5749<p>
5750The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5751values with an even number of bytes (positive multiple of 16 bits). These are
5752useful for performing operations on data that is not in the target's native
5753byte order.
5754</p>
5755
5756<h5>Semantics:</h5>
5757
5758<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5759The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5760and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5761intrinsic returns an i32 value that has the four bytes of the input i32
5762swapped, 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]5763i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5764<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5765additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5766</p>
5767
5768</div>
5769
5770<!-- _______________________________________________________________________ -->
5771<div class="doc_subsubsection">
5772 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5773</div>
5774
5775<div class="doc_text">
5776
5777<h5>Syntax:</h5>
5778<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5779width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5780<pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5781 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5782 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5783 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5784 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5785 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5786</pre>
5787
5788<h5>Overview:</h5>
5789
5790<p>
5791The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5792value.
5793</p>
5794
5795<h5>Arguments:</h5>
5796
5797<p>
5798The only argument is the value to be counted. The argument may be of any
5799integer type. The return type must match the argument type.
5800</p>
5801
5802<h5>Semantics:</h5>
5803
5804<p>
5805The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5806</p>
5807</div>
5808
5809<!-- _______________________________________________________________________ -->
5810<div class="doc_subsubsection">
5811 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5812</div>
5813
5814<div class="doc_text">
5815
5816<h5>Syntax:</h5>
5817<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5818integer bit width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5819<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5820 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5821 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5822 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5823 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5824 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5825</pre>
5826
5827<h5>Overview:</h5>
5828
5829<p>
5830The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5831leading zeros in a variable.
5832</p>
5833
5834<h5>Arguments:</h5>
5835
5836<p>
5837The only argument is the value to be counted. The argument may be of any
5838integer type. The return type must match the argument type.
5839</p>
5840
5841<h5>Semantics:</h5>
5842
5843<p>
5844The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5845in a variable. If the src == 0 then the result is the size in bits of the type
5846of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5847</p>
5848</div>
5849
5850
5851
5852<!-- _______________________________________________________________________ -->
5853<div class="doc_subsubsection">
5854 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5855</div>
5856
5857<div class="doc_text">
5858
5859<h5>Syntax:</h5>
5860<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5861integer bit width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5862<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5863 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5864 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5865 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5866 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5867 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5868</pre>
5869
5870<h5>Overview:</h5>
5871
5872<p>
5873The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5874trailing zeros.
5875</p>
5876
5877<h5>Arguments:</h5>
5878
5879<p>
5880The only argument is the value to be counted. The argument may be of any
5881integer type. The return type must match the argument type.
5882</p>
5883
5884<h5>Semantics:</h5>
5885
5886<p>
5887The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5888in a variable. If the src == 0 then the result is the size in bits of the type
5889of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5890</p>
5891</div>
5892
5893<!-- _______________________________________________________________________ -->
5894<div class="doc_subsubsection">
5895 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5896</div>
5897
5898<div class="doc_text">
5899
5900<h5>Syntax:</h5>
5901<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5902on any integer bit width.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5903<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5904 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5905 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5906</pre>
5907
5908<h5>Overview:</h5>
5909<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5910range of bits from an integer value and returns them in the same bit width as
5911the original value.</p>
5912
5913<h5>Arguments:</h5>
5914<p>The first argument, <tt>%val</tt> and the result may be integer types of
5915any bit width but they must have the same bit width. The second and third
5916arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5917
5918<h5>Semantics:</h5>
5919<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5920of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5921<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5922operates in forward mode.</p>
5923<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5924right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5925only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5926<ol>
5927 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5928 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5929 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5930 to determine the number of bits to retain.</li>
5931 <li>A mask of the retained bits is created by shifting a -1 value.</li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5932 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5933</ol>
5934<p>In reverse mode, a similar computation is made except that the bits are
5935returned in the reverse order. So, for example, if <tt>X</tt> has the value
5936<tt>i16 0x0ACF (101011001111)</tt> and we apply
5937<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5938<tt>i16 0x0026 (000000100110)</tt>.</p>
5939</div>
5940
5941<div class="doc_subsubsection">
5942 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5943</div>
5944
5945<div class="doc_text">
5946
5947<h5>Syntax:</h5>
5948<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5949on any integer bit width.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5950<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5951 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5952 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]5953</pre>
5954
5955<h5>Overview:</h5>
5956<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5957of bits in an integer value with another integer value. It returns the integer
5958with the replaced bits.</p>
5959
5960<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5961<p>The first argument, <tt>%val</tt>, and the result may be integer types of
5962any bit width, but they must have the same bit width. <tt>%val</tt> is the value
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5963whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5964integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5965type since they specify only a bit index.</p>
5966
5967<h5>Semantics:</h5>
5968<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5969of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5970<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5971operates in forward mode.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5972
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5973<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5974truncating it down to the size of the replacement area or zero extending it
5975up to that size.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5976
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5977<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5978are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5979in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5980to the <tt>%hi</tt>th bit.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5981
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5982<p>In reverse mode, a similar computation is made except that the bits are
5983reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5984<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5985
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5986<h5>Examples:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5987
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5988<pre>
5989 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5990 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5991 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5992 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5993 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5994</pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5995
5996</div>
5997
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]5998<!-- ======================================================================= -->
5999<div class="doc_subsection">
6000 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6001</div>
6002
6003<div class="doc_text">
6004<p>
6005LLVM provides intrinsics for some arithmetic with overflow operations.
6006</p>
6007
6008</div>
6009
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6010<!-- _______________________________________________________________________ -->
6011<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6012 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6013</div>
6014
6015<div class="doc_text">
6016
6017<h5>Syntax:</h5>
6018
6019<p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6020on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6021
6022<pre>
6023 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6024 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6025 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6026</pre>
6027
6028<h5>Overview:</h5>
6029
6030<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6031a signed addition of the two arguments, and indicate whether an overflow
6032occurred during the signed summation.</p>
6033
6034<h5>Arguments:</h5>
6035
6036<p>The arguments (%a and %b) and the first element of the result structure may
6037be of integer types of any bit width, but they must have the same bit width. The
6038second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6039and <tt>%b</tt> are the two values that will undergo signed addition.</p>
6040
6041<h5>Semantics:</h5>
6042
6043<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6044a signed addition of the two variables. They return a structure &mdash; the
6045first element of which is the signed summation, and the second element of which
6046is a bit specifying if the signed summation resulted in an overflow.</p>
6047
6048<h5>Examples:</h5>
6049<pre>
6050 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6051 %sum = extractvalue {i32, i1} %res, 0
6052 %obit = extractvalue {i32, i1} %res, 1
6053 br i1 %obit, label %overflow, label %normal
6054</pre>
6055
6056</div>
6057
6058<!-- _______________________________________________________________________ -->
6059<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6060 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6061</div>
6062
6063<div class="doc_text">
6064
6065<h5>Syntax:</h5>
6066
6067<p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6068on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6069
6070<pre>
6071 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6072 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6073 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6074</pre>
6075
6076<h5>Overview:</h5>
6077
6078<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6079an unsigned addition of the two arguments, and indicate whether a carry occurred
6080during the unsigned summation.</p>
6081
6082<h5>Arguments:</h5>
6083
6084<p>The arguments (%a and %b) and the first element of the result structure may
6085be of integer types of any bit width, but they must have the same bit width. The
6086second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6087and <tt>%b</tt> are the two values that will undergo unsigned addition.</p>
6088
6089<h5>Semantics:</h5>
6090
6091<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6092an unsigned addition of the two arguments. They return a structure &mdash; the
6093first element of which is the sum, and the second element of which is a bit
6094specifying if the unsigned summation resulted in a carry.</p>
6095
6096<h5>Examples:</h5>
6097<pre>
6098 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6099 %sum = extractvalue {i32, i1} %res, 0
6100 %obit = extractvalue {i32, i1} %res, 1
6101 br i1 %obit, label %carry, label %normal
6102</pre>
6103
6104</div>
6105
6106<!-- _______________________________________________________________________ -->
6107<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6108 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6109</div>
6110
6111<div class="doc_text">
6112
6113<h5>Syntax:</h5>
6114
6115<p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6116on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6117
6118<pre>
6119 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6120 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6121 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6122</pre>
6123
6124<h5>Overview:</h5>
6125
6126<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6127a signed subtraction of the two arguments, and indicate whether an overflow
6128occurred during the signed subtraction.</p>
6129
6130<h5>Arguments:</h5>
6131
6132<p>The arguments (%a and %b) and the first element of the result structure may
6133be of integer types of any bit width, but they must have the same bit width. The
6134second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6135and <tt>%b</tt> are the two values that will undergo signed subtraction.</p>
6136
6137<h5>Semantics:</h5>
6138
6139<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6140a signed subtraction of the two arguments. They return a structure &mdash; the
6141first element of which is the subtraction, and the second element of which is a bit
6142specifying if the signed subtraction resulted in an overflow.</p>
6143
6144<h5>Examples:</h5>
6145<pre>
6146 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6147 %sum = extractvalue {i32, i1} %res, 0
6148 %obit = extractvalue {i32, i1} %res, 1
6149 br i1 %obit, label %overflow, label %normal
6150</pre>
6151
6152</div>
6153
6154<!-- _______________________________________________________________________ -->
6155<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6156 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6157</div>
6158
6159<div class="doc_text">
6160
6161<h5>Syntax:</h5>
6162
6163<p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6164on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6165
6166<pre>
6167 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6168 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6169 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6170</pre>
6171
6172<h5>Overview:</h5>
6173
6174<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6175an unsigned subtraction of the two arguments, and indicate whether an overflow
6176occurred during the unsigned subtraction.</p>
6177
6178<h5>Arguments:</h5>
6179
6180<p>The arguments (%a and %b) and the first element of the result structure may
6181be of integer types of any bit width, but they must have the same bit width. The
6182second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6183and <tt>%b</tt> are the two values that will undergo unsigned subtraction.</p>
6184
6185<h5>Semantics:</h5>
6186
6187<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6188an unsigned subtraction of the two arguments. They return a structure &mdash; the
6189first element of which is the subtraction, and the second element of which is a bit
6190specifying if the unsigned subtraction resulted in an overflow.</p>
6191
6192<h5>Examples:</h5>
6193<pre>
6194 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6195 %sum = extractvalue {i32, i1} %res, 0
6196 %obit = extractvalue {i32, i1} %res, 1
6197 br i1 %obit, label %overflow, label %normal
6198</pre>
6199
6200</div>
6201
6202<!-- _______________________________________________________________________ -->
6203<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6204 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6205</div>
6206
6207<div class="doc_text">
6208
6209<h5>Syntax:</h5>
6210
6211<p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6212on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6213
6214<pre>
6215 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6216 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6217 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6218</pre>
6219
6220<h5>Overview:</h5>
6221
6222<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6223a signed multiplication of the two arguments, and indicate whether an overflow
6224occurred during the signed multiplication.</p>
6225
6226<h5>Arguments:</h5>
6227
6228<p>The arguments (%a and %b) and the first element of the result structure may
6229be of integer types of any bit width, but they must have the same bit width. The
6230second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6231and <tt>%b</tt> are the two values that will undergo signed multiplication.</p>
6232
6233<h5>Semantics:</h5>
6234
6235<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6236a signed multiplication of the two arguments. They return a structure &mdash;
6237the first element of which is the multiplication, and the second element of
6238which is a bit specifying if the signed multiplication resulted in an
6239overflow.</p>
6240
6241<h5>Examples:</h5>
6242<pre>
6243 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6244 %sum = extractvalue {i32, i1} %res, 0
6245 %obit = extractvalue {i32, i1} %res, 1
6246 br i1 %obit, label %overflow, label %normal
6247</pre>
6248
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6249</div>
6250
Bill Wendlingbda98b62009-02-08 23:00:09 +0000[diff] [blame]6251<!-- _______________________________________________________________________ -->
6252<div class="doc_subsubsection">
6253 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6254</div>
6255
6256<div class="doc_text">
6257
6258<h5>Syntax:</h5>
6259
6260<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6261on any integer bit width.</p>
6262
6263<pre>
6264 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6265 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6266 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6267</pre>
6268
6269<h5>Overview:</h5>
6270
6271<p><i><b>Warning:</b> '<tt>llvm.umul.with.overflow</tt>' is badly broken. It is
6272actively being fixed, but it should not currently be used!</i></p>
6273
6274<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6275a unsigned multiplication of the two arguments, and indicate whether an overflow
6276occurred during the unsigned multiplication.</p>
6277
6278<h5>Arguments:</h5>
6279
6280<p>The arguments (%a and %b) and the first element of the result structure may
6281be of integer types of any bit width, but they must have the same bit width. The
6282second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6283and <tt>%b</tt> are the two values that will undergo unsigned
6284multiplication.</p>
6285
6286<h5>Semantics:</h5>
6287
6288<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6289an unsigned multiplication of the two arguments. They return a structure &mdash;
6290the first element of which is the multiplication, and the second element of
6291which is a bit specifying if the unsigned multiplication resulted in an
6292overflow.</p>
6293
6294<h5>Examples:</h5>
6295<pre>
6296 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6297 %sum = extractvalue {i32, i1} %res, 0
6298 %obit = extractvalue {i32, i1} %res, 1
6299 br i1 %obit, label %overflow, label %normal
6300</pre>
6301
6302</div>
6303
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6304<!-- ======================================================================= -->
6305<div class="doc_subsection">
6306 <a name="int_debugger">Debugger Intrinsics</a>
6307</div>
6308
6309<div class="doc_text">
6310<p>
6311The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
6312are described in the <a
6313href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
6314Debugging</a> document.
6315</p>
6316</div>
6317
6318
6319<!-- ======================================================================= -->
6320<div class="doc_subsection">
6321 <a name="int_eh">Exception Handling Intrinsics</a>
6322</div>
6323
6324<div class="doc_text">
6325<p> The LLVM exception handling intrinsics (which all start with
6326<tt>llvm.eh.</tt> prefix), are described in the <a
6327href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6328Handling</a> document. </p>
6329</div>
6330
6331<!-- ======================================================================= -->
6332<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6333 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6334</div>
6335
6336<div class="doc_text">
6337<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6338 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6339 the <tt>nest</tt> attribute, from a function. The result is a callable
6340 function pointer lacking the nest parameter - the caller does not need
6341 to provide a value for it. Instead, the value to use is stored in
6342 advance in a "trampoline", a block of memory usually allocated
6343 on the stack, which also contains code to splice the nest value into the
6344 argument list. This is used to implement the GCC nested function address
6345 extension.
6346</p>
6347<p>
6348 For example, if the function is
6349 <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]6350 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]6351<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6352 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6353 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6354 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6355 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6356</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]6357 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6358 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6359</div>
6360
6361<!-- _______________________________________________________________________ -->
6362<div class="doc_subsubsection">
6363 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6364</div>
6365<div class="doc_text">
6366<h5>Syntax:</h5>
6367<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6368declare 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]6369</pre>
6370<h5>Overview:</h5>
6371<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6372 This fills the memory pointed to by <tt>tramp</tt> with code
6373 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6374</p>
6375<h5>Arguments:</h5>
6376<p>
6377 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6378 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
6379 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]6380 intrinsic. Note that the size and the alignment are target-specific - LLVM
6381 currently provides no portable way of determining them, so a front-end that
6382 generates this intrinsic needs to have some target-specific knowledge.
6383 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]6384</p>
6385<h5>Semantics:</h5>
6386<p>
6387 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6388 dependent code, turning it into a function. A pointer to this function is
6389 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6390 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6391 before being called. The new function's signature is the same as that of
6392 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
6393 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
6394 of pointer type. Calling the new function is equivalent to calling
6395 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
6396 missing <tt>nest</tt> argument. If, after calling
6397 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
6398 modified, then the effect of any later call to the returned function pointer is
6399 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6400</p>
6401</div>
6402
6403<!-- ======================================================================= -->
6404<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6405 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6406</div>
6407
6408<div class="doc_text">
6409<p>
6410 These intrinsic functions expand the "universal IR" of LLVM to represent
6411 hardware constructs for atomic operations and memory synchronization. This
6412 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]6413 is aimed at a low enough level to allow any programming models or APIs
6414 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6415 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6416 hardware behavior. Just as hardware provides a "universal IR" for source
6417 languages, it also provides a starting point for developing a "universal"
6418 atomic operation and synchronization IR.
6419</p>
6420<p>
6421 These do <em>not</em> form an API such as high-level threading libraries,
6422 software transaction memory systems, atomic primitives, and intrinsic
6423 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6424 application libraries. The hardware interface provided by LLVM should allow
6425 a clean implementation of all of these APIs and parallel programming models.
6426 No one model or paradigm should be selected above others unless the hardware
6427 itself ubiquitously does so.
6428
6429</p>
6430</div>
6431
6432<!-- _______________________________________________________________________ -->
6433<div class="doc_subsubsection">
6434 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6435</div>
6436<div class="doc_text">
6437<h5>Syntax:</h5>
6438<pre>
6439declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
6440i1 &lt;device&gt; )
6441
6442</pre>
6443<h5>Overview:</h5>
6444<p>
6445 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6446 specific pairs of memory access types.
6447</p>
6448<h5>Arguments:</h5>
6449<p>
6450 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6451 The first four arguments enables a specific barrier as listed below. The fith
6452 argument specifies that the barrier applies to io or device or uncached memory.
6453
6454</p>
6455 <ul>
6456 <li><tt>ll</tt>: load-load barrier</li>
6457 <li><tt>ls</tt>: load-store barrier</li>
6458 <li><tt>sl</tt>: store-load barrier</li>
6459 <li><tt>ss</tt>: store-store barrier</li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6460 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6461 </ul>
6462<h5>Semantics:</h5>
6463<p>
6464 This intrinsic causes the system to enforce some ordering constraints upon
6465 the loads and stores of the program. This barrier does not indicate
6466 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6467 which they occur. For any of the specified pairs of load and store operations
6468 (f.ex. load-load, or store-load), all of the first operations preceding the
6469 barrier will complete before any of the second operations succeeding the
6470 barrier begin. Specifically the semantics for each pairing is as follows:
6471</p>
6472 <ul>
6473 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6474 after the barrier begins.</li>
6475
6476 <li><tt>ls</tt>: All loads before the barrier must complete before any
6477 store after the barrier begins.</li>
6478 <li><tt>ss</tt>: All stores before the barrier must complete before any
6479 store after the barrier begins.</li>
6480 <li><tt>sl</tt>: All stores before the barrier must complete before any
6481 load after the barrier begins.</li>
6482 </ul>
6483<p>
6484 These semantics are applied with a logical "and" behavior when more than one
6485 is enabled in a single memory barrier intrinsic.
6486</p>
6487<p>
6488 Backends may implement stronger barriers than those requested when they do not
6489 support as fine grained a barrier as requested. Some architectures do not
6490 need all types of barriers and on such architectures, these become noops.
6491</p>
6492<h5>Example:</h5>
6493<pre>
6494%ptr = malloc i32
6495 store i32 4, %ptr
6496
6497%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6498 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6499 <i>; guarantee the above finishes</i>
6500 store i32 8, %ptr <i>; before this begins</i>
6501</pre>
6502</div>
6503
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6504<!-- _______________________________________________________________________ -->
6505<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6506 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6507</div>
6508<div class="doc_text">
6509<h5>Syntax:</h5>
6510<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6511 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6512 any integer bit width and for different address spaces. Not all targets
6513 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6514
6515<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6516declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
6517declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
6518declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
6519declare 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]6520
6521</pre>
6522<h5>Overview:</h5>
6523<p>
6524 This loads a value in memory and compares it to a given value. If they are
6525 equal, it stores a new value into the memory.
6526</p>
6527<h5>Arguments:</h5>
6528<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6529 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6530 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6531 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6532 this integer type. While any bit width integer may be used, targets may only
6533 lower representations they support in hardware.
6534
6535</p>
6536<h5>Semantics:</h5>
6537<p>
6538 This entire intrinsic must be executed atomically. It first loads the value
6539 in memory pointed to by <tt>ptr</tt> and compares it with the value
6540 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6541 loaded value is yielded in all cases. This provides the equivalent of an
6542 atomic compare-and-swap operation within the SSA framework.
6543</p>
6544<h5>Examples:</h5>
6545
6546<pre>
6547%ptr = malloc i32
6548 store i32 4, %ptr
6549
6550%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6551%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6552 <i>; yields {i32}:result1 = 4</i>
6553%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6554%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6555
6556%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6557%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6558 <i>; yields {i32}:result2 = 8</i>
6559%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6560
6561%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6562</pre>
6563</div>
6564
6565<!-- _______________________________________________________________________ -->
6566<div class="doc_subsubsection">
6567 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6568</div>
6569<div class="doc_text">
6570<h5>Syntax:</h5>
6571
6572<p>
6573 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6574 integer bit width. Not all targets support all bit widths however.</p>
6575<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6576declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
6577declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
6578declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
6579declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6580
6581</pre>
6582<h5>Overview:</h5>
6583<p>
6584 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6585 the value from memory. It then stores the value in <tt>val</tt> in the memory
6586 at <tt>ptr</tt>.
6587</p>
6588<h5>Arguments:</h5>
6589
6590<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6591 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6592 <tt>val</tt> argument and the result must be integers of the same bit width.
6593 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6594 integer type. The targets may only lower integer representations they
6595 support.
6596</p>
6597<h5>Semantics:</h5>
6598<p>
6599 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6600 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6601 equivalent of an atomic swap operation within the SSA framework.
6602
6603</p>
6604<h5>Examples:</h5>
6605<pre>
6606%ptr = malloc i32
6607 store i32 4, %ptr
6608
6609%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6610%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6611 <i>; yields {i32}:result1 = 4</i>
6612%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6613%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6614
6615%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6616%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6617 <i>; yields {i32}:result2 = 8</i>
6618
6619%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6620%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6621</pre>
6622</div>
6623
6624<!-- _______________________________________________________________________ -->
6625<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6626 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6627
6628</div>
6629<div class="doc_text">
6630<h5>Syntax:</h5>
6631<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6632 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]6633 integer bit width. Not all targets support all bit widths however.</p>
6634<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6635declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6636declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6637declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6638declare 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]6639
6640</pre>
6641<h5>Overview:</h5>
6642<p>
6643 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6644 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6645</p>
6646<h5>Arguments:</h5>
6647<p>
6648
6649 The intrinsic takes two arguments, the first a pointer to an integer value
6650 and the second an integer value. The result is also an integer value. These
6651 integer types can have any bit width, but they must all have the same bit
6652 width. The targets may only lower integer representations they support.
6653</p>
6654<h5>Semantics:</h5>
6655<p>
6656 This intrinsic does a series of operations atomically. It first loads the
6657 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6658 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6659</p>
6660
6661<h5>Examples:</h5>
6662<pre>
6663%ptr = malloc i32
6664 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6665%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6666 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6667%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6668 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6669%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6670 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6671%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6672</pre>
6673</div>
6674
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6675<!-- _______________________________________________________________________ -->
6676<div class="doc_subsubsection">
6677 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6678
6679</div>
6680<div class="doc_text">
6681<h5>Syntax:</h5>
6682<p>
6683 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]6684 any integer bit width and for different address spaces. Not all targets
6685 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6686<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6687declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6688declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6689declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6690declare 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]6691
6692</pre>
6693<h5>Overview:</h5>
6694<p>
6695 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6696 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6697</p>
6698<h5>Arguments:</h5>
6699<p>
6700
6701 The intrinsic takes two arguments, the first a pointer to an integer value
6702 and the second an integer value. The result is also an integer value. These
6703 integer types can have any bit width, but they must all have the same bit
6704 width. The targets may only lower integer representations they support.
6705</p>
6706<h5>Semantics:</h5>
6707<p>
6708 This intrinsic does a series of operations atomically. It first loads the
6709 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6710 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6711</p>
6712
6713<h5>Examples:</h5>
6714<pre>
6715%ptr = malloc i32
6716 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6717%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6718 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6719%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6720 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6721%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6722 <i>; yields {i32}:result3 = 2</i>
6723%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6724</pre>
6725</div>
6726
6727<!-- _______________________________________________________________________ -->
6728<div class="doc_subsubsection">
6729 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6730 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6731 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6732 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6733
6734</div>
6735<div class="doc_text">
6736<h5>Syntax:</h5>
6737<p>
6738 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6739 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6740 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6741 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6742<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6743declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6744declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6745declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6746declare 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]6747
6748</pre>
6749
6750<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6751declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6752declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6753declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6754declare 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]6755
6756</pre>
6757
6758<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6759declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6760declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6761declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6762declare 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]6763
6764</pre>
6765
6766<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6767declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6768declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6769declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6770declare 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]6771
6772</pre>
6773<h5>Overview:</h5>
6774<p>
6775 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6776 the value stored in memory at <tt>ptr</tt>. It yields the original value
6777 at <tt>ptr</tt>.
6778</p>
6779<h5>Arguments:</h5>
6780<p>
6781
6782 These intrinsics take two arguments, the first a pointer to an integer value
6783 and the second an integer value. The result is also an integer value. These
6784 integer types can have any bit width, but they must all have the same bit
6785 width. The targets may only lower integer representations they support.
6786</p>
6787<h5>Semantics:</h5>
6788<p>
6789 These intrinsics does a series of operations atomically. They first load the
6790 value stored at <tt>ptr</tt>. They then do the bitwise operation
6791 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6792 value stored at <tt>ptr</tt>.
6793</p>
6794
6795<h5>Examples:</h5>
6796<pre>
6797%ptr = malloc i32
6798 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6799%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6800 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6801%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6802 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6803%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6804 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6805%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6806 <i>; yields {i32}:result3 = FF</i>
6807%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6808</pre>
6809</div>
6810
6811
6812<!-- _______________________________________________________________________ -->
6813<div class="doc_subsubsection">
6814 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6815 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6816 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6817 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6818
6819</div>
6820<div class="doc_text">
6821<h5>Syntax:</h5>
6822<p>
6823 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6824 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6825 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6826 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6827 support all bit widths however.</p>
6828<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6829declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6830declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6831declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6832declare 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]6833
6834</pre>
6835
6836<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6837declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6838declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6839declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6840declare 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]6841
6842</pre>
6843
6844<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6845declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6846declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6847declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6848declare 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]6849
6850</pre>
6851
6852<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6853declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6854declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6855declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6856declare 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]6857
6858</pre>
6859<h5>Overview:</h5>
6860<p>
6861 These intrinsics takes the signed or unsigned minimum or maximum of
6862 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6863 original value at <tt>ptr</tt>.
6864</p>
6865<h5>Arguments:</h5>
6866<p>
6867
6868 These intrinsics take two arguments, the first a pointer to an integer value
6869 and the second an integer value. The result is also an integer value. These
6870 integer types can have any bit width, but they must all have the same bit
6871 width. The targets may only lower integer representations they support.
6872</p>
6873<h5>Semantics:</h5>
6874<p>
6875 These intrinsics does a series of operations atomically. They first load the
6876 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6877 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6878 the original value stored at <tt>ptr</tt>.
6879</p>
6880
6881<h5>Examples:</h5>
6882<pre>
6883%ptr = malloc i32
6884 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6885%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6886 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6887%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6888 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6889%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6890 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6891%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6892 <i>; yields {i32}:result3 = 8</i>
6893%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6894</pre>
6895</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6896
6897<!-- ======================================================================= -->
6898<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6899 <a name="int_general">General Intrinsics</a>
6900</div>
6901
6902<div class="doc_text">
6903<p> This class of intrinsics is designed to be generic and has
6904no specific purpose. </p>
6905</div>
6906
6907<!-- _______________________________________________________________________ -->
6908<div class="doc_subsubsection">
6909 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6910</div>
6911
6912<div class="doc_text">
6913
6914<h5>Syntax:</h5>
6915<pre>
6916 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6917</pre>
6918
6919<h5>Overview:</h5>
6920
6921<p>
6922The '<tt>llvm.var.annotation</tt>' intrinsic
6923</p>
6924
6925<h5>Arguments:</h5>
6926
6927<p>
6928The first argument is a pointer to a value, the second is a pointer to a
6929global string, the third is a pointer to a global string which is the source
6930file name, and the last argument is the line number.
6931</p>
6932
6933<h5>Semantics:</h5>
6934
6935<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6936This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6937This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6938annotations. These have no other defined use, they are ignored by code
6939generation and optimization.
6940</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6941</div>
6942
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6943<!-- _______________________________________________________________________ -->
6944<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6945 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6946</div>
6947
6948<div class="doc_text">
6949
6950<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6951<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6952any integer bit width.
6953</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6954<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6955 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6956 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6957 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6958 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6959 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]6960</pre>
6961
6962<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6963
6964<p>
6965The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6966</p>
6967
6968<h5>Arguments:</h5>
6969
6970<p>
6971The first argument is an integer value (result of some expression),
6972the second is a pointer to a global string, the third is a pointer to a global
6973string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6974It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6975</p>
6976
6977<h5>Semantics:</h5>
6978
6979<p>
6980This intrinsic allows annotations to be put on arbitrary expressions
6981with arbitrary strings. This can be useful for special purpose optimizations
6982that want to look for these annotations. These have no other defined use, they
6983are ignored by code generation and optimization.
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6984</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6985</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6986
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6987<!-- _______________________________________________________________________ -->
6988<div class="doc_subsubsection">
6989 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6990</div>
6991
6992<div class="doc_text">
6993
6994<h5>Syntax:</h5>
6995<pre>
6996 declare void @llvm.trap()
6997</pre>
6998
6999<h5>Overview:</h5>
7000
7001<p>
7002The '<tt>llvm.trap</tt>' intrinsic
7003</p>
7004
7005<h5>Arguments:</h5>
7006
7007<p>
7008None
7009</p>
7010
7011<h5>Semantics:</h5>
7012
7013<p>
7014This intrinsics is lowered to the target dependent trap instruction. If the
7015target does not have a trap instruction, this intrinsic will be lowered to the
7016call of the abort() function.
7017</p>
7018</div>
7019
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]7020<!-- _______________________________________________________________________ -->
7021<div class="doc_subsubsection">
Misha Brukman5dd7f4d2008-11-22 23:55:29 +0000[diff] [blame]7022 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]7023</div>
7024<div class="doc_text">
7025<h5>Syntax:</h5>
7026<pre>
7027declare void @llvm.stackprotector( i8* &lt;guard&gt;, i8** &lt;slot&gt; )
7028
7029</pre>
7030<h5>Overview:</h5>
7031<p>
7032 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
7033 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
7034 it is placed on the stack before local variables.
7035</p>
7036<h5>Arguments:</h5>
7037<p>
7038 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
7039 first argument is the value loaded from the stack guard
7040 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
7041 has enough space to hold the value of the guard.
7042</p>
7043<h5>Semantics:</h5>
7044<p>
7045 This intrinsic causes the prologue/epilogue inserter to force the position of
7046 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7047 stack. This is to ensure that if a local variable on the stack is overwritten,
7048 it will destroy the value of the guard. When the function exits, the guard on
7049 the stack is checked against the original guard. If they're different, then
7050 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
7051</p>
7052</div>
7053
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7054<!-- *********************************************************************** -->
7055<hr>
7056<address>
7057 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]7058 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7059 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]7060 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7061
7062 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7063 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7064 Last modified: $Date$
7065</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]7066
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7067</body>
7068</html>