Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 50609942838306d7571b6a31aca82c26d9ae3b53 / . / docs / LangRef.html
blob: 6b7038dbccbdcc17d7d87dedef9f354d05d81fa3 [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>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]68 </ol>
69 </li>
70 <li><a href="#othervalues">Other Values</a>
71 <ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]72 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]73 </ol>
74 </li>
75 <li><a href="#instref">Instruction Reference</a>
76 <ol>
77 <li><a href="#terminators">Terminator Instructions</a>
78 <ol>
79 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
80 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
81 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
82 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
83 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
84 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
85 </ol>
86 </li>
87 <li><a href="#binaryops">Binary Operations</a>
88 <ol>
89 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
90 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
91 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
92 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
93 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
94 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
95 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
96 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
97 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
98 </ol>
99 </li>
100 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
101 <ol>
102 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
103 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
104 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
105 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
106 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
107 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
108 </ol>
109 </li>
110 <li><a href="#vectorops">Vector Operations</a>
111 <ol>
112 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
113 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
114 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
115 </ol>
116 </li>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]117 <li><a href="#aggregateops">Aggregate Operations</a>
118 <ol>
119 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
120 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
121 </ol>
122 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]123 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
124 <ol>
125 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
126 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
127 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
128 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
129 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
130 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
131 </ol>
132 </li>
133 <li><a href="#convertops">Conversion Operations</a>
134 <ol>
135 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
136 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
137 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
140 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
141 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
142 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
143 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
144 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
145 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
146 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
147 </ol>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]148 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]149 <li><a href="#otherops">Other Operations</a>
150 <ol>
151 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
152 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]153 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
154 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]155 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
156 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
157 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
158 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
159 </ol>
160 </li>
161 </ol>
162 </li>
163 <li><a href="#intrinsics">Intrinsic Functions</a>
164 <ol>
165 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
166 <ol>
167 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
168 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
169 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
170 </ol>
171 </li>
172 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
173 <ol>
174 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
175 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
176 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
177 </ol>
178 </li>
179 <li><a href="#int_codegen">Code Generator Intrinsics</a>
180 <ol>
181 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
182 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
183 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
184 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
185 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
186 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
187 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
188 </ol>
189 </li>
190 <li><a href="#int_libc">Standard C Library Intrinsics</a>
191 <ol>
192 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
194 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]197 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
198 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
199 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]200 </ol>
201 </li>
202 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
203 <ol>
204 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
205 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
207 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
208 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
209 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
210 </ol>
211 </li>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]212 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
213 <ol>
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]214 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
215 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
216 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
217 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
218 <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]219 <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]220 </ol>
221 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]222 <li><a href="#int_debugger">Debugger intrinsics</a></li>
223 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]224 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]225 <ol>
226 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]227 </ol>
228 </li>
Bill Wendling9127adb2008-11-18 22:10:53 +0000[diff] [blame]229 <li><a href="#int_atomics">Atomic intrinsics</a>
230 <ol>
231 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
232 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
233 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
234 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
235 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
236 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
237 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
238 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
239 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
240 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
241 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
242 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
243 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
244 </ol>
245 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]246 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]247 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]248 <li><a href="#int_var_annotation">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]249 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]250 <li><a href="#int_annotation">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]251 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]252 <li><a href="#int_trap">
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]253 '<tt>llvm.trap</tt>' Intrinsic</a></li>
254 <li><a href="#int_stackprotector">
255 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]256 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]257 </li>
258 </ol>
259 </li>
260</ol>
261
262<div class="doc_author">
263 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
264 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
265</div>
266
267<!-- *********************************************************************** -->
268<div class="doc_section"> <a name="abstract">Abstract </a></div>
269<!-- *********************************************************************** -->
270
271<div class="doc_text">
272<p>This document is a reference manual for the LLVM assembly language.
Bill Wendlinge7846a52008-08-05 22:29:16 +0000[diff] [blame]273LLVM is a Static Single Assignment (SSA) based representation that provides
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]274type safety, low-level operations, flexibility, and the capability of
275representing 'all' high-level languages cleanly. It is the common code
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]276representation used throughout all phases of the LLVM compilation
277strategy.</p>
278</div>
279
280<!-- *********************************************************************** -->
281<div class="doc_section"> <a name="introduction">Introduction</a> </div>
282<!-- *********************************************************************** -->
283
284<div class="doc_text">
285
286<p>The LLVM code representation is designed to be used in three
287different forms: as an in-memory compiler IR, as an on-disk bitcode
288representation (suitable for fast loading by a Just-In-Time compiler),
289and as a human readable assembly language representation. This allows
290LLVM to provide a powerful intermediate representation for efficient
291compiler transformations and analysis, while providing a natural means
292to debug and visualize the transformations. The three different forms
293of LLVM are all equivalent. This document describes the human readable
294representation and notation.</p>
295
296<p>The LLVM representation aims to be light-weight and low-level
297while being expressive, typed, and extensible at the same time. It
298aims to be a "universal IR" of sorts, by being at a low enough level
299that high-level ideas may be cleanly mapped to it (similar to how
300microprocessors are "universal IR's", allowing many source languages to
301be mapped to them). By providing type information, LLVM can be used as
302the target of optimizations: for example, through pointer analysis, it
303can be proven that a C automatic variable is never accessed outside of
304the current function... allowing it to be promoted to a simple SSA
305value instead of a memory location.</p>
306
307</div>
308
309<!-- _______________________________________________________________________ -->
310<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
311
312<div class="doc_text">
313
314<p>It is important to note that this document describes 'well formed'
315LLVM assembly language. There is a difference between what the parser
316accepts and what is considered 'well formed'. For example, the
317following instruction is syntactically okay, but not well formed:</p>
318
319<div class="doc_code">
320<pre>
321%x = <a href="#i_add">add</a> i32 1, %x
322</pre>
323</div>
324
325<p>...because the definition of <tt>%x</tt> does not dominate all of
326its uses. The LLVM infrastructure provides a verification pass that may
327be used to verify that an LLVM module is well formed. This pass is
328automatically run by the parser after parsing input assembly and by
329the optimizer before it outputs bitcode. The violations pointed out
330by the verifier pass indicate bugs in transformation passes or input to
331the parser.</p>
332</div>
333
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]334<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]335
336<!-- *********************************************************************** -->
337<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
338<!-- *********************************************************************** -->
339
340<div class="doc_text">
341
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]342 <p>LLVM identifiers come in two basic types: global and local. Global
343 identifiers (functions, global variables) begin with the @ character. Local
344 identifiers (register names, types) begin with the % character. Additionally,
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]345 there are three different formats for identifiers, for different purposes:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]346
347<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]348 <li>Named values are represented as a string of characters with their prefix.
349 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
350 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]351 Identifiers which require other characters in their names can be surrounded
Daniel Dunbar73f9a772008-10-14 23:51:43 +0000[diff] [blame]352 with quotes. Special characters may be escaped using "\xx" where xx is the
353 ASCII code for the character in hexadecimal. In this way, any character can
354 be used in a name value, even quotes themselves.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]355
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]356 <li>Unnamed values are represented as an unsigned numeric value with their
357 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]358
359 <li>Constants, which are described in a <a href="#constants">section about
360 constants</a>, below.</li>
361</ol>
362
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]363<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]364don't need to worry about name clashes with reserved words, and the set of
365reserved words may be expanded in the future without penalty. Additionally,
366unnamed identifiers allow a compiler to quickly come up with a temporary
367variable without having to avoid symbol table conflicts.</p>
368
369<p>Reserved words in LLVM are very similar to reserved words in other
370languages. There are keywords for different opcodes
371('<tt><a href="#i_add">add</a></tt>',
372 '<tt><a href="#i_bitcast">bitcast</a></tt>',
373 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
374href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
375and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]376none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]377
378<p>Here is an example of LLVM code to multiply the integer variable
379'<tt>%X</tt>' by 8:</p>
380
381<p>The easy way:</p>
382
383<div class="doc_code">
384<pre>
385%result = <a href="#i_mul">mul</a> i32 %X, 8
386</pre>
387</div>
388
389<p>After strength reduction:</p>
390
391<div class="doc_code">
392<pre>
393%result = <a href="#i_shl">shl</a> i32 %X, i8 3
394</pre>
395</div>
396
397<p>And the hard way:</p>
398
399<div class="doc_code">
400<pre>
401<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
402<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
403%result = <a href="#i_add">add</a> i32 %1, %1
404</pre>
405</div>
406
407<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
408important lexical features of LLVM:</p>
409
410<ol>
411
412 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
413 line.</li>
414
415 <li>Unnamed temporaries are created when the result of a computation is not
416 assigned to a named value.</li>
417
418 <li>Unnamed temporaries are numbered sequentially</li>
419
420</ol>
421
422<p>...and it also shows a convention that we follow in this document. When
423demonstrating instructions, we will follow an instruction with a comment that
424defines the type and name of value produced. Comments are shown in italic
425text.</p>
426
427</div>
428
429<!-- *********************************************************************** -->
430<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
431<!-- *********************************************************************** -->
432
433<!-- ======================================================================= -->
434<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
435</div>
436
437<div class="doc_text">
438
439<p>LLVM programs are composed of "Module"s, each of which is a
440translation unit of the input programs. Each module consists of
441functions, global variables, and symbol table entries. Modules may be
442combined together with the LLVM linker, which merges function (and
443global variable) definitions, resolves forward declarations, and merges
444symbol table entries. Here is an example of the "hello world" module:</p>
445
446<div class="doc_code">
447<pre><i>; Declare the string constant as a global constant...</i>
448<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
449 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
450
451<i>; External declaration of the puts function</i>
452<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
453
454<i>; Definition of main function</i>
455define i32 @main() { <i>; i32()* </i>
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]456 <i>; Convert [13 x i8]* to i8 *...</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]457 %cast210 = <a
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]458 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]459
460 <i>; Call puts function to write out the string to stdout...</i>
461 <a
462 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
463 <a
464 href="#i_ret">ret</a> i32 0<br>}<br>
465</pre>
466</div>
467
468<p>This example is made up of a <a href="#globalvars">global variable</a>
469named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
470function, and a <a href="#functionstructure">function definition</a>
471for "<tt>main</tt>".</p>
472
473<p>In general, a module is made up of a list of global values,
474where both functions and global variables are global values. Global values are
475represented by a pointer to a memory location (in this case, a pointer to an
476array of char, and a pointer to a function), and have one of the following <a
477href="#linkage">linkage types</a>.</p>
478
479</div>
480
481<!-- ======================================================================= -->
482<div class="doc_subsection">
483 <a name="linkage">Linkage Types</a>
484</div>
485
486<div class="doc_text">
487
488<p>
489All Global Variables and Functions have one of the following types of linkage:
490</p>
491
492<dl>
493
Rafael Espindolaa168fc92009-01-15 20:18:42 +0000[diff] [blame]494 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
495
496 <dd>Global values with private linkage are only directly accessible by
497 objects in the current module. In particular, linking code into a module with
498 an private global value may cause the private to be renamed as necessary to
499 avoid collisions. Because the symbol is private to the module, all
500 references can be updated. This doesn't show up in any symbol table in the
501 object file.
502 </dd>
503
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]504 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]505
Duncan Sandsa75223a2009-01-16 09:29:46 +0000[diff] [blame]506 <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]507 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]508 '<tt>static</tt>' keyword in C.
509 </dd>
510
511 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
512
513 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
514 the same name when linkage occurs. This is typically used to implement
515 inline functions, templates, or other code which must be generated in each
516 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
517 allowed to be discarded.
518 </dd>
519
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]520 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
521
522 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
523 linkage, except that unreferenced <tt>common</tt> globals may not be
524 discarded. This is used for globals that may be emitted in multiple
525 translation units, but that are not guaranteed to be emitted into every
526 translation unit that uses them. One example of this is tentative
527 definitions in C, such as "<tt>int X;</tt>" at global scope.
528 </dd>
529
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]530 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
531
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]532 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
533 that some targets may choose to emit different assembly sequences for them
534 for target-dependent reasons. This is used for globals that are declared
535 "weak" in C source code.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]536 </dd>
537
538 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
539
540 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
541 pointer to array type. When two global variables with appending linkage are
542 linked together, the two global arrays are appended together. This is the
543 LLVM, typesafe, equivalent of having the system linker append together
544 "sections" with identical names when .o files are linked.
545 </dd>
546
547 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]548
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]549 <dd>The semantics of this linkage follow the ELF object file model: the
550 symbol is weak until linked, if not linked, the symbol becomes null instead
551 of being an undefined reference.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]552 </dd>
553
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]554 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
555 <dt><tt><b><a name="linkage_common">common_odr</a></b></tt>: </dt>
556 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
557 <dt><tt><b><a name="linkage_externweak">extern_weak_odr</a></b></tt>: </dt>
558 <dd>Some languages allow inequivalent globals to be merged, such as two
559 functions with different semantics. Other languages, such as <tt>C++</tt>,
560 ensure that only equivalent globals are ever merged (the "one definition
561 rule" - <tt>odr</tt>). Such languages can use the <tt>linkonce_odr</tt>,
562 <tt>common_odr</tt>, <tt>weak_odr</tt> and <tt>extern_weak_odr</tt> linkage
563 types to indicate that the global will only be merged with equivalent
564 globals. These linkage types are otherwise the same as their
565 non-<tt>odr</tt> versions.
566 </dd>
567
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]568 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
569
570 <dd>If none of the above identifiers are used, the global is externally
571 visible, meaning that it participates in linkage and can be used to resolve
572 external symbol references.
573 </dd>
574</dl>
575
576 <p>
577 The next two types of linkage are targeted for Microsoft Windows platform
578 only. They are designed to support importing (exporting) symbols from (to)
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]579 DLLs (Dynamic Link Libraries).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]580 </p>
581
582 <dl>
583 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
584
585 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
586 or variable via a global pointer to a pointer that is set up by the DLL
587 exporting the symbol. On Microsoft Windows targets, the pointer name is
Dan Gohman5ec99832009-01-12 21:35:55 +0000[diff] [blame]588 formed by combining <code>__imp_</code> and the function or variable name.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]589 </dd>
590
591 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
592
593 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
594 pointer to a pointer in a DLL, so that it can be referenced with the
595 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
Dan Gohman5ec99832009-01-12 21:35:55 +0000[diff] [blame]596 name is formed by combining <code>__imp_</code> and the function or variable
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]597 name.
598 </dd>
599
600</dl>
601
Dan Gohman4dfac702008-11-24 17:18:39 +0000[diff] [blame]602<p>For example, since the "<tt>.LC0</tt>"
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]603variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
604variable and was linked with this one, one of the two would be renamed,
605preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
606external (i.e., lacking any linkage declarations), they are accessible
607outside of the current module.</p>
608<p>It is illegal for a function <i>declaration</i>
609to have any linkage type other than "externally visible", <tt>dllimport</tt>,
Duncan Sands19d161f2009-03-07 15:45:40 +0000[diff] [blame]610<tt>extern_weak</tt> or <tt>extern_weak_odr</tt>.</p>
611<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
612or <tt>weak_odr</tt> linkages.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]613</div>
614
615<!-- ======================================================================= -->
616<div class="doc_subsection">
617 <a name="callingconv">Calling Conventions</a>
618</div>
619
620<div class="doc_text">
621
622<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
623and <a href="#i_invoke">invokes</a> can all have an optional calling convention
624specified for the call. The calling convention of any pair of dynamic
625caller/callee must match, or the behavior of the program is undefined. The
626following calling conventions are supported by LLVM, and more may be added in
627the future:</p>
628
629<dl>
630 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
631
632 <dd>This calling convention (the default if no other calling convention is
633 specified) matches the target C calling conventions. This calling convention
634 supports varargs function calls and tolerates some mismatch in the declared
635 prototype and implemented declaration of the function (as does normal C).
636 </dd>
637
638 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
639
640 <dd>This calling convention attempts to make calls as fast as possible
641 (e.g. by passing things in registers). This calling convention allows the
642 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]643 without having to conform to an externally specified ABI (Application Binary
644 Interface). Implementations of this convention should allow arbitrary
Arnold Schwaighofer07444922008-05-14 09:17:12 +0000[diff] [blame]645 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
646 supported. This calling convention does not support varargs and requires the
647 prototype of all callees to exactly match the prototype of the function
648 definition.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]649 </dd>
650
651 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
652
653 <dd>This calling convention attempts to make code in the caller as efficient
654 as possible under the assumption that the call is not commonly executed. As
655 such, these calls often preserve all registers so that the call does not break
656 any live ranges in the caller side. This calling convention does not support
657 varargs and requires the prototype of all callees to exactly match the
658 prototype of the function definition.
659 </dd>
660
661 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
662
663 <dd>Any calling convention may be specified by number, allowing
664 target-specific calling conventions to be used. Target specific calling
665 conventions start at 64.
666 </dd>
667</dl>
668
669<p>More calling conventions can be added/defined on an as-needed basis, to
670support pascal conventions or any other well-known target-independent
671convention.</p>
672
673</div>
674
675<!-- ======================================================================= -->
676<div class="doc_subsection">
677 <a name="visibility">Visibility Styles</a>
678</div>
679
680<div class="doc_text">
681
682<p>
683All Global Variables and Functions have one of the following visibility styles:
684</p>
685
686<dl>
687 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
688
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]689 <dd>On targets that use the ELF object file format, default visibility means
690 that the declaration is visible to other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]691 modules and, in shared libraries, means that the declared entity may be
692 overridden. On Darwin, default visibility means that the declaration is
693 visible to other modules. Default visibility corresponds to "external
694 linkage" in the language.
695 </dd>
696
697 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
698
699 <dd>Two declarations of an object with hidden visibility refer to the same
700 object if they are in the same shared object. Usually, hidden visibility
701 indicates that the symbol will not be placed into the dynamic symbol table,
702 so no other module (executable or shared library) can reference it
703 directly.
704 </dd>
705
706 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
707
708 <dd>On ELF, protected visibility indicates that the symbol will be placed in
709 the dynamic symbol table, but that references within the defining module will
710 bind to the local symbol. That is, the symbol cannot be overridden by another
711 module.
712 </dd>
713</dl>
714
715</div>
716
717<!-- ======================================================================= -->
718<div class="doc_subsection">
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]719 <a name="namedtypes">Named Types</a>
720</div>
721
722<div class="doc_text">
723
724<p>LLVM IR allows you to specify name aliases for certain types. This can make
725it easier to read the IR and make the IR more condensed (particularly when
726recursive types are involved). An example of a name specification is:
727</p>
728
729<div class="doc_code">
730<pre>
731%mytype = type { %mytype*, i32 }
732</pre>
733</div>
734
735<p>You may give a name to any <a href="#typesystem">type</a> except "<a
736href="t_void">void</a>". Type name aliases may be used anywhere a type is
737expected with the syntax "%mytype".</p>
738
739<p>Note that type names are aliases for the structural type that they indicate,
740and that you can therefore specify multiple names for the same type. This often
741leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
742structural typing, the name is not part of the type. When printing out LLVM IR,
743the printer will pick <em>one name</em> to render all types of a particular
744shape. This means that if you have code where two different source types end up
745having the same LLVM type, that the dumper will sometimes print the "wrong" or
746unexpected type. This is an important design point and isn't going to
747change.</p>
748
749</div>
750
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]751<!-- ======================================================================= -->
752<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]753 <a name="globalvars">Global Variables</a>
754</div>
755
756<div class="doc_text">
757
758<p>Global variables define regions of memory allocated at compilation time
759instead of run-time. Global variables may optionally be initialized, may have
760an explicit section to be placed in, and may have an optional explicit alignment
761specified. A variable may be defined as "thread_local", which means that it
762will not be shared by threads (each thread will have a separated copy of the
763variable). A variable may be defined as a global "constant," which indicates
764that the contents of the variable will <b>never</b> be modified (enabling better
765optimization, allowing the global data to be placed in the read-only section of
766an executable, etc). Note that variables that need runtime initialization
767cannot be marked "constant" as there is a store to the variable.</p>
768
769<p>
770LLVM explicitly allows <em>declarations</em> of global variables to be marked
771constant, even if the final definition of the global is not. This capability
772can be used to enable slightly better optimization of the program, but requires
773the language definition to guarantee that optimizations based on the
774'constantness' are valid for the translation units that do not include the
775definition.
776</p>
777
778<p>As SSA values, global variables define pointer values that are in
779scope (i.e. they dominate) all basic blocks in the program. Global
780variables always define a pointer to their "content" type because they
781describe a region of memory, and all memory objects in LLVM are
782accessed through pointers.</p>
783
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]784<p>A global variable may be declared to reside in a target-specifc numbered
785address space. For targets that support them, address spaces may affect how
786optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]787the variable. The default address space is zero. The address space qualifier
788must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]789
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]790<p>LLVM allows an explicit section to be specified for globals. If the target
791supports it, it will emit globals to the section specified.</p>
792
793<p>An explicit alignment may be specified for a global. If not present, or if
794the alignment is set to zero, the alignment of the global is set by the target
795to whatever it feels convenient. If an explicit alignment is specified, the
796global is forced to have at least that much alignment. All alignments must be
797a power of 2.</p>
798
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]799<p>For example, the following defines a global in a numbered address space with
800an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]801
802<div class="doc_code">
803<pre>
Dan Gohman21ef02c2009-01-11 00:40:00 +0000[diff] [blame]804@G = addrspace(5) constant float 1.0, section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]805</pre>
806</div>
807
808</div>
809
810
811<!-- ======================================================================= -->
812<div class="doc_subsection">
813 <a name="functionstructure">Functions</a>
814</div>
815
816<div class="doc_text">
817
818<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
819an optional <a href="#linkage">linkage type</a>, an optional
820<a href="#visibility">visibility style</a>, an optional
821<a href="#callingconv">calling convention</a>, a return type, an optional
822<a href="#paramattrs">parameter attribute</a> for the return type, a function
823name, a (possibly empty) argument list (each with optional
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]824<a href="#paramattrs">parameter attributes</a>), optional
825<a href="#fnattrs">function attributes</a>, an optional section,
826an optional alignment, an optional <a href="#gc">garbage collector name</a>,
Chris Lattneree202542008-10-04 18:10:21 +0000[diff] [blame]827an opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]828
829LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
830optional <a href="#linkage">linkage type</a>, an optional
831<a href="#visibility">visibility style</a>, an optional
832<a href="#callingconv">calling convention</a>, a return type, an optional
833<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]834name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]835<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]836
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]837<p>A function definition contains a list of basic blocks, forming the CFG
838(Control Flow Graph) for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]839the function. Each basic block may optionally start with a label (giving the
840basic block a symbol table entry), contains a list of instructions, and ends
841with a <a href="#terminators">terminator</a> instruction (such as a branch or
842function return).</p>
843
844<p>The first basic block in a function is special in two ways: it is immediately
845executed on entrance to the function, and it is not allowed to have predecessor
846basic blocks (i.e. there can not be any branches to the entry block of a
847function). Because the block can have no predecessors, it also cannot have any
848<a href="#i_phi">PHI nodes</a>.</p>
849
850<p>LLVM allows an explicit section to be specified for functions. If the target
851supports it, it will emit functions to the section specified.</p>
852
853<p>An explicit alignment may be specified for a function. If not present, or if
854the alignment is set to zero, the alignment of the function is set by the target
855to whatever it feels convenient. If an explicit alignment is specified, the
856function is forced to have at least that much alignment. All alignments must be
857a power of 2.</p>
858
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]859 <h5>Syntax:</h5>
860
861<div class="doc_code">
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000[diff] [blame]862<tt>
863define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
864 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
865 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
866 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
867 [<a href="#gc">gc</a>] { ... }
868</tt>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]869</div>
870
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]871</div>
872
873
874<!-- ======================================================================= -->
875<div class="doc_subsection">
876 <a name="aliasstructure">Aliases</a>
877</div>
878<div class="doc_text">
879 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]880 function, global variable, another alias or bitcast of global value). Aliases
881 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]882 optional <a href="#visibility">visibility style</a>.</p>
883
884 <h5>Syntax:</h5>
885
886<div class="doc_code">
887<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000[diff] [blame]888@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]889</pre>
890</div>
891
892</div>
893
894
895
896<!-- ======================================================================= -->
897<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
898<div class="doc_text">
899 <p>The return type and each parameter of a function type may have a set of
900 <i>parameter attributes</i> associated with them. Parameter attributes are
901 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]902 a function. Parameter attributes are considered to be part of the function,
903 not of the function type, so functions with different parameter attributes
904 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]905
906 <p>Parameter attributes are simple keywords that follow the type specified. If
907 multiple parameter attributes are needed, they are space separated. For
908 example:</p>
909
910<div class="doc_code">
911<pre>
Nick Lewycky3022a742009-02-15 23:06:14 +0000[diff] [blame]912declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]913declare i32 @atoi(i8 zeroext)
914declare signext i8 @returns_signed_char()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]915</pre>
916</div>
917
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]918 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
919 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]920
921 <p>Currently, only the following parameter attributes are defined:</p>
922 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]923 <dt><tt>zeroext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]924 <dd>This indicates to the code generator that the parameter or return value
925 should be zero-extended to a 32-bit value by the caller (for a parameter)
926 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]927
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]928 <dt><tt>signext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]929 <dd>This indicates to the code generator that the parameter or return value
930 should be sign-extended to a 32-bit value by the caller (for a parameter)
931 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]932
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]933 <dt><tt>inreg</tt></dt>
Dale Johannesenc08a0e22008-09-25 20:47:45 +0000[diff] [blame]934 <dd>This indicates that this parameter or return value should be treated
935 in a special target-dependent fashion during while emitting code for a
936 function call or return (usually, by putting it in a register as opposed
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]937 to memory, though some targets use it to distinguish between two different
938 kinds of registers). Use of this attribute is target-specific.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]939
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]940 <dt><tt><a name="byval">byval</a></tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]941 <dd>This indicates that the pointer parameter should really be passed by
942 value to the function. The attribute implies that a hidden copy of the
943 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]944 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]945 pointer arguments. It is generally used to pass structs and arrays by
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]946 value, but is also valid on pointers to scalars. The copy is considered to
947 belong to the caller not the callee (for example,
948 <tt><a href="#readonly">readonly</a></tt> functions should not write to
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]949 <tt>byval</tt> parameters). This is not a valid attribute for return
Chris Lattnere39c5bf2009-02-05 05:42:28 +0000[diff] [blame]950 values. The byval attribute also supports specifying an alignment with the
951 align attribute. This has a target-specific effect on the code generator
952 that usually indicates a desired alignment for the synthesized stack
953 slot.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]954
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]955 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]956 <dd>This indicates that the pointer parameter specifies the address of a
957 structure that is the return value of the function in the source program.
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]958 This pointer must be guaranteed by the caller to be valid: loads and stores
959 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]960 be applied to the first parameter. This is not a valid attribute for
961 return values. </dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]962
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]963 <dt><tt>noalias</tt></dt>
Nick Lewyckyff384472008-11-24 03:41:24 +0000[diff] [blame]964 <dd>This indicates that the pointer does not alias any global or any other
965 parameter. The caller is responsible for ensuring that this is the
Nick Lewyckya604a942008-11-24 05:00:44 +0000[diff] [blame]966 case. On a function return value, <tt>noalias</tt> additionally indicates
967 that the pointer does not alias any other pointers visible to the
Nick Lewycky40fb6f22008-12-19 06:39:12 +0000[diff] [blame]968 caller. For further details, please see the discussion of the NoAlias
969 response in
970 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
971 analysis</a>.</dd>
972
973 <dt><tt>nocapture</tt></dt>
974 <dd>This indicates that the callee does not make any copies of the pointer
975 that outlive the callee itself. This is not a valid attribute for return
976 values.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]977
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]978 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]979 <dd>This indicates that the pointer parameter can be excised using the
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]980 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
981 attribute for return values.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]982 </dl>
983
984</div>
985
986<!-- ======================================================================= -->
987<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]988 <a name="gc">Garbage Collector Names</a>
989</div>
990
991<div class="doc_text">
992<p>Each function may specify a garbage collector name, which is simply a
993string.</p>
994
995<div class="doc_code"><pre
996>define void @f() gc "name" { ...</pre></div>
997
998<p>The compiler declares the supported values of <i>name</i>. Specifying a
999collector which will cause the compiler to alter its output in order to support
1000the named garbage collection algorithm.</p>
1001</div>
1002
1003<!-- ======================================================================= -->
1004<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1005 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1006</div>
1007
1008<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1009
1010<p>Function attributes are set to communicate additional information about
1011 a function. Function attributes are considered to be part of the function,
1012 not of the function type, so functions with different parameter attributes
1013 can have the same function type.</p>
1014
1015 <p>Function attributes are simple keywords that follow the type specified. If
1016 multiple attributes are needed, they are space separated. For
1017 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1018
1019<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1020<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1021define void @f() noinline { ... }
1022define void @f() alwaysinline { ... }
1023define void @f() alwaysinline optsize { ... }
1024define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1025</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1026</div>
1027
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1028<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1029<dt><tt>alwaysinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1030<dd>This attribute indicates that the inliner should attempt to inline this
1031function into callers whenever possible, ignoring any active inlining size
1032threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1033
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1034<dt><tt>noinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1035<dd>This attribute indicates that the inliner should never inline this function
Chris Lattner377f3ae2008-10-05 17:14:59 +0000[diff] [blame]1036in any situation. This attribute may not be used together with the
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1037<tt>alwaysinline</tt> attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1038
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1039<dt><tt>optsize</tt></dt>
Devang Patelc29099b2008-09-29 18:34:44 +0000[diff] [blame]1040<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1041make choices that keep the code size of this function low, and otherwise do
1042optimizations specifically to reduce code size.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1043
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1044<dt><tt>noreturn</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1045<dd>This function attribute indicates that the function never returns normally.
1046This produces undefined behavior at runtime if the function ever does
1047dynamically return.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1048
1049<dt><tt>nounwind</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1050<dd>This function attribute indicates that the function never returns with an
1051unwind or exceptional control flow. If the function does unwind, its runtime
1052behavior is undefined.</dd>
1053
1054<dt><tt>readnone</tt></dt>
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]1055<dd>This attribute indicates that the function computes its result (or the
1056exception it throws) based strictly on its arguments, without dereferencing any
1057pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1058registers, etc) visible to caller functions. It does not write through any
1059pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
1060never changes any state visible to callers.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1061
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]1062<dt><tt><a name="readonly">readonly</a></tt></dt>
1063<dd>This attribute indicates that the function does not write through any
1064pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1065or otherwise modify any state (e.g. memory, control registers, etc) visible to
1066caller functions. It may dereference pointer arguments and read state that may
1067be set in the caller. A readonly function always returns the same value (or
1068throws the same exception) when called with the same set of arguments and global
1069state.</dd>
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1070
1071<dt><tt><a name="ssp">ssp</a></tt></dt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1072<dd>This attribute indicates that the function should emit a stack smashing
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1073protector. It is in the form of a "canary"&mdash;a random value placed on the
1074stack before the local variables that's checked upon return from the function to
1075see if it has been overwritten. A heuristic is used to determine if a function
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1076needs stack protectors or not.
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1077
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +0000[diff] [blame]1078<p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1079that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1080have an <tt>ssp</tt> attribute.</p></dd>
1081
1082<dt><tt>sspreq</tt></dt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1083<dd>This attribute indicates that the function should <em>always</em> emit a
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1084stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1085function attribute.
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +0000[diff] [blame]1086
1087<p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1088function that doesn't have an <tt>sspreq</tt> attribute or which has
1089an <tt>ssp</tt> attribute, then the resulting function will have
1090an <tt>sspreq</tt> attribute.</p></dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1091</dl>
1092
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1093</div>
1094
1095<!-- ======================================================================= -->
1096<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1097 <a name="moduleasm">Module-Level Inline Assembly</a>
1098</div>
1099
1100<div class="doc_text">
1101<p>
1102Modules may contain "module-level inline asm" blocks, which corresponds to the
1103GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1104LLVM and treated as a single unit, but may be separated in the .ll file if
1105desired. The syntax is very simple:
1106</p>
1107
1108<div class="doc_code">
1109<pre>
1110module asm "inline asm code goes here"
1111module asm "more can go here"
1112</pre>
1113</div>
1114
1115<p>The strings can contain any character by escaping non-printable characters.
1116 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1117 for the number.
1118</p>
1119
1120<p>
1121 The inline asm code is simply printed to the machine code .s file when
1122 assembly code is generated.
1123</p>
1124</div>
1125
1126<!-- ======================================================================= -->
1127<div class="doc_subsection">
1128 <a name="datalayout">Data Layout</a>
1129</div>
1130
1131<div class="doc_text">
1132<p>A module may specify a target specific data layout string that specifies how
1133data is to be laid out in memory. The syntax for the data layout is simply:</p>
1134<pre> target datalayout = "<i>layout specification</i>"</pre>
1135<p>The <i>layout specification</i> consists of a list of specifications
1136separated by the minus sign character ('-'). Each specification starts with a
1137letter and may include other information after the letter to define some
1138aspect of the data layout. The specifications accepted are as follows: </p>
1139<dl>
1140 <dt><tt>E</tt></dt>
1141 <dd>Specifies that the target lays out data in big-endian form. That is, the
1142 bits with the most significance have the lowest address location.</dd>
1143 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1144 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1145 the bits with the least significance have the lowest address location.</dd>
1146 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1147 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1148 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1149 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1150 too.</dd>
1151 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1152 <dd>This specifies the alignment for an integer type of a given bit
1153 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1154 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1155 <dd>This specifies the alignment for a vector type of a given bit
1156 <i>size</i>.</dd>
1157 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1158 <dd>This specifies the alignment for a floating point type of a given bit
1159 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1160 (double).</dd>
1161 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1162 <dd>This specifies the alignment for an aggregate type of a given bit
1163 <i>size</i>.</dd>
1164</dl>
1165<p>When constructing the data layout for a given target, LLVM starts with a
1166default set of specifications which are then (possibly) overriden by the
1167specifications in the <tt>datalayout</tt> keyword. The default specifications
1168are given in this list:</p>
1169<ul>
1170 <li><tt>E</tt> - big endian</li>
1171 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1172 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1173 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1174 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1175 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1176 <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]1177 alignment of 64-bits</li>
1178 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1179 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1180 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1181 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1182 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1183</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1184<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1185following rules:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1186<ol>
1187 <li>If the type sought is an exact match for one of the specifications, that
1188 specification is used.</li>
1189 <li>If no match is found, and the type sought is an integer type, then the
1190 smallest integer type that is larger than the bitwidth of the sought type is
1191 used. If none of the specifications are larger than the bitwidth then the the
1192 largest integer type is used. For example, given the default specifications
1193 above, the i7 type will use the alignment of i8 (next largest) while both
1194 i65 and i256 will use the alignment of i64 (largest specified).</li>
1195 <li>If no match is found, and the type sought is a vector type, then the
1196 largest vector type that is smaller than the sought vector type will be used
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1197 as a fall back. This happens because &lt;128 x double&gt; can be implemented
1198 in terms of 64 &lt;2 x double&gt;, for example.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1199</ol>
1200</div>
1201
1202<!-- *********************************************************************** -->
1203<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1204<!-- *********************************************************************** -->
1205
1206<div class="doc_text">
1207
1208<p>The LLVM type system is one of the most important features of the
1209intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1210optimizations to be performed on the intermediate representation directly,
1211without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1212extra analyses on the side before the transformation. A strong type
1213system makes it easier to read the generated code and enables novel
1214analyses and transformations that are not feasible to perform on normal
1215three address code representations.</p>
1216
1217</div>
1218
1219<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1220<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1221Classifications</a> </div>
1222<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1223<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1224classifications:</p>
1225
1226<table border="1" cellspacing="0" cellpadding="4">
1227 <tbody>
1228 <tr><th>Classification</th><th>Types</th></tr>
1229 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1230 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1231 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1232 </tr>
1233 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1234 <td><a href="#t_floating">floating point</a></td>
1235 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1236 </tr>
1237 <tr>
1238 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1239 <td><a href="#t_integer">integer</a>,
1240 <a href="#t_floating">floating point</a>,
1241 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1242 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1243 <a href="#t_struct">structure</a>,
1244 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1245 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1246 </td>
1247 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1248 <tr>
1249 <td><a href="#t_primitive">primitive</a></td>
1250 <td><a href="#t_label">label</a>,
1251 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1252 <a href="#t_floating">floating point</a>.</td>
1253 </tr>
1254 <tr>
1255 <td><a href="#t_derived">derived</a></td>
1256 <td><a href="#t_integer">integer</a>,
1257 <a href="#t_array">array</a>,
1258 <a href="#t_function">function</a>,
1259 <a href="#t_pointer">pointer</a>,
1260 <a href="#t_struct">structure</a>,
1261 <a href="#t_pstruct">packed structure</a>,
1262 <a href="#t_vector">vector</a>,
1263 <a href="#t_opaque">opaque</a>.
Dan Gohman032ba852008-10-14 16:32:04 +0000[diff] [blame]1264 </td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1265 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1266 </tbody>
1267</table>
1268
1269<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1270most important. Values of these types are the only ones which can be
1271produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1272instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1273</div>
1274
1275<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1276<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1277
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1278<div class="doc_text">
1279<p>The primitive types are the fundamental building blocks of the LLVM
1280system.</p>
1281
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1282</div>
1283
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1284<!-- _______________________________________________________________________ -->
1285<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1286
1287<div class="doc_text">
1288 <table>
1289 <tbody>
1290 <tr><th>Type</th><th>Description</th></tr>
1291 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1292 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1293 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1294 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1295 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1296 </tbody>
1297 </table>
1298</div>
1299
1300<!-- _______________________________________________________________________ -->
1301<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1302
1303<div class="doc_text">
1304<h5>Overview:</h5>
1305<p>The void type does not represent any value and has no size.</p>
1306
1307<h5>Syntax:</h5>
1308
1309<pre>
1310 void
1311</pre>
1312</div>
1313
1314<!-- _______________________________________________________________________ -->
1315<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1316
1317<div class="doc_text">
1318<h5>Overview:</h5>
1319<p>The label type represents code labels.</p>
1320
1321<h5>Syntax:</h5>
1322
1323<pre>
1324 label
1325</pre>
1326</div>
1327
1328
1329<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1330<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1331
1332<div class="doc_text">
1333
1334<p>The real power in LLVM comes from the derived types in the system.
1335This is what allows a programmer to represent arrays, functions,
1336pointers, and other useful types. Note that these derived types may be
1337recursive: For example, it is possible to have a two dimensional array.</p>
1338
1339</div>
1340
1341<!-- _______________________________________________________________________ -->
1342<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1343
1344<div class="doc_text">
1345
1346<h5>Overview:</h5>
1347<p>The integer type is a very simple derived type that simply specifies an
1348arbitrary bit width for the integer type desired. Any bit width from 1 bit to
13492^23-1 (about 8 million) can be specified.</p>
1350
1351<h5>Syntax:</h5>
1352
1353<pre>
1354 iN
1355</pre>
1356
1357<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1358value.</p>
1359
1360<h5>Examples:</h5>
1361<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1362 <tbody>
1363 <tr>
1364 <td><tt>i1</tt></td>
1365 <td>a single-bit integer.</td>
1366 </tr><tr>
1367 <td><tt>i32</tt></td>
1368 <td>a 32-bit integer.</td>
1369 </tr><tr>
1370 <td><tt>i1942652</tt></td>
1371 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1372 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1373 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1374</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1375
1376<p>Note that the code generator does not yet support large integer types
1377to be used as function return types. The specific limit on how large a
1378return type the code generator can currently handle is target-dependent;
1379currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1380targets.</p>
1381
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1382</div>
1383
1384<!-- _______________________________________________________________________ -->
1385<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1386
1387<div class="doc_text">
1388
1389<h5>Overview:</h5>
1390
1391<p>The array type is a very simple derived type that arranges elements
1392sequentially in memory. The array type requires a size (number of
1393elements) and an underlying data type.</p>
1394
1395<h5>Syntax:</h5>
1396
1397<pre>
1398 [&lt;# elements&gt; x &lt;elementtype&gt;]
1399</pre>
1400
1401<p>The number of elements is a constant integer value; elementtype may
1402be any type with a size.</p>
1403
1404<h5>Examples:</h5>
1405<table class="layout">
1406 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1407 <td class="left"><tt>[40 x i32]</tt></td>
1408 <td class="left">Array of 40 32-bit integer values.</td>
1409 </tr>
1410 <tr class="layout">
1411 <td class="left"><tt>[41 x i32]</tt></td>
1412 <td class="left">Array of 41 32-bit integer values.</td>
1413 </tr>
1414 <tr class="layout">
1415 <td class="left"><tt>[4 x i8]</tt></td>
1416 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1417 </tr>
1418</table>
1419<p>Here are some examples of multidimensional arrays:</p>
1420<table class="layout">
1421 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1422 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1423 <td class="left">3x4 array of 32-bit integer values.</td>
1424 </tr>
1425 <tr class="layout">
1426 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1427 <td class="left">12x10 array of single precision floating point values.</td>
1428 </tr>
1429 <tr class="layout">
1430 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1431 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1432 </tr>
1433</table>
1434
1435<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1436length array. Normally, accesses past the end of an array are undefined in
1437LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1438As a special case, however, zero length arrays are recognized to be variable
1439length. This allows implementation of 'pascal style arrays' with the LLVM
1440type "{ i32, [0 x float]}", for example.</p>
1441
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1442<p>Note that the code generator does not yet support large aggregate types
1443to be used as function return types. The specific limit on how large an
1444aggregate return type the code generator can currently handle is
1445target-dependent, and also dependent on the aggregate element types.</p>
1446
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1447</div>
1448
1449<!-- _______________________________________________________________________ -->
1450<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1451<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1452
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1453<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1454
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1455<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1456consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1457return 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]1458If 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]1459class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1460
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1461<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1462
1463<pre>
1464 &lt;returntype list&gt; (&lt;parameter list&gt;)
1465</pre>
1466
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1467<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1468specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1469which indicates that the function takes a variable number of arguments.
1470Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1471 href="#int_varargs">variable argument handling intrinsic</a> functions.
1472'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1473<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1474
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1475<h5>Examples:</h5>
1476<table class="layout">
1477 <tr class="layout">
1478 <td class="left"><tt>i32 (i32)</tt></td>
1479 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1480 </td>
1481 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1482 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1483 </tt></td>
1484 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1485 an <tt>i16</tt> that should be sign extended and a
1486 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1487 <tt>float</tt>.
1488 </td>
1489 </tr><tr class="layout">
1490 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1491 <td class="left">A vararg function that takes at least one
1492 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1493 which returns an integer. This is the signature for <tt>printf</tt> in
1494 LLVM.
1495 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1496 </tr><tr class="layout">
1497 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Misha Brukmanafc88b02008-11-27 06:41:20 +0000[diff] [blame]1498 <td class="left">A function taking an <tt>i32</tt>, returning two
1499 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1500 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1501 </tr>
1502</table>
1503
1504</div>
1505<!-- _______________________________________________________________________ -->
1506<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1507<div class="doc_text">
1508<h5>Overview:</h5>
1509<p>The structure type is used to represent a collection of data members
1510together in memory. The packing of the field types is defined to match
1511the ABI of the underlying processor. The elements of a structure may
1512be any type that has a size.</p>
1513<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1514and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1515field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1516instruction.</p>
1517<h5>Syntax:</h5>
1518<pre> { &lt;type list&gt; }<br></pre>
1519<h5>Examples:</h5>
1520<table class="layout">
1521 <tr class="layout">
1522 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1523 <td class="left">A triple of three <tt>i32</tt> values</td>
1524 </tr><tr class="layout">
1525 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1526 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1527 second element is a <a href="#t_pointer">pointer</a> to a
1528 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1529 an <tt>i32</tt>.</td>
1530 </tr>
1531</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1532
1533<p>Note that the code generator does not yet support large aggregate types
1534to be used as function return types. The specific limit on how large an
1535aggregate return type the code generator can currently handle is
1536target-dependent, and also dependent on the aggregate element types.</p>
1537
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1538</div>
1539
1540<!-- _______________________________________________________________________ -->
1541<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1542</div>
1543<div class="doc_text">
1544<h5>Overview:</h5>
1545<p>The packed structure type is used to represent a collection of data members
1546together in memory. There is no padding between fields. Further, the alignment
1547of a packed structure is 1 byte. The elements of a packed structure may
1548be any type that has a size.</p>
1549<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1550and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1551field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1552instruction.</p>
1553<h5>Syntax:</h5>
1554<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1555<h5>Examples:</h5>
1556<table class="layout">
1557 <tr class="layout">
1558 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1559 <td class="left">A triple of three <tt>i32</tt> values</td>
1560 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1561 <td class="left">
1562<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1563 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1564 second element is a <a href="#t_pointer">pointer</a> to a
1565 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1566 an <tt>i32</tt>.</td>
1567 </tr>
1568</table>
1569</div>
1570
1571<!-- _______________________________________________________________________ -->
1572<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1573<div class="doc_text">
1574<h5>Overview:</h5>
1575<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1576reference to another object, which must live in memory. Pointer types may have
1577an optional address space attribute defining the target-specific numbered
1578address space where the pointed-to object resides. The default address space is
1579zero.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +0000[diff] [blame]1580
1581<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]1582it permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +0000[diff] [blame]1583
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1584<h5>Syntax:</h5>
1585<pre> &lt;type&gt; *<br></pre>
1586<h5>Examples:</h5>
1587<table class="layout">
1588 <tr class="layout">
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]1589 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1590 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1591 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1592 </tr>
1593 <tr class="layout">
1594 <td class="left"><tt>i32 (i32 *) *</tt></td>
1595 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1596 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1597 <tt>i32</tt>.</td>
1598 </tr>
1599 <tr class="layout">
1600 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1601 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1602 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1603 </tr>
1604</table>
1605</div>
1606
1607<!-- _______________________________________________________________________ -->
1608<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1609<div class="doc_text">
1610
1611<h5>Overview:</h5>
1612
1613<p>A vector type is a simple derived type that represents a vector
1614of elements. Vector types are used when multiple primitive data
1615are operated in parallel using a single instruction (SIMD).
1616A vector type requires a size (number of
1617elements) and an underlying primitive data type. Vectors must have a power
1618of two length (1, 2, 4, 8, 16 ...). Vector types are
1619considered <a href="#t_firstclass">first class</a>.</p>
1620
1621<h5>Syntax:</h5>
1622
1623<pre>
1624 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1625</pre>
1626
1627<p>The number of elements is a constant integer value; elementtype may
1628be any integer or floating point type.</p>
1629
1630<h5>Examples:</h5>
1631
1632<table class="layout">
1633 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1634 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1635 <td class="left">Vector of 4 32-bit integer values.</td>
1636 </tr>
1637 <tr class="layout">
1638 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1639 <td class="left">Vector of 8 32-bit floating-point values.</td>
1640 </tr>
1641 <tr class="layout">
1642 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1643 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1644 </tr>
1645</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1646
1647<p>Note that the code generator does not yet support large vector types
1648to be used as function return types. The specific limit on how large a
1649vector return type codegen can currently handle is target-dependent;
1650currently it's often a few times longer than a hardware vector register.</p>
1651
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1652</div>
1653
1654<!-- _______________________________________________________________________ -->
1655<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1656<div class="doc_text">
1657
1658<h5>Overview:</h5>
1659
1660<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1661corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1662In LLVM, opaque types can eventually be resolved to any type (not just a
1663structure type).</p>
1664
1665<h5>Syntax:</h5>
1666
1667<pre>
1668 opaque
1669</pre>
1670
1671<h5>Examples:</h5>
1672
1673<table class="layout">
1674 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1675 <td class="left"><tt>opaque</tt></td>
1676 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1677 </tr>
1678</table>
1679</div>
1680
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]1681<!-- ======================================================================= -->
1682<div class="doc_subsection">
1683 <a name="t_uprefs">Type Up-references</a>
1684</div>
1685
1686<div class="doc_text">
1687<h5>Overview:</h5>
1688<p>
1689An "up reference" allows you to refer to a lexically enclosing type without
1690requiring it to have a name. For instance, a structure declaration may contain a
1691pointer to any of the types it is lexically a member of. Example of up
1692references (with their equivalent as named type declarations) include:</p>
1693
1694<pre>
Chris Lattner5ad632d2009-02-09 10:00:56 +0000[diff] [blame]1695 { \2 * } %x = type { %x* }
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]1696 { \2 }* %y = type { %y }*
1697 \1* %z = type %z*
1698</pre>
1699
1700<p>
1701An up reference is needed by the asmprinter for printing out cyclic types when
1702there is no declared name for a type in the cycle. Because the asmprinter does
1703not want to print out an infinite type string, it needs a syntax to handle
1704recursive types that have no names (all names are optional in llvm IR).
1705</p>
1706
1707<h5>Syntax:</h5>
1708<pre>
1709 \&lt;level&gt;
1710</pre>
1711
1712<p>
1713The level is the count of the lexical type that is being referred to.
1714</p>
1715
1716<h5>Examples:</h5>
1717
1718<table class="layout">
1719 <tr class="layout">
1720 <td class="left"><tt>\1*</tt></td>
1721 <td class="left">Self-referential pointer.</td>
1722 </tr>
1723 <tr class="layout">
1724 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1725 <td class="left">Recursive structure where the upref refers to the out-most
1726 structure.</td>
1727 </tr>
1728</table>
1729</div>
1730
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1731
1732<!-- *********************************************************************** -->
1733<div class="doc_section"> <a name="constants">Constants</a> </div>
1734<!-- *********************************************************************** -->
1735
1736<div class="doc_text">
1737
1738<p>LLVM has several different basic types of constants. This section describes
1739them all and their syntax.</p>
1740
1741</div>
1742
1743<!-- ======================================================================= -->
1744<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1745
1746<div class="doc_text">
1747
1748<dl>
1749 <dt><b>Boolean constants</b></dt>
1750
1751 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1752 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1753 </dd>
1754
1755 <dt><b>Integer constants</b></dt>
1756
1757 <dd>Standard integers (such as '4') are constants of the <a
1758 href="#t_integer">integer</a> type. Negative numbers may be used with
1759 integer types.
1760 </dd>
1761
1762 <dt><b>Floating point constants</b></dt>
1763
1764 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1765 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1766 notation (see below). The assembler requires the exact decimal value of
1767 a floating-point constant. For example, the assembler accepts 1.25 but
1768 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1769 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1770
1771 <dt><b>Null pointer constants</b></dt>
1772
1773 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1774 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1775
1776</dl>
1777
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1778<p>The one non-intuitive notation for constants is the hexadecimal form
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1779of floating point constants. For example, the form '<tt>double
17800x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
17814.5e+15</tt>'. The only time hexadecimal floating point constants are required
1782(and the only time that they are generated by the disassembler) is when a
1783floating point constant must be emitted but it cannot be represented as a
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1784decimal floating point number in a reasonable number of digits. For example,
1785NaN's, infinities, and other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1786special values are represented in their IEEE hexadecimal format so that
1787assembly and disassembly do not cause any bits to change in the constants.</p>
Dale Johannesenf82a52f2009-02-11 22:14:51 +0000[diff] [blame]1788<p>When using the hexadecimal form, constants of types float and double are
1789represented using the 16-digit form shown above (which matches the IEEE754
1790representation for double); float values must, however, be exactly representable
1791as IEE754 single precision.
1792Hexadecimal format is always used for long
1793double, and there are three forms of long double. The 80-bit
1794format used by x86 is represented as <tt>0xK</tt>
1795followed by 20 hexadecimal digits.
1796The 128-bit format used by PowerPC (two adjacent doubles) is represented
1797by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit
1798format is represented
1799by <tt>0xL</tt> followed by 32 hexadecimal digits; no currently supported
1800target uses this format. Long doubles will only work if they match
1801the long double format on your target. All hexadecimal formats are big-endian
1802(sign bit at the left).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1803</div>
1804
1805<!-- ======================================================================= -->
Chris Lattner97063852009-02-28 18:32:25 +0000[diff] [blame]1806<div class="doc_subsection">
1807<a name="aggregateconstants"> <!-- old anchor -->
1808<a name="complexconstants">Complex Constants</a></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1809</div>
1810
1811<div class="doc_text">
Chris Lattner97063852009-02-28 18:32:25 +0000[diff] [blame]1812<p>Complex constants are a (potentially recursive) combination of simple
1813constants and smaller complex constants.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1814
1815<dl>
1816 <dt><b>Structure constants</b></dt>
1817
1818 <dd>Structure constants are represented with notation similar to structure
1819 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1820 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1821 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]1822 must have <a href="#t_struct">structure type</a>, and the number and
1823 types of elements must match those specified by the type.
1824 </dd>
1825
1826 <dt><b>Array constants</b></dt>
1827
1828 <dd>Array constants are represented with notation similar to array type
1829 definitions (a comma separated list of elements, surrounded by square brackets
1830 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1831 constants must have <a href="#t_array">array type</a>, and the number and
1832 types of elements must match those specified by the type.
1833 </dd>
1834
1835 <dt><b>Vector constants</b></dt>
1836
1837 <dd>Vector constants are represented with notation similar to vector type
1838 definitions (a comma separated list of elements, surrounded by
1839 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1840 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1841 href="#t_vector">vector type</a>, and the number and types of elements must
1842 match those specified by the type.
1843 </dd>
1844
1845 <dt><b>Zero initialization</b></dt>
1846
1847 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1848 value to zero of <em>any</em> type, including scalar and aggregate types.
1849 This is often used to avoid having to print large zero initializers (e.g. for
1850 large arrays) and is always exactly equivalent to using explicit zero
1851 initializers.
1852 </dd>
1853</dl>
1854
1855</div>
1856
1857<!-- ======================================================================= -->
1858<div class="doc_subsection">
1859 <a name="globalconstants">Global Variable and Function Addresses</a>
1860</div>
1861
1862<div class="doc_text">
1863
1864<p>The addresses of <a href="#globalvars">global variables</a> and <a
1865href="#functionstructure">functions</a> are always implicitly valid (link-time)
1866constants. These constants are explicitly referenced when the <a
1867href="#identifiers">identifier for the global</a> is used and always have <a
1868href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1869file:</p>
1870
1871<div class="doc_code">
1872<pre>
1873@X = global i32 17
1874@Y = global i32 42
1875@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1876</pre>
1877</div>
1878
1879</div>
1880
1881<!-- ======================================================================= -->
1882<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1883<div class="doc_text">
1884 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1885 no specific value. Undefined values may be of any type and be used anywhere
1886 a constant is permitted.</p>
1887
1888 <p>Undefined values indicate to the compiler that the program is well defined
1889 no matter what value is used, giving the compiler more freedom to optimize.
1890 </p>
1891</div>
1892
1893<!-- ======================================================================= -->
1894<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1895</div>
1896
1897<div class="doc_text">
1898
1899<p>Constant expressions are used to allow expressions involving other constants
1900to be used as constants. Constant expressions may be of any <a
1901href="#t_firstclass">first class</a> type and may involve any LLVM operation
1902that does not have side effects (e.g. load and call are not supported). The
1903following is the syntax for constant expressions:</p>
1904
1905<dl>
1906 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1907 <dd>Truncate a constant to another type. The bit size of CST must be larger
1908 than the bit size of TYPE. Both types must be integers.</dd>
1909
1910 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1911 <dd>Zero extend a constant to another type. The bit size of CST must be
1912 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1913
1914 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1915 <dd>Sign extend a constant to another type. The bit size of CST must be
1916 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1917
1918 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1919 <dd>Truncate a floating point constant to another floating point type. The
1920 size of CST must be larger than the size of TYPE. Both types must be
1921 floating point.</dd>
1922
1923 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1924 <dd>Floating point extend a constant to another type. The size of CST must be
1925 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1926
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1927 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1928 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1929 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1930 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1931 of the same number of elements. If the value won't fit in the integer type,
1932 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1933
1934 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1935 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1936 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1937 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1938 of the same number of elements. If the value won't fit in the integer type,
1939 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1940
1941 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1942 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1943 constant. TYPE must be a scalar or vector floating point type. CST must be of
1944 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1945 of the same number of elements. If the value won't fit in the floating point
1946 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1947
1948 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1949 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1950 constant. TYPE must be a scalar or vector floating point type. CST must be of
1951 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1952 of the same number of elements. If the value won't fit in the floating point
1953 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1954
1955 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1956 <dd>Convert a pointer typed constant to the corresponding integer constant
1957 TYPE must be an integer type. CST must be of pointer type. The CST value is
1958 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1959
1960 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1961 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1962 pointer type. CST must be of integer type. The CST value is zero extended,
1963 truncated, or unchanged to make it fit in a pointer size. This one is
1964 <i>really</i> dangerous!</dd>
1965
1966 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
Chris Lattner557bc5d2009-02-28 18:27:03 +0000[diff] [blame]1967 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
1968 are the same as those for the <a href="#i_bitcast">bitcast
1969 instruction</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1970
1971 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1972
1973 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1974 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1975 instruction, the index list may have zero or more indexes, which are required
1976 to make sense for the type of "CSTPTR".</dd>
1977
1978 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1979
1980 <dd>Perform the <a href="#i_select">select operation</a> on
1981 constants.</dd>
1982
1983 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1984 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1985
1986 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1987 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1988
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1989 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1990 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1991
1992 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1993 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1994
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1995 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1996
1997 <dd>Perform the <a href="#i_extractelement">extractelement
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1998 operation</a> on constants.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1999
2000 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2001
2002 <dd>Perform the <a href="#i_insertelement">insertelement
2003 operation</a> on constants.</dd>
2004
2005
2006 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2007
2008 <dd>Perform the <a href="#i_shufflevector">shufflevector
2009 operation</a> on constants.</dd>
2010
2011 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2012
2013 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2014 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
2015 binary</a> operations. The constraints on operands are the same as those for
2016 the corresponding instruction (e.g. no bitwise operations on floating point
2017 values are allowed).</dd>
2018</dl>
2019</div>
2020
2021<!-- *********************************************************************** -->
2022<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2023<!-- *********************************************************************** -->
2024
2025<!-- ======================================================================= -->
2026<div class="doc_subsection">
2027<a name="inlineasm">Inline Assembler Expressions</a>
2028</div>
2029
2030<div class="doc_text">
2031
2032<p>
2033LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
2034Module-Level Inline Assembly</a>) through the use of a special value. This
2035value represents the inline assembler as a string (containing the instructions
2036to emit), a list of operand constraints (stored as a string), and a flag that
2037indicates whether or not the inline asm expression has side effects. An example
2038inline assembler expression is:
2039</p>
2040
2041<div class="doc_code">
2042<pre>
2043i32 (i32) asm "bswap $0", "=r,r"
2044</pre>
2045</div>
2046
2047<p>
2048Inline assembler expressions may <b>only</b> be used as the callee operand of
2049a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2050</p>
2051
2052<div class="doc_code">
2053<pre>
2054%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2055</pre>
2056</div>
2057
2058<p>
2059Inline asms with side effects not visible in the constraint list must be marked
2060as having side effects. This is done through the use of the
2061'<tt>sideeffect</tt>' keyword, like so:
2062</p>
2063
2064<div class="doc_code">
2065<pre>
2066call void asm sideeffect "eieio", ""()
2067</pre>
2068</div>
2069
2070<p>TODO: The format of the asm and constraints string still need to be
2071documented here. Constraints on what can be done (e.g. duplication, moving, etc
Chris Lattner1cdd64e2008-10-04 18:36:02 +0000[diff] [blame]2072need to be documented). This is probably best done by reference to another
2073document that covers inline asm from a holistic perspective.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2074</p>
2075
2076</div>
2077
2078<!-- *********************************************************************** -->
2079<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2080<!-- *********************************************************************** -->
2081
2082<div class="doc_text">
2083
2084<p>The LLVM instruction set consists of several different
2085classifications of instructions: <a href="#terminators">terminator
2086instructions</a>, <a href="#binaryops">binary instructions</a>,
2087<a href="#bitwiseops">bitwise binary instructions</a>, <a
2088 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2089instructions</a>.</p>
2090
2091</div>
2092
2093<!-- ======================================================================= -->
2094<div class="doc_subsection"> <a name="terminators">Terminator
2095Instructions</a> </div>
2096
2097<div class="doc_text">
2098
2099<p>As mentioned <a href="#functionstructure">previously</a>, every
2100basic block in a program ends with a "Terminator" instruction, which
2101indicates which block should be executed after the current block is
2102finished. These terminator instructions typically yield a '<tt>void</tt>'
2103value: they produce control flow, not values (the one exception being
2104the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2105<p>There are six different terminator instructions: the '<a
2106 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2107instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
2108the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2109 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2110 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2111
2112</div>
2113
2114<!-- _______________________________________________________________________ -->
2115<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2116Instruction</a> </div>
2117<div class="doc_text">
2118<h5>Syntax:</h5>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2119<pre>
2120 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]2121 ret void <i>; Return from void function</i>
2122</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2123
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2124<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2125
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2126<p>The '<tt>ret</tt>' instruction is used to return control flow (and
2127optionally a value) from a function back to the caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2128<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2129returns a value and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2130control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2131
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2132<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2133
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2134<p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2135the return value. The type of the return value must be a
2136'<a href="#t_firstclass">first class</a>' type.</p>
2137
2138<p>A function is not <a href="#wellformed">well formed</a> if
2139it it has a non-void return type and contains a '<tt>ret</tt>'
2140instruction with no return value or a return value with a type that
2141does not match its type, or if it has a void return type and contains
2142a '<tt>ret</tt>' instruction with a return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2143
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2144<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2145
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2146<p>When the '<tt>ret</tt>' instruction is executed, control flow
2147returns back to the calling function's context. If the caller is a "<a
2148 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2149the instruction after the call. If the caller was an "<a
2150 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2151at the beginning of the "normal" destination block. If the instruction
2152returns a value, that value shall set the call or invoke instruction's
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2153return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2154
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2155<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2156
2157<pre>
2158 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2159 ret void <i>; Return from a void function</i>
Bill Wendlingd163e2d2009-02-28 22:12:54 +0000[diff] [blame]2160 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]2161</pre>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]2162
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]2163<p>Note that the code generator does not yet fully support large
2164 return values. The specific sizes that are currently supported are
2165 dependent on the target. For integers, on 32-bit targets the limit
2166 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2167 For aggregate types, the current limits are dependent on the element
2168 types; for example targets are often limited to 2 total integer
2169 elements and 2 total floating-point elements.</p>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]2170
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2171</div>
2172<!-- _______________________________________________________________________ -->
2173<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2174<div class="doc_text">
2175<h5>Syntax:</h5>
2176<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
2177</pre>
2178<h5>Overview:</h5>
2179<p>The '<tt>br</tt>' instruction is used to cause control flow to
2180transfer to a different basic block in the current function. There are
2181two forms of this instruction, corresponding to a conditional branch
2182and an unconditional branch.</p>
2183<h5>Arguments:</h5>
2184<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2185single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2186unconditional form of the '<tt>br</tt>' instruction takes a single
2187'<tt>label</tt>' value as a target.</p>
2188<h5>Semantics:</h5>
2189<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2190argument is evaluated. If the value is <tt>true</tt>, control flows
2191to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2192control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2193<h5>Example:</h5>
2194<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
2195 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2196</div>
2197<!-- _______________________________________________________________________ -->
2198<div class="doc_subsubsection">
2199 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2200</div>
2201
2202<div class="doc_text">
2203<h5>Syntax:</h5>
2204
2205<pre>
2206 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
2207</pre>
2208
2209<h5>Overview:</h5>
2210
2211<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2212several different places. It is a generalization of the '<tt>br</tt>'
2213instruction, allowing a branch to occur to one of many possible
2214destinations.</p>
2215
2216
2217<h5>Arguments:</h5>
2218
2219<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2220comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2221an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2222table is not allowed to contain duplicate constant entries.</p>
2223
2224<h5>Semantics:</h5>
2225
2226<p>The <tt>switch</tt> instruction specifies a table of values and
2227destinations. When the '<tt>switch</tt>' instruction is executed, this
2228table is searched for the given value. If the value is found, control flow is
2229transfered to the corresponding destination; otherwise, control flow is
2230transfered to the default destination.</p>
2231
2232<h5>Implementation:</h5>
2233
2234<p>Depending on properties of the target machine and the particular
2235<tt>switch</tt> instruction, this instruction may be code generated in different
2236ways. For example, it could be generated as a series of chained conditional
2237branches or with a lookup table.</p>
2238
2239<h5>Example:</h5>
2240
2241<pre>
2242 <i>; Emulate a conditional br instruction</i>
2243 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]2244 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2245
2246 <i>; Emulate an unconditional br instruction</i>
2247 switch i32 0, label %dest [ ]
2248
2249 <i>; Implement a jump table:</i>
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]2250 switch i32 %val, label %otherwise [ i32 0, label %onzero
2251 i32 1, label %onone
2252 i32 2, label %ontwo ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2253</pre>
2254</div>
2255
2256<!-- _______________________________________________________________________ -->
2257<div class="doc_subsubsection">
2258 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2259</div>
2260
2261<div class="doc_text">
2262
2263<h5>Syntax:</h5>
2264
2265<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2266 &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]2267 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2268</pre>
2269
2270<h5>Overview:</h5>
2271
2272<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2273function, with the possibility of control flow transfer to either the
2274'<tt>normal</tt>' label or the
2275'<tt>exception</tt>' label. If the callee function returns with the
2276"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2277"normal" label. If the callee (or any indirect callees) returns with the "<a
2278href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2279continued at the dynamically nearest "exception" label.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2280
2281<h5>Arguments:</h5>
2282
2283<p>This instruction requires several arguments:</p>
2284
2285<ol>
2286 <li>
2287 The optional "cconv" marker indicates which <a href="#callingconv">calling
2288 convention</a> the call should use. If none is specified, the call defaults
2289 to using C calling conventions.
2290 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2291
2292 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2293 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2294 and '<tt>inreg</tt>' attributes are valid here.</li>
2295
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2296 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2297 function value being invoked. In most cases, this is a direct function
2298 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2299 an arbitrary pointer to function value.
2300 </li>
2301
2302 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2303 function to be invoked. </li>
2304
2305 <li>'<tt>function args</tt>': argument list whose types match the function
2306 signature argument types. If the function signature indicates the function
2307 accepts a variable number of arguments, the extra arguments can be
2308 specified. </li>
2309
2310 <li>'<tt>normal label</tt>': the label reached when the called function
2311 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2312
2313 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2314 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2315
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2316 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2317 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2318 '<tt>readnone</tt>' attributes are valid here.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2319</ol>
2320
2321<h5>Semantics:</h5>
2322
2323<p>This instruction is designed to operate as a standard '<tt><a
2324href="#i_call">call</a></tt>' instruction in most regards. The primary
2325difference is that it establishes an association with a label, which is used by
2326the runtime library to unwind the stack.</p>
2327
2328<p>This instruction is used in languages with destructors to ensure that proper
2329cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2330exception. Additionally, this is important for implementation of
2331'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2332
2333<h5>Example:</h5>
2334<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2335 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2336 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2337 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2338 unwind label %TestCleanup <i>; {i32}:retval set</i>
2339</pre>
2340</div>
2341
2342
2343<!-- _______________________________________________________________________ -->
2344
2345<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2346Instruction</a> </div>
2347
2348<div class="doc_text">
2349
2350<h5>Syntax:</h5>
2351<pre>
2352 unwind
2353</pre>
2354
2355<h5>Overview:</h5>
2356
2357<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2358at the first callee in the dynamic call stack which used an <a
2359href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2360primarily used to implement exception handling.</p>
2361
2362<h5>Semantics:</h5>
2363
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2364<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2365immediately halt. The dynamic call stack is then searched for the first <a
2366href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2367execution continues at the "exceptional" destination block specified by the
2368<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2369dynamic call chain, undefined behavior results.</p>
2370</div>
2371
2372<!-- _______________________________________________________________________ -->
2373
2374<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2375Instruction</a> </div>
2376
2377<div class="doc_text">
2378
2379<h5>Syntax:</h5>
2380<pre>
2381 unreachable
2382</pre>
2383
2384<h5>Overview:</h5>
2385
2386<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2387instruction is used to inform the optimizer that a particular portion of the
2388code is not reachable. This can be used to indicate that the code after a
2389no-return function cannot be reached, and other facts.</p>
2390
2391<h5>Semantics:</h5>
2392
2393<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2394</div>
2395
2396
2397
2398<!-- ======================================================================= -->
2399<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2400<div class="doc_text">
2401<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2402program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2403produce a single value. The operands might represent
2404multiple 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]2405The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2406<p>There are several different binary operators:</p>
2407</div>
2408<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2409<div class="doc_subsubsection">
2410 <a name="i_add">'<tt>add</tt>' Instruction</a>
2411</div>
2412
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2413<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2414
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2415<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2416
2417<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2418 &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]2419</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2420
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2421<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2422
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2423<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2424
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2425<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2426
2427<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2428 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2429 <a href="#t_vector">vector</a> values. Both arguments must have identical
2430 types.</p>
2431
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2432<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2433
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2434<p>The value produced is the integer or floating point sum of the two
2435operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2436
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2437<p>If an integer sum has unsigned overflow, the result returned is the
2438mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2439the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2440
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2441<p>Because LLVM integers use a two's complement representation, this
2442instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2443
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2444<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2445
2446<pre>
2447 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2448</pre>
2449</div>
2450<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2451<div class="doc_subsubsection">
2452 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2453</div>
2454
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2455<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2456
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2457<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2458
2459<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2460 &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]2461</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2462
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2463<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2464
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2465<p>The '<tt>sub</tt>' instruction returns the difference of its two
2466operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2467
2468<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2469'<tt>neg</tt>' instruction present in most other intermediate
2470representations.</p>
2471
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2472<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2473
2474<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2475 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2476 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2477 types.</p>
2478
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2479<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2480
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2481<p>The value produced is the integer or floating point difference of
2482the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2483
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2484<p>If an integer difference has unsigned overflow, the result returned is the
2485mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2486the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2487
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2488<p>Because LLVM integers use a two's complement representation, this
2489instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2490
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2491<h5>Example:</h5>
2492<pre>
2493 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2494 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2495</pre>
2496</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2497
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2498<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2499<div class="doc_subsubsection">
2500 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2501</div>
2502
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2503<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2504
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2505<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2506<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]2507</pre>
2508<h5>Overview:</h5>
2509<p>The '<tt>mul</tt>' instruction returns the product of its two
2510operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2511
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2512<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2513
2514<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2515href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2516or <a href="#t_vector">vector</a> values. Both arguments must have identical
2517types.</p>
2518
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2519<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2520
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2521<p>The value produced is the integer or floating point product of the
2522two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2523
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2524<p>If the result of an integer multiplication has unsigned overflow,
2525the result returned is the mathematical result modulo
25262<sup>n</sup>, where n is the bit width of the result.</p>
2527<p>Because LLVM integers use a two's complement representation, and the
2528result is the same width as the operands, this instruction returns the
2529correct result for both signed and unsigned integers. If a full product
2530(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2531should be sign-extended or zero-extended as appropriate to the
2532width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2533<h5>Example:</h5>
2534<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2535</pre>
2536</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2537
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2538<!-- _______________________________________________________________________ -->
2539<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2540</a></div>
2541<div class="doc_text">
2542<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2543<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]2544</pre>
2545<h5>Overview:</h5>
2546<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2547operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2548
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2549<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2550
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2551<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2552<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2553values. Both arguments must have identical types.</p>
2554
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2555<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2556
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2557<p>The value produced is the unsigned integer quotient of the two operands.</p>
2558<p>Note that unsigned integer division and signed integer division are distinct
2559operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2560<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2561<h5>Example:</h5>
2562<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2563</pre>
2564</div>
2565<!-- _______________________________________________________________________ -->
2566<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2567</a> </div>
2568<div class="doc_text">
2569<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2570<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2571 &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]2572</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2573
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2574<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2575
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2576<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2577operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2578
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2579<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2580
2581<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2582<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2583values. Both arguments must have identical types.</p>
2584
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2585<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2586<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]2587<p>Note that signed integer division and unsigned integer division are distinct
2588operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2589<p>Division by zero leads to undefined behavior. Overflow also leads to
2590undefined behavior; this is a rare case, but can occur, for example,
2591by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2592<h5>Example:</h5>
2593<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2594</pre>
2595</div>
2596<!-- _______________________________________________________________________ -->
2597<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2598Instruction</a> </div>
2599<div class="doc_text">
2600<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2601<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2602 &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]2603</pre>
2604<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2605
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2606<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2607operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2608
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2609<h5>Arguments:</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 two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2612<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2613of floating point values. Both arguments must have identical types.</p>
2614
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2615<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2616
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2617<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2618
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2619<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2620
2621<pre>
2622 &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]2623</pre>
2624</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2625
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2626<!-- _______________________________________________________________________ -->
2627<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2628</div>
2629<div class="doc_text">
2630<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2631<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]2632</pre>
2633<h5>Overview:</h5>
2634<p>The '<tt>urem</tt>' instruction returns the remainder from the
2635unsigned division of its two arguments.</p>
2636<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2637<p>The two arguments to the '<tt>urem</tt>' instruction must be
2638<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2639values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2640<h5>Semantics:</h5>
2641<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2642This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2643<p>Note that unsigned integer remainder and signed integer remainder are
2644distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2645<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2646<h5>Example:</h5>
2647<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2648</pre>
2649
2650</div>
2651<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2652<div class="doc_subsubsection">
2653 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2654</div>
2655
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2656<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2657
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2658<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2659
2660<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2661 &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]2662</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2663
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2664<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2665
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2666<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2667signed division of its two operands. This instruction can also take
2668<a href="#t_vector">vector</a> versions of the values in which case
2669the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2670
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2671<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2672
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2673<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2674<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2675values. Both arguments must have identical types.</p>
2676
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2677<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2678
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2679<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2680has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2681operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2682a value. For more information about the difference, see <a
2683 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2684Math Forum</a>. For a table of how this is implemented in various languages,
2685please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2686Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2687<p>Note that signed integer remainder and unsigned integer remainder are
2688distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2689<p>Taking the remainder of a division by zero leads to undefined behavior.
2690Overflow also leads to undefined behavior; this is a rare case, but can occur,
2691for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2692(The remainder doesn't actually overflow, but this rule lets srem be
2693implemented using instructions that return both the result of the division
2694and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2695<h5>Example:</h5>
2696<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2697</pre>
2698
2699</div>
2700<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2701<div class="doc_subsubsection">
2702 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2703
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2704<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2705
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2706<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2707<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]2708</pre>
2709<h5>Overview:</h5>
2710<p>The '<tt>frem</tt>' instruction returns the remainder from the
2711division of its two operands.</p>
2712<h5>Arguments:</h5>
2713<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2714<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2715of floating point values. Both arguments must have identical types.</p>
2716
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2717<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2718
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2719<p>This instruction returns the <i>remainder</i> of a division.
2720The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2721
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2722<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2723
2724<pre>
2725 &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]2726</pre>
2727</div>
2728
2729<!-- ======================================================================= -->
2730<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2731Operations</a> </div>
2732<div class="doc_text">
2733<p>Bitwise binary operators are used to do various forms of
2734bit-twiddling in a program. They are generally very efficient
2735instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2736instructions. They require two operands of the same type, execute an operation on them,
2737and 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]2738</div>
2739
2740<!-- _______________________________________________________________________ -->
2741<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2742Instruction</a> </div>
2743<div class="doc_text">
2744<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2745<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]2746</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2747
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2748<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2749
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2750<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2751the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2752
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2753<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2754
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2755<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2756 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2757type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2758
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2759<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2760
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2761<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2762where 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]2763equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2764If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2765corresponding shift amount in <tt>op2</tt>.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2766
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2767<h5>Example:</h5><pre>
2768 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2769 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2770 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2771 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2772 &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]2773</pre>
2774</div>
2775<!-- _______________________________________________________________________ -->
2776<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</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; = 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]2781</pre>
2782
2783<h5>Overview:</h5>
2784<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2785operand shifted to the right a specified number of bits with zero fill.</p>
2786
2787<h5>Arguments:</h5>
2788<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2789<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]2790type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2791
2792<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2793
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2794<p>This instruction always performs a logical shift right operation. The most
2795significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2796shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
Mon P Wangb0f51822008-12-10 08:55:09 +0000[diff] [blame]2797the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2798vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2799amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2800
2801<h5>Example:</h5>
2802<pre>
2803 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2804 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2805 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2806 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2807 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2808 &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]2809</pre>
2810</div>
2811
2812<!-- _______________________________________________________________________ -->
2813<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2814Instruction</a> </div>
2815<div class="doc_text">
2816
2817<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2818<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]2819</pre>
2820
2821<h5>Overview:</h5>
2822<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2823operand shifted to the right a specified number of bits with sign extension.</p>
2824
2825<h5>Arguments:</h5>
2826<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2827<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]2828type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2829
2830<h5>Semantics:</h5>
2831<p>This instruction always performs an arithmetic shift right operation,
2832The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2833of <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]2834larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2835arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2836corresponding shift amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2837
2838<h5>Example:</h5>
2839<pre>
2840 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2841 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2842 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2843 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2844 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2845 &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]2846</pre>
2847</div>
2848
2849<!-- _______________________________________________________________________ -->
2850<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2851Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2852
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2853<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2854
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2855<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2856
2857<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2858 &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]2859</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2860
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2861<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2862
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2863<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2864its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2865
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2866<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2867
2868<p>The two arguments to the '<tt>and</tt>' instruction must be
2869<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2870values. Both arguments must have identical types.</p>
2871
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2872<h5>Semantics:</h5>
2873<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2874<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2875<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2876<table border="1" cellspacing="0" cellpadding="4">
2877 <tbody>
2878 <tr>
2879 <td>In0</td>
2880 <td>In1</td>
2881 <td>Out</td>
2882 </tr>
2883 <tr>
2884 <td>0</td>
2885 <td>0</td>
2886 <td>0</td>
2887 </tr>
2888 <tr>
2889 <td>0</td>
2890 <td>1</td>
2891 <td>0</td>
2892 </tr>
2893 <tr>
2894 <td>1</td>
2895 <td>0</td>
2896 <td>0</td>
2897 </tr>
2898 <tr>
2899 <td>1</td>
2900 <td>1</td>
2901 <td>1</td>
2902 </tr>
2903 </tbody>
2904</table>
2905</div>
2906<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2907<pre>
2908 &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]2909 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2910 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2911</pre>
2912</div>
2913<!-- _______________________________________________________________________ -->
2914<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2915<div class="doc_text">
2916<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2917<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]2918</pre>
2919<h5>Overview:</h5>
2920<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2921or of its two operands.</p>
2922<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2923
2924<p>The two arguments to the '<tt>or</tt>' instruction must be
2925<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2926values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2927<h5>Semantics:</h5>
2928<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2929<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2930<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2931<table border="1" cellspacing="0" cellpadding="4">
2932 <tbody>
2933 <tr>
2934 <td>In0</td>
2935 <td>In1</td>
2936 <td>Out</td>
2937 </tr>
2938 <tr>
2939 <td>0</td>
2940 <td>0</td>
2941 <td>0</td>
2942 </tr>
2943 <tr>
2944 <td>0</td>
2945 <td>1</td>
2946 <td>1</td>
2947 </tr>
2948 <tr>
2949 <td>1</td>
2950 <td>0</td>
2951 <td>1</td>
2952 </tr>
2953 <tr>
2954 <td>1</td>
2955 <td>1</td>
2956 <td>1</td>
2957 </tr>
2958 </tbody>
2959</table>
2960</div>
2961<h5>Example:</h5>
2962<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2963 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2964 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2965</pre>
2966</div>
2967<!-- _______________________________________________________________________ -->
2968<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2969Instruction</a> </div>
2970<div class="doc_text">
2971<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2972<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]2973</pre>
2974<h5>Overview:</h5>
2975<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2976or of its two operands. The <tt>xor</tt> is used to implement the
2977"one's complement" operation, which is the "~" operator in C.</p>
2978<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2979<p>The two arguments to the '<tt>xor</tt>' instruction must be
2980<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2981values. Both arguments must have identical types.</p>
2982
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2983<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2984
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2985<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2986<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2987<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2988<table border="1" cellspacing="0" cellpadding="4">
2989 <tbody>
2990 <tr>
2991 <td>In0</td>
2992 <td>In1</td>
2993 <td>Out</td>
2994 </tr>
2995 <tr>
2996 <td>0</td>
2997 <td>0</td>
2998 <td>0</td>
2999 </tr>
3000 <tr>
3001 <td>0</td>
3002 <td>1</td>
3003 <td>1</td>
3004 </tr>
3005 <tr>
3006 <td>1</td>
3007 <td>0</td>
3008 <td>1</td>
3009 </tr>
3010 <tr>
3011 <td>1</td>
3012 <td>1</td>
3013 <td>0</td>
3014 </tr>
3015 </tbody>
3016</table>
3017</div>
3018<p> </p>
3019<h5>Example:</h5>
3020<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3021 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3022 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3023 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3024</pre>
3025</div>
3026
3027<!-- ======================================================================= -->
3028<div class="doc_subsection">
3029 <a name="vectorops">Vector Operations</a>
3030</div>
3031
3032<div class="doc_text">
3033
3034<p>LLVM supports several instructions to represent vector operations in a
3035target-independent manner. These instructions cover the element-access and
3036vector-specific operations needed to process vectors effectively. While LLVM
3037does directly support these vector operations, many sophisticated algorithms
3038will want to use target-specific intrinsics to take full advantage of a specific
3039target.</p>
3040
3041</div>
3042
3043<!-- _______________________________________________________________________ -->
3044<div class="doc_subsubsection">
3045 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3046</div>
3047
3048<div class="doc_text">
3049
3050<h5>Syntax:</h5>
3051
3052<pre>
3053 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
3054</pre>
3055
3056<h5>Overview:</h5>
3057
3058<p>
3059The '<tt>extractelement</tt>' instruction extracts a single scalar
3060element from a vector at a specified index.
3061</p>
3062
3063
3064<h5>Arguments:</h5>
3065
3066<p>
3067The first operand of an '<tt>extractelement</tt>' instruction is a
3068value of <a href="#t_vector">vector</a> type. The second operand is
3069an index indicating the position from which to extract the element.
3070The index may be a variable.</p>
3071
3072<h5>Semantics:</h5>
3073
3074<p>
3075The result is a scalar of the same type as the element type of
3076<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3077<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3078results are undefined.
3079</p>
3080
3081<h5>Example:</h5>
3082
3083<pre>
3084 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
3085</pre>
3086</div>
3087
3088
3089<!-- _______________________________________________________________________ -->
3090<div class="doc_subsubsection">
3091 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3092</div>
3093
3094<div class="doc_text">
3095
3096<h5>Syntax:</h5>
3097
3098<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]3099 &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]3100</pre>
3101
3102<h5>Overview:</h5>
3103
3104<p>
3105The '<tt>insertelement</tt>' instruction inserts a scalar
3106element into a vector at a specified index.
3107</p>
3108
3109
3110<h5>Arguments:</h5>
3111
3112<p>
3113The first operand of an '<tt>insertelement</tt>' instruction is a
3114value of <a href="#t_vector">vector</a> type. The second operand is a
3115scalar value whose type must equal the element type of the first
3116operand. The third operand is an index indicating the position at
3117which to insert the value. The index may be a variable.</p>
3118
3119<h5>Semantics:</h5>
3120
3121<p>
3122The result is a vector of the same type as <tt>val</tt>. Its
3123element values are those of <tt>val</tt> except at position
3124<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3125exceeds the length of <tt>val</tt>, the results are undefined.
3126</p>
3127
3128<h5>Example:</h5>
3129
3130<pre>
3131 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
3132</pre>
3133</div>
3134
3135<!-- _______________________________________________________________________ -->
3136<div class="doc_subsubsection">
3137 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3138</div>
3139
3140<div class="doc_text">
3141
3142<h5>Syntax:</h5>
3143
3144<pre>
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3145 &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]3146</pre>
3147
3148<h5>Overview:</h5>
3149
3150<p>
3151The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3152from two input vectors, returning a vector with the same element type as
3153the input and length that is the same as the shuffle mask.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3154</p>
3155
3156<h5>Arguments:</h5>
3157
3158<p>
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3159The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3160with types that match each other. The third argument is a shuffle mask whose
3161element type is always 'i32'. The result of the instruction is a vector whose
3162length is the same as the shuffle mask and whose element type is the same as
3163the element type of the first two operands.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3164</p>
3165
3166<p>
3167The shuffle mask operand is required to be a constant vector with either
3168constant integer or undef values.
3169</p>
3170
3171<h5>Semantics:</h5>
3172
3173<p>
3174The elements of the two input vectors are numbered from left to right across
3175both of the vectors. The shuffle mask operand specifies, for each element of
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3176the result vector, which element of the two input vectors the result element
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3177gets. The element selector may be undef (meaning "don't care") and the second
3178operand may be undef if performing a shuffle from only one vector.
3179</p>
3180
3181<h5>Example:</h5>
3182
3183<pre>
3184 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3185 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
3186 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
3187 &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]3188 %result = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
3189 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i>
3190 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3191 &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]3192</pre>
3193</div>
3194
3195
3196<!-- ======================================================================= -->
3197<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3198 <a name="aggregateops">Aggregate Operations</a>
3199</div>
3200
3201<div class="doc_text">
3202
3203<p>LLVM supports several instructions for working with aggregate values.
3204</p>
3205
3206</div>
3207
3208<!-- _______________________________________________________________________ -->
3209<div class="doc_subsubsection">
3210 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3211</div>
3212
3213<div class="doc_text">
3214
3215<h5>Syntax:</h5>
3216
3217<pre>
3218 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
3219</pre>
3220
3221<h5>Overview:</h5>
3222
3223<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3224The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3225or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3226</p>
3227
3228
3229<h5>Arguments:</h5>
3230
3231<p>
3232The first operand of an '<tt>extractvalue</tt>' instruction is a
3233value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3234type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3235in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3236'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3237</p>
3238
3239<h5>Semantics:</h5>
3240
3241<p>
3242The result is the value at the position in the aggregate specified by
3243the index operands.
3244</p>
3245
3246<h5>Example:</h5>
3247
3248<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3249 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3250</pre>
3251</div>
3252
3253
3254<!-- _______________________________________________________________________ -->
3255<div class="doc_subsubsection">
3256 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3257</div>
3258
3259<div class="doc_text">
3260
3261<h5>Syntax:</h5>
3262
3263<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3264 &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]3265</pre>
3266
3267<h5>Overview:</h5>
3268
3269<p>
3270The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3271into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3272</p>
3273
3274
3275<h5>Arguments:</h5>
3276
3277<p>
3278The first operand of an '<tt>insertvalue</tt>' instruction is a
3279value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3280The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3281The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3282indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3283indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3284'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3285The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3286by the indices.
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3287</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3288
3289<h5>Semantics:</h5>
3290
3291<p>
3292The result is an aggregate of the same type as <tt>val</tt>. Its
3293value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3294specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3295</p>
3296
3297<h5>Example:</h5>
3298
3299<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3300 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3301</pre>
3302</div>
3303
3304
3305<!-- ======================================================================= -->
3306<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3307 <a name="memoryops">Memory Access and Addressing Operations</a>
3308</div>
3309
3310<div class="doc_text">
3311
3312<p>A key design point of an SSA-based representation is how it
3313represents memory. In LLVM, no memory locations are in SSA form, which
3314makes things very simple. This section describes how to read, write,
3315allocate, and free memory in LLVM.</p>
3316
3317</div>
3318
3319<!-- _______________________________________________________________________ -->
3320<div class="doc_subsubsection">
3321 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3322</div>
3323
3324<div class="doc_text">
3325
3326<h5>Syntax:</h5>
3327
3328<pre>
3329 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3330</pre>
3331
3332<h5>Overview:</h5>
3333
3334<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3335heap and returns a pointer to it. The object is always allocated in the generic
3336address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3337
3338<h5>Arguments:</h5>
3339
3340<p>The '<tt>malloc</tt>' instruction allocates
3341<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3342bytes of memory from the operating system and returns a pointer of the
3343appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3344number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3345If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3346be aligned to at least that boundary. If not specified, or if zero, the target can
3347choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3348
3349<p>'<tt>type</tt>' must be a sized type.</p>
3350
3351<h5>Semantics:</h5>
3352
3353<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Nick Lewyckyff384472008-11-24 03:41:24 +0000[diff] [blame]3354a pointer is returned. The result of a zero byte allocation is undefined. The
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3355result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3356
3357<h5>Example:</h5>
3358
3359<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3360 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3361
3362 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3363 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3364 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3365 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3366 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3367</pre>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]3368
3369<p>Note that the code generator does not yet respect the
3370 alignment value.</p>
3371
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3372</div>
3373
3374<!-- _______________________________________________________________________ -->
3375<div class="doc_subsubsection">
3376 <a name="i_free">'<tt>free</tt>' Instruction</a>
3377</div>
3378
3379<div class="doc_text">
3380
3381<h5>Syntax:</h5>
3382
3383<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3384 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3385</pre>
3386
3387<h5>Overview:</h5>
3388
3389<p>The '<tt>free</tt>' instruction returns memory back to the unused
3390memory heap to be reallocated in the future.</p>
3391
3392<h5>Arguments:</h5>
3393
3394<p>'<tt>value</tt>' shall be a pointer value that points to a value
3395that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3396instruction.</p>
3397
3398<h5>Semantics:</h5>
3399
3400<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3401after this instruction executes. If the pointer is null, the operation
3402is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3403
3404<h5>Example:</h5>
3405
3406<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3407 %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]3408 free [4 x i8]* %array
3409</pre>
3410</div>
3411
3412<!-- _______________________________________________________________________ -->
3413<div class="doc_subsubsection">
3414 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3415</div>
3416
3417<div class="doc_text">
3418
3419<h5>Syntax:</h5>
3420
3421<pre>
3422 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3423</pre>
3424
3425<h5>Overview:</h5>
3426
3427<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3428currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3429returns to its caller. The object is always allocated in the generic address
3430space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3431
3432<h5>Arguments:</h5>
3433
3434<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3435bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3436appropriate type to the program. If "NumElements" is specified, it is the
3437number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3438If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3439to be aligned to at least that boundary. If not specified, or if zero, the target
3440can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3441
3442<p>'<tt>type</tt>' may be any sized type.</p>
3443
3444<h5>Semantics:</h5>
3445
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3446<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3447there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3448memory is automatically released when the function returns. The '<tt>alloca</tt>'
3449instruction is commonly used to represent automatic variables that must
3450have an address available. When the function returns (either with the <tt><a
3451 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]3452instructions), the memory is reclaimed. Allocating zero bytes
3453is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3454
3455<h5>Example:</h5>
3456
3457<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3458 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3459 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3460 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3461 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3462</pre>
3463</div>
3464
3465<!-- _______________________________________________________________________ -->
3466<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3467Instruction</a> </div>
3468<div class="doc_text">
3469<h5>Syntax:</h5>
3470<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>
3471<h5>Overview:</h5>
3472<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3473<h5>Arguments:</h5>
3474<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3475address from which to load. The pointer must point to a <a
3476 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3477marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3478the number or order of execution of this <tt>load</tt> with other
3479volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3480instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3481<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3482The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3483(that is, the alignment of the memory address). A value of 0 or an
3484omitted "align" argument means that the operation has the preferential
3485alignment for the target. It is the responsibility of the code emitter
3486to ensure that the alignment information is correct. Overestimating
3487the alignment results in an undefined behavior. Underestimating the
3488alignment may produce less efficient code. An alignment of 1 is always
3489safe.
3490</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3491<h5>Semantics:</h5>
3492<p>The location of memory pointed to is loaded.</p>
3493<h5>Examples:</h5>
3494<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3495 <a
3496 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3497 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3498</pre>
3499</div>
3500<!-- _______________________________________________________________________ -->
3501<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3502Instruction</a> </div>
3503<div class="doc_text">
3504<h5>Syntax:</h5>
3505<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3506 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3507</pre>
3508<h5>Overview:</h5>
3509<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3510<h5>Arguments:</h5>
3511<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3512to 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]3513operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3514of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3515operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3516optimizer is not allowed to modify the number or order of execution of
3517this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3518 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3519<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3520The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3521(that is, the alignment of the memory address). A value of 0 or an
3522omitted "align" argument means that the operation has the preferential
3523alignment for the target. It is the responsibility of the code emitter
3524to ensure that the alignment information is correct. Overestimating
3525the alignment results in an undefined behavior. Underestimating the
3526alignment may produce less efficient code. An alignment of 1 is always
3527safe.
3528</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3529<h5>Semantics:</h5>
3530<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3531at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3532<h5>Example:</h5>
3533<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3534 store i32 3, i32* %ptr <i>; yields {void}</i>
3535 %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]3536</pre>
3537</div>
3538
3539<!-- _______________________________________________________________________ -->
3540<div class="doc_subsubsection">
3541 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3542</div>
3543
3544<div class="doc_text">
3545<h5>Syntax:</h5>
3546<pre>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3547 &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]3548</pre>
3549
3550<h5>Overview:</h5>
3551
3552<p>
3553The '<tt>getelementptr</tt>' instruction is used to get the address of a
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3554subelement of an aggregate data structure. It performs address calculation only
3555and does not access memory.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3556
3557<h5>Arguments:</h5>
3558
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3559<p>The first argument is always a pointer, and forms the basis of the
3560calculation. The remaining arguments are indices, that indicate which of the
3561elements of the aggregate object are indexed. The interpretation of each index
3562is dependent on the type being indexed into. The first index always indexes the
3563pointer value given as the first argument, the second index indexes a value of
3564the type pointed to (not necessarily the value directly pointed to, since the
3565first index can be non-zero), etc. The first type indexed into must be a pointer
3566value, subsequent types can be arrays, vectors and structs. Note that subsequent
3567types being indexed into can never be pointers, since that would require loading
3568the pointer before continuing calculation.</p>
3569
3570<p>The type of each index argument depends on the type it is indexing into.
3571When indexing into a (packed) structure, only <tt>i32</tt> integer
3572<b>constants</b> are allowed. When indexing into an array, pointer or vector,
3573only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3574will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3575
3576<p>For example, let's consider a C code fragment and how it gets
3577compiled to LLVM:</p>
3578
3579<div class="doc_code">
3580<pre>
3581struct RT {
3582 char A;
3583 int B[10][20];
3584 char C;
3585};
3586struct ST {
3587 int X;
3588 double Y;
3589 struct RT Z;
3590};
3591
3592int *foo(struct ST *s) {
3593 return &amp;s[1].Z.B[5][13];
3594}
3595</pre>
3596</div>
3597
3598<p>The LLVM code generated by the GCC frontend is:</p>
3599
3600<div class="doc_code">
3601<pre>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]3602%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3603%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3604
3605define i32* %foo(%ST* %s) {
3606entry:
3607 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3608 ret i32* %reg
3609}
3610</pre>
3611</div>
3612
3613<h5>Semantics:</h5>
3614
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3615<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3616type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3617}</tt>' type, a structure. The second index indexes into the third element of
3618the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3619i8 }</tt>' type, another structure. The third index indexes into the second
3620element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3621array. The two dimensions of the array are subscripted into, yielding an
3622'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3623to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3624
3625<p>Note that it is perfectly legal to index partially through a
3626structure, returning a pointer to an inner element. Because of this,
3627the LLVM code for the given testcase is equivalent to:</p>
3628
3629<pre>
3630 define i32* %foo(%ST* %s) {
3631 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3632 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3633 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3634 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3635 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3636 ret i32* %t5
3637 }
3638</pre>
3639
Chris Lattner50609942009-03-09 20:55:18 +0000[diff] [blame^]3640<p>Note that it is undefined to access an array out of bounds: array
3641and pointer indexes must always be within the defined bounds of the
3642array type when accessed with an instruction that dereferences the
3643pointer (e.g. a load or store instruction). The one exception for
3644this rule is zero length arrays. These arrays are defined to be
3645accessible as variable length arrays, which requires access beyond the
3646zero'th element.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3647
3648<p>The getelementptr instruction is often confusing. For some more insight
3649into how it works, see <a href="GetElementPtr.html">the getelementptr
3650FAQ</a>.</p>
3651
3652<h5>Example:</h5>
3653
3654<pre>
3655 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3656 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3657 <i>; yields i8*:vptr</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3658 %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]3659 <i>; yields i8*:eptr</i>
3660 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3661</pre>
3662</div>
3663
3664<!-- ======================================================================= -->
3665<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3666</div>
3667<div class="doc_text">
3668<p>The instructions in this category are the conversion instructions (casting)
3669which all take a single operand and a type. They perform various bit conversions
3670on the operand.</p>
3671</div>
3672
3673<!-- _______________________________________________________________________ -->
3674<div class="doc_subsubsection">
3675 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3676</div>
3677<div class="doc_text">
3678
3679<h5>Syntax:</h5>
3680<pre>
3681 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3682</pre>
3683
3684<h5>Overview:</h5>
3685<p>
3686The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3687</p>
3688
3689<h5>Arguments:</h5>
3690<p>
3691The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3692be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3693and type of the result, which must be an <a href="#t_integer">integer</a>
3694type. The bit size of <tt>value</tt> must be larger than the bit size of
3695<tt>ty2</tt>. Equal sized types are not allowed.</p>
3696
3697<h5>Semantics:</h5>
3698<p>
3699The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3700and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3701larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3702It will always truncate bits.</p>
3703
3704<h5>Example:</h5>
3705<pre>
3706 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3707 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3708 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3709</pre>
3710</div>
3711
3712<!-- _______________________________________________________________________ -->
3713<div class="doc_subsubsection">
3714 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3715</div>
3716<div class="doc_text">
3717
3718<h5>Syntax:</h5>
3719<pre>
3720 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3721</pre>
3722
3723<h5>Overview:</h5>
3724<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3725<tt>ty2</tt>.</p>
3726
3727
3728<h5>Arguments:</h5>
3729<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3730<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3731also be of <a href="#t_integer">integer</a> type. The bit size of the
3732<tt>value</tt> must be smaller than the bit size of the destination type,
3733<tt>ty2</tt>.</p>
3734
3735<h5>Semantics:</h5>
3736<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3737bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3738
3739<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3740
3741<h5>Example:</h5>
3742<pre>
3743 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3744 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3745</pre>
3746</div>
3747
3748<!-- _______________________________________________________________________ -->
3749<div class="doc_subsubsection">
3750 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3751</div>
3752<div class="doc_text">
3753
3754<h5>Syntax:</h5>
3755<pre>
3756 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3757</pre>
3758
3759<h5>Overview:</h5>
3760<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3761
3762<h5>Arguments:</h5>
3763<p>
3764The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3765<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3766also be of <a href="#t_integer">integer</a> type. The bit size of the
3767<tt>value</tt> must be smaller than the bit size of the destination type,
3768<tt>ty2</tt>.</p>
3769
3770<h5>Semantics:</h5>
3771<p>
3772The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3773bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3774the type <tt>ty2</tt>.</p>
3775
3776<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3777
3778<h5>Example:</h5>
3779<pre>
3780 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3781 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3782</pre>
3783</div>
3784
3785<!-- _______________________________________________________________________ -->
3786<div class="doc_subsubsection">
3787 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3788</div>
3789
3790<div class="doc_text">
3791
3792<h5>Syntax:</h5>
3793
3794<pre>
3795 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3796</pre>
3797
3798<h5>Overview:</h5>
3799<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3800<tt>ty2</tt>.</p>
3801
3802
3803<h5>Arguments:</h5>
3804<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3805 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3806cast it to. The size of <tt>value</tt> must be larger than the size of
3807<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3808<i>no-op cast</i>.</p>
3809
3810<h5>Semantics:</h5>
3811<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3812<a href="#t_floating">floating point</a> type to a smaller
3813<a href="#t_floating">floating point</a> type. If the value cannot fit within
3814the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3815
3816<h5>Example:</h5>
3817<pre>
3818 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3819 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3820</pre>
3821</div>
3822
3823<!-- _______________________________________________________________________ -->
3824<div class="doc_subsubsection">
3825 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3826</div>
3827<div class="doc_text">
3828
3829<h5>Syntax:</h5>
3830<pre>
3831 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3832</pre>
3833
3834<h5>Overview:</h5>
3835<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3836floating point value.</p>
3837
3838<h5>Arguments:</h5>
3839<p>The '<tt>fpext</tt>' instruction takes a
3840<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3841and a <a href="#t_floating">floating point</a> type to cast it to. The source
3842type must be smaller than the destination type.</p>
3843
3844<h5>Semantics:</h5>
3845<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3846<a href="#t_floating">floating point</a> type to a larger
3847<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3848used to make a <i>no-op cast</i> because it always changes bits. Use
3849<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3850
3851<h5>Example:</h5>
3852<pre>
3853 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3854 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3855</pre>
3856</div>
3857
3858<!-- _______________________________________________________________________ -->
3859<div class="doc_subsubsection">
3860 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3861</div>
3862<div class="doc_text">
3863
3864<h5>Syntax:</h5>
3865<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3866 &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]3867</pre>
3868
3869<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3870<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3871unsigned integer equivalent of type <tt>ty2</tt>.
3872</p>
3873
3874<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3875<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]3876scalar or vector <a href="#t_floating">floating point</a> value, and a type
3877to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3878type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3879vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3880
3881<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3882<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3883<a href="#t_floating">floating point</a> operand into the nearest (rounding
3884towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3885the results are undefined.</p>
3886
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3887<h5>Example:</h5>
3888<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3889 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3890 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3891 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3892</pre>
3893</div>
3894
3895<!-- _______________________________________________________________________ -->
3896<div class="doc_subsubsection">
3897 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3898</div>
3899<div class="doc_text">
3900
3901<h5>Syntax:</h5>
3902<pre>
3903 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3904</pre>
3905
3906<h5>Overview:</h5>
3907<p>The '<tt>fptosi</tt>' instruction converts
3908<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3909</p>
3910
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3911<h5>Arguments:</h5>
3912<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]3913scalar or vector <a href="#t_floating">floating point</a> value, and a type
3914to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3915type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3916vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3917
3918<h5>Semantics:</h5>
3919<p>The '<tt>fptosi</tt>' instruction converts its
3920<a href="#t_floating">floating point</a> operand into the nearest (rounding
3921towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3922the results are undefined.</p>
3923
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3924<h5>Example:</h5>
3925<pre>
3926 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3927 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3928 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3929</pre>
3930</div>
3931
3932<!-- _______________________________________________________________________ -->
3933<div class="doc_subsubsection">
3934 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3935</div>
3936<div class="doc_text">
3937
3938<h5>Syntax:</h5>
3939<pre>
3940 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3941</pre>
3942
3943<h5>Overview:</h5>
3944<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3945integer and converts that value to the <tt>ty2</tt> type.</p>
3946
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3947<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3948<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3949scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3950to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3951type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3952floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3953
3954<h5>Semantics:</h5>
3955<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3956integer quantity and converts it to the corresponding floating point value. If
3957the value cannot fit in the floating point value, the results are undefined.</p>
3958
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3959<h5>Example:</h5>
3960<pre>
3961 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3962 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3963</pre>
3964</div>
3965
3966<!-- _______________________________________________________________________ -->
3967<div class="doc_subsubsection">
3968 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3969</div>
3970<div class="doc_text">
3971
3972<h5>Syntax:</h5>
3973<pre>
3974 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3975</pre>
3976
3977<h5>Overview:</h5>
3978<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3979integer and converts that value to the <tt>ty2</tt> type.</p>
3980
3981<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3982<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3983scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3984to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3985type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3986floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3987
3988<h5>Semantics:</h5>
3989<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3990integer quantity and converts it to the corresponding floating point value. If
3991the value cannot fit in the floating point value, the results are undefined.</p>
3992
3993<h5>Example:</h5>
3994<pre>
3995 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3996 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3997</pre>
3998</div>
3999
4000<!-- _______________________________________________________________________ -->
4001<div class="doc_subsubsection">
4002 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4003</div>
4004<div class="doc_text">
4005
4006<h5>Syntax:</h5>
4007<pre>
4008 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4009</pre>
4010
4011<h5>Overview:</h5>
4012<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4013the integer type <tt>ty2</tt>.</p>
4014
4015<h5>Arguments:</h5>
4016<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4017must 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]4018<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]4019
4020<h5>Semantics:</h5>
4021<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4022<tt>ty2</tt> by interpreting the pointer value as an integer and either
4023truncating or zero extending that value to the size of the integer type. If
4024<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4025<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4026are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4027change.</p>
4028
4029<h5>Example:</h5>
4030<pre>
4031 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4032 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4033</pre>
4034</div>
4035
4036<!-- _______________________________________________________________________ -->
4037<div class="doc_subsubsection">
4038 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4039</div>
4040<div class="doc_text">
4041
4042<h5>Syntax:</h5>
4043<pre>
4044 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4045</pre>
4046
4047<h5>Overview:</h5>
4048<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4049a pointer type, <tt>ty2</tt>.</p>
4050
4051<h5>Arguments:</h5>
4052<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4053value to cast, and a type to cast it to, which must be a
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4054<a href="#t_pointer">pointer</a> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4055
4056<h5>Semantics:</h5>
4057<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4058<tt>ty2</tt> by applying either a zero extension or a truncation depending on
4059the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4060size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4061the size of a pointer then a zero extension is done. If they are the same size,
4062nothing is done (<i>no-op cast</i>).</p>
4063
4064<h5>Example:</h5>
4065<pre>
4066 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4067 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4068 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4069</pre>
4070</div>
4071
4072<!-- _______________________________________________________________________ -->
4073<div class="doc_subsubsection">
4074 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4075</div>
4076<div class="doc_text">
4077
4078<h5>Syntax:</h5>
4079<pre>
4080 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4081</pre>
4082
4083<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4084
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4085<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4086<tt>ty2</tt> without changing any bits.</p>
4087
4088<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4089
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4090<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]4091a non-aggregate first class value, and a type to cast it to, which must also be
4092a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4093<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4094and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4095type is a pointer, the destination type must also be a pointer. This
4096instruction supports bitwise conversion of vectors to integers and to vectors
4097of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4098
4099<h5>Semantics:</h5>
4100<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4101<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4102this conversion. The conversion is done as if the <tt>value</tt> had been
4103stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4104converted to other pointer types with this instruction. To convert pointers to
4105other types, use the <a href="#i_inttoptr">inttoptr</a> or
4106<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4107
4108<h5>Example:</h5>
4109<pre>
4110 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4111 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4112 %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]4113</pre>
4114</div>
4115
4116<!-- ======================================================================= -->
4117<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4118<div class="doc_text">
4119<p>The instructions in this category are the "miscellaneous"
4120instructions, which defy better classification.</p>
4121</div>
4122
4123<!-- _______________________________________________________________________ -->
4124<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4125</div>
4126<div class="doc_text">
4127<h5>Syntax:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4128<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]4129</pre>
4130<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4131<p>The '<tt>icmp</tt>' instruction returns a boolean value or
4132a vector of boolean values based on comparison
4133of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4134<h5>Arguments:</h5>
4135<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4136the condition code indicating the kind of comparison to perform. It is not
4137a value, just a keyword. The possible condition code are:
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4138</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4139<ol>
4140 <li><tt>eq</tt>: equal</li>
4141 <li><tt>ne</tt>: not equal </li>
4142 <li><tt>ugt</tt>: unsigned greater than</li>
4143 <li><tt>uge</tt>: unsigned greater or equal</li>
4144 <li><tt>ult</tt>: unsigned less than</li>
4145 <li><tt>ule</tt>: unsigned less or equal</li>
4146 <li><tt>sgt</tt>: signed greater than</li>
4147 <li><tt>sge</tt>: signed greater or equal</li>
4148 <li><tt>slt</tt>: signed less than</li>
4149 <li><tt>sle</tt>: signed less or equal</li>
4150</ol>
4151<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4152<a href="#t_pointer">pointer</a>
4153or integer <a href="#t_vector">vector</a> typed.
4154They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4155<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4156<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]4157the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4158yields 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]4159</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4160<ol>
4161 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4162 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4163 </li>
4164 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4165 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4166 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4167 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4168 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4169 <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]4170 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4171 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4172 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4173 <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]4174 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4175 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4176 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4177 <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]4178 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4179 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4180 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4181 <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]4182</ol>
4183<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4184values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4185<p>If the operands are integer vectors, then they are compared
4186element by element. The result is an <tt>i1</tt> vector with
4187the same number of elements as the values being compared.
4188Otherwise, the result is an <tt>i1</tt>.
4189</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4190
4191<h5>Example:</h5>
4192<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
4193 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
4194 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
4195 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4196 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4197 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
4198</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4199
4200<p>Note that the code generator does not yet support vector types with
4201 the <tt>icmp</tt> instruction.</p>
4202
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4203</div>
4204
4205<!-- _______________________________________________________________________ -->
4206<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4207</div>
4208<div class="doc_text">
4209<h5>Syntax:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4210<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]4211</pre>
4212<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4213<p>The '<tt>fcmp</tt>' instruction returns a boolean value
4214or vector of boolean values based on comparison
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4215of its operands.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4216<p>
4217If the operands are floating point scalars, then the result
4218type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4219</p>
4220<p>If the operands are floating point vectors, then the result type
4221is a vector of boolean with the same number of elements as the
4222operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4223<h5>Arguments:</h5>
4224<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4225the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4226a value, just a keyword. The possible condition code are:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4227<ol>
4228 <li><tt>false</tt>: no comparison, always returns false</li>
4229 <li><tt>oeq</tt>: ordered and equal</li>
4230 <li><tt>ogt</tt>: ordered and greater than </li>
4231 <li><tt>oge</tt>: ordered and greater than or equal</li>
4232 <li><tt>olt</tt>: ordered and less than </li>
4233 <li><tt>ole</tt>: ordered and less than or equal</li>
4234 <li><tt>one</tt>: ordered and not equal</li>
4235 <li><tt>ord</tt>: ordered (no nans)</li>
4236 <li><tt>ueq</tt>: unordered or equal</li>
4237 <li><tt>ugt</tt>: unordered or greater than </li>
4238 <li><tt>uge</tt>: unordered or greater than or equal</li>
4239 <li><tt>ult</tt>: unordered or less than </li>
4240 <li><tt>ule</tt>: unordered or less than or equal</li>
4241 <li><tt>une</tt>: unordered or not equal</li>
4242 <li><tt>uno</tt>: unordered (either nans)</li>
4243 <li><tt>true</tt>: no comparison, always returns true</li>
4244</ol>
4245<p><i>Ordered</i> means that neither operand is a QNAN while
4246<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4247<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4248either a <a href="#t_floating">floating point</a> type
4249or a <a href="#t_vector">vector</a> of floating point type.
4250They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4251<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4252<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4253according to the condition code given as <tt>cond</tt>.
4254If the operands are vectors, then the vectors are compared
4255element by element.
4256Each comparison performed
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4257always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4258<ol>
4259 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4260 <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]4261 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4262 <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]4263 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4264 <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]4265 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4266 <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]4267 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4268 <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]4269 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4270 <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]4271 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4272 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4273 <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]4274 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4275 <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]4276 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4277 <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]4278 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4279 <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]4280 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4281 <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]4282 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4283 <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]4284 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4285 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4286 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4287</ol>
4288
4289<h5>Example:</h5>
4290<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]4291 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4292 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4293 &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]4294</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4295
4296<p>Note that the code generator does not yet support vector types with
4297 the <tt>fcmp</tt> instruction.</p>
4298
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4299</div>
4300
4301<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4302<div class="doc_subsubsection">
4303 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4304</div>
4305<div class="doc_text">
4306<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4307<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]4308</pre>
4309<h5>Overview:</h5>
4310<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4311element-wise comparison of its two integer vector operands.</p>
4312<h5>Arguments:</h5>
4313<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4314the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4315a value, just a keyword. The possible condition code are:</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4316<ol>
4317 <li><tt>eq</tt>: equal</li>
4318 <li><tt>ne</tt>: not equal </li>
4319 <li><tt>ugt</tt>: unsigned greater than</li>
4320 <li><tt>uge</tt>: unsigned greater or equal</li>
4321 <li><tt>ult</tt>: unsigned less than</li>
4322 <li><tt>ule</tt>: unsigned less or equal</li>
4323 <li><tt>sgt</tt>: signed greater than</li>
4324 <li><tt>sge</tt>: signed greater or equal</li>
4325 <li><tt>slt</tt>: signed less than</li>
4326 <li><tt>sle</tt>: signed less or equal</li>
4327</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4328<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4329<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4330<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4331<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4332according to the condition code given as <tt>cond</tt>. The comparison yields a
4333<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4334identical type as the values being compared. The most significant bit in each
4335element is 1 if the element-wise comparison evaluates to true, and is 0
4336otherwise. All other bits of the result are undefined. The condition codes
4337are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4338instruction</a>.</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4339
4340<h5>Example:</h5>
4341<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4342 &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>
4343 &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]4344</pre>
4345</div>
4346
4347<!-- _______________________________________________________________________ -->
4348<div class="doc_subsubsection">
4349 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4350</div>
4351<div class="doc_text">
4352<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4353<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]4354<h5>Overview:</h5>
4355<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4356element-wise comparison of its two floating point vector operands. The output
4357elements have the same width as the input elements.</p>
4358<h5>Arguments:</h5>
4359<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4360the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4361a value, just a keyword. The possible condition code are:</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4362<ol>
4363 <li><tt>false</tt>: no comparison, always returns false</li>
4364 <li><tt>oeq</tt>: ordered and equal</li>
4365 <li><tt>ogt</tt>: ordered and greater than </li>
4366 <li><tt>oge</tt>: ordered and greater than or equal</li>
4367 <li><tt>olt</tt>: ordered and less than </li>
4368 <li><tt>ole</tt>: ordered and less than or equal</li>
4369 <li><tt>one</tt>: ordered and not equal</li>
4370 <li><tt>ord</tt>: ordered (no nans)</li>
4371 <li><tt>ueq</tt>: unordered or equal</li>
4372 <li><tt>ugt</tt>: unordered or greater than </li>
4373 <li><tt>uge</tt>: unordered or greater than or equal</li>
4374 <li><tt>ult</tt>: unordered or less than </li>
4375 <li><tt>ule</tt>: unordered or less than or equal</li>
4376 <li><tt>une</tt>: unordered or not equal</li>
4377 <li><tt>uno</tt>: unordered (either nans)</li>
4378 <li><tt>true</tt>: no comparison, always returns true</li>
4379</ol>
4380<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4381<a href="#t_floating">floating point</a> typed. They must also be identical
4382types.</p>
4383<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4384<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4385according to the condition code given as <tt>cond</tt>. The comparison yields a
4386<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4387an identical number of elements as the values being compared, and each element
4388having identical with to the width of the floating point elements. The most
4389significant bit in each element is 1 if the element-wise comparison evaluates to
4390true, and is 0 otherwise. All other bits of the result are undefined. The
4391condition codes are evaluated identically to the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4392<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4393
4394<h5>Example:</h5>
4395<pre>
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000[diff] [blame]4396 <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4397 &lt;result&gt; = vfcmp oeq &lt;2 x float&gt; &lt; float 4, float 0 &gt;, &lt; float 5, float 0 &gt;
4398
4399 <i>; yields: result=&lt;2 x i64&gt; &lt; i64 -1, i64 0 &gt;</i>
4400 &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]4401</pre>
4402</div>
4403
4404<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4405<div class="doc_subsubsection">
4406 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4407</div>
4408
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4409<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4410
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4411<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4412
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4413<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4414<h5>Overview:</h5>
4415<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4416the SSA graph representing the function.</p>
4417<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4418
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4419<p>The type of the incoming values is specified with the first type
4420field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4421as arguments, with one pair for each predecessor basic block of the
4422current block. Only values of <a href="#t_firstclass">first class</a>
4423type may be used as the value arguments to the PHI node. Only labels
4424may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4425
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4426<p>There must be no non-phi instructions between the start of a basic
4427block and the PHI instructions: i.e. PHI instructions must be first in
4428a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4429
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4430<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4431
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4432<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4433specified by the pair corresponding to the predecessor basic block that executed
4434just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4435
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4436<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4437<pre>
4438Loop: ; Infinite loop that counts from 0 on up...
4439 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4440 %nextindvar = add i32 %indvar, 1
4441 br label %Loop
4442</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4443</div>
4444
4445<!-- _______________________________________________________________________ -->
4446<div class="doc_subsubsection">
4447 <a name="i_select">'<tt>select</tt>' Instruction</a>
4448</div>
4449
4450<div class="doc_text">
4451
4452<h5>Syntax:</h5>
4453
4454<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4455 &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>
4456
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4457 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4458</pre>
4459
4460<h5>Overview:</h5>
4461
4462<p>
4463The '<tt>select</tt>' instruction is used to choose one value based on a
4464condition, without branching.
4465</p>
4466
4467
4468<h5>Arguments:</h5>
4469
4470<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4471The '<tt>select</tt>' instruction requires an 'i1' value or
4472a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4473condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4474type. If the val1/val2 are vectors and
4475the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4476individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4477</p>
4478
4479<h5>Semantics:</h5>
4480
4481<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4482If 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]4483value argument; otherwise, it returns the second value argument.
4484</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4485<p>
4486If the condition is a vector of i1, then the value arguments must
4487be vectors of the same size, and the selection is done element
4488by element.
4489</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4490
4491<h5>Example:</h5>
4492
4493<pre>
4494 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4495</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4496
4497<p>Note that the code generator does not yet support conditions
4498 with vector type.</p>
4499
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4500</div>
4501
4502
4503<!-- _______________________________________________________________________ -->
4504<div class="doc_subsubsection">
4505 <a name="i_call">'<tt>call</tt>' Instruction</a>
4506</div>
4507
4508<div class="doc_text">
4509
4510<h5>Syntax:</h5>
4511<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4512 &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]4513</pre>
4514
4515<h5>Overview:</h5>
4516
4517<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4518
4519<h5>Arguments:</h5>
4520
4521<p>This instruction requires several arguments:</p>
4522
4523<ol>
4524 <li>
4525 <p>The optional "tail" marker indicates whether the callee function accesses
4526 any allocas or varargs in the caller. If the "tail" marker is present, the
4527 function call is eligible for tail call optimization. Note that calls may
4528 be marked "tail" even if they do not occur before a <a
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4529 href="#i_ret"><tt>ret</tt></a> instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4530 </li>
4531 <li>
4532 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4533 convention</a> the call should use. If none is specified, the call defaults
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4534 to using C calling conventions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4535 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4536
4537 <li>
4538 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4539 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4540 and '<tt>inreg</tt>' attributes are valid here.</p>
4541 </li>
4542
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4543 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4544 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4545 the type of the return value. Functions that return no value are marked
4546 <tt><a href="#t_void">void</a></tt>.</p>
4547 </li>
4548 <li>
4549 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4550 value being invoked. The argument types must match the types implied by
4551 this signature. This type can be omitted if the function is not varargs
4552 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4553 </li>
4554 <li>
4555 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4556 be invoked. In most cases, this is a direct function invocation, but
4557 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4558 to function value.</p>
4559 </li>
4560 <li>
4561 <p>'<tt>function args</tt>': argument list whose types match the
4562 function signature argument types. All arguments must be of
4563 <a href="#t_firstclass">first class</a> type. If the function signature
4564 indicates the function accepts a variable number of arguments, the extra
4565 arguments can be specified.</p>
4566 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4567 <li>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4568 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4569 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4570 '<tt>readnone</tt>' attributes are valid here.</p>
4571 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4572</ol>
4573
4574<h5>Semantics:</h5>
4575
4576<p>The '<tt>call</tt>' instruction is used to cause control flow to
4577transfer to a specified function, with its incoming arguments bound to
4578the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4579instruction in the called function, control flow continues with the
4580instruction after the function call, and the return value of the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4581function is bound to the result argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4582
4583<h5>Example:</h5>
4584
4585<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4586 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4587 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4588 %X = tail call i32 @foo() <i>; yields i32</i>
4589 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4590 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4591
4592 %struct.A = type { i32, i8 }
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4593 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4594 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4595 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattnerac454b32008-10-08 06:26:11 +0000[diff] [blame]4596 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijman2c4e05a2008-10-07 10:03:45 +0000[diff] [blame]4597 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4598</pre>
4599
4600</div>
4601
4602<!-- _______________________________________________________________________ -->
4603<div class="doc_subsubsection">
4604 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4605</div>
4606
4607<div class="doc_text">
4608
4609<h5>Syntax:</h5>
4610
4611<pre>
4612 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4613</pre>
4614
4615<h5>Overview:</h5>
4616
4617<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4618the "variable argument" area of a function call. It is used to implement the
4619<tt>va_arg</tt> macro in C.</p>
4620
4621<h5>Arguments:</h5>
4622
4623<p>This instruction takes a <tt>va_list*</tt> value and the type of
4624the argument. It returns a value of the specified argument type and
4625increments the <tt>va_list</tt> to point to the next argument. The
4626actual type of <tt>va_list</tt> is target specific.</p>
4627
4628<h5>Semantics:</h5>
4629
4630<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4631type from the specified <tt>va_list</tt> and causes the
4632<tt>va_list</tt> to point to the next argument. For more information,
4633see the variable argument handling <a href="#int_varargs">Intrinsic
4634Functions</a>.</p>
4635
4636<p>It is legal for this instruction to be called in a function which does not
4637take a variable number of arguments, for example, the <tt>vfprintf</tt>
4638function.</p>
4639
4640<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4641href="#intrinsics">intrinsic function</a> because it takes a type as an
4642argument.</p>
4643
4644<h5>Example:</h5>
4645
4646<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4647
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]4648<p>Note that the code generator does not yet fully support va_arg
4649 on many targets. Also, it does not currently support va_arg with
4650 aggregate types on any target.</p>
4651
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4652</div>
4653
4654<!-- *********************************************************************** -->
4655<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4656<!-- *********************************************************************** -->
4657
4658<div class="doc_text">
4659
4660<p>LLVM supports the notion of an "intrinsic function". These functions have
4661well known names and semantics and are required to follow certain restrictions.
4662Overall, these intrinsics represent an extension mechanism for the LLVM
4663language that does not require changing all of the transformations in LLVM when
4664adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4665
4666<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4667prefix is reserved in LLVM for intrinsic names; thus, function names may not
4668begin with this prefix. Intrinsic functions must always be external functions:
4669you cannot define the body of intrinsic functions. Intrinsic functions may
4670only be used in call or invoke instructions: it is illegal to take the address
4671of an intrinsic function. Additionally, because intrinsic functions are part
4672of the LLVM language, it is required if any are added that they be documented
4673here.</p>
4674
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4675<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4676a family of functions that perform the same operation but on different data
4677types. Because LLVM can represent over 8 million different integer types,
4678overloading is used commonly to allow an intrinsic function to operate on any
4679integer type. One or more of the argument types or the result type can be
4680overloaded to accept any integer type. Argument types may also be defined as
4681exactly matching a previous argument's type or the result type. This allows an
4682intrinsic function which accepts multiple arguments, but needs all of them to
4683be of the same type, to only be overloaded with respect to a single argument or
4684the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4685
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4686<p>Overloaded intrinsics will have the names of its overloaded argument types
4687encoded into its function name, each preceded by a period. Only those types
4688which are overloaded result in a name suffix. Arguments whose type is matched
4689against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4690take an integer of any width and returns an integer of exactly the same integer
4691width. This leads to a family of functions such as
4692<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4693Only one type, the return type, is overloaded, and only one type suffix is
4694required. Because the argument's type is matched against the return type, it
4695does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4696
4697<p>To learn how to add an intrinsic function, please see the
4698<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4699</p>
4700
4701</div>
4702
4703<!-- ======================================================================= -->
4704<div class="doc_subsection">
4705 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4706</div>
4707
4708<div class="doc_text">
4709
4710<p>Variable argument support is defined in LLVM with the <a
4711 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4712intrinsic functions. These functions are related to the similarly
4713named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4714
4715<p>All of these functions operate on arguments that use a
4716target-specific value type "<tt>va_list</tt>". The LLVM assembly
4717language reference manual does not define what this type is, so all
4718transformations should be prepared to handle these functions regardless of
4719the type used.</p>
4720
4721<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4722instruction and the variable argument handling intrinsic functions are
4723used.</p>
4724
4725<div class="doc_code">
4726<pre>
4727define i32 @test(i32 %X, ...) {
4728 ; Initialize variable argument processing
4729 %ap = alloca i8*
4730 %ap2 = bitcast i8** %ap to i8*
4731 call void @llvm.va_start(i8* %ap2)
4732
4733 ; Read a single integer argument
4734 %tmp = va_arg i8** %ap, i32
4735
4736 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4737 %aq = alloca i8*
4738 %aq2 = bitcast i8** %aq to i8*
4739 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4740 call void @llvm.va_end(i8* %aq2)
4741
4742 ; Stop processing of arguments.
4743 call void @llvm.va_end(i8* %ap2)
4744 ret i32 %tmp
4745}
4746
4747declare void @llvm.va_start(i8*)
4748declare void @llvm.va_copy(i8*, i8*)
4749declare void @llvm.va_end(i8*)
4750</pre>
4751</div>
4752
4753</div>
4754
4755<!-- _______________________________________________________________________ -->
4756<div class="doc_subsubsection">
4757 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4758</div>
4759
4760
4761<div class="doc_text">
4762<h5>Syntax:</h5>
4763<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4764<h5>Overview:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4765<p>The '<tt>llvm.va_start</tt>' intrinsic initializes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4766<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4767href="#i_va_arg">va_arg</a></tt>.</p>
4768
4769<h5>Arguments:</h5>
4770
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4771<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]4772
4773<h5>Semantics:</h5>
4774
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4775<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]4776macro available in C. In a target-dependent way, it initializes the
4777<tt>va_list</tt> element to which the argument points, so that the next call to
4778<tt>va_arg</tt> will produce the first variable argument passed to the function.
4779Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4780last argument of the function as the compiler can figure that out.</p>
4781
4782</div>
4783
4784<!-- _______________________________________________________________________ -->
4785<div class="doc_subsubsection">
4786 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4787</div>
4788
4789<div class="doc_text">
4790<h5>Syntax:</h5>
4791<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4792<h5>Overview:</h5>
4793
4794<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4795which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4796or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4797
4798<h5>Arguments:</h5>
4799
4800<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4801
4802<h5>Semantics:</h5>
4803
4804<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4805macro available in C. In a target-dependent way, it destroys the
4806<tt>va_list</tt> element to which the argument points. Calls to <a
4807href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4808<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4809<tt>llvm.va_end</tt>.</p>
4810
4811</div>
4812
4813<!-- _______________________________________________________________________ -->
4814<div class="doc_subsubsection">
4815 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4816</div>
4817
4818<div class="doc_text">
4819
4820<h5>Syntax:</h5>
4821
4822<pre>
4823 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4824</pre>
4825
4826<h5>Overview:</h5>
4827
4828<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4829from the source argument list to the destination argument list.</p>
4830
4831<h5>Arguments:</h5>
4832
4833<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4834The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4835
4836
4837<h5>Semantics:</h5>
4838
4839<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4840macro available in C. In a target-dependent way, it copies the source
4841<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4842intrinsic is necessary because the <tt><a href="#int_va_start">
4843llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4844example, memory allocation.</p>
4845
4846</div>
4847
4848<!-- ======================================================================= -->
4849<div class="doc_subsection">
4850 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4851</div>
4852
4853<div class="doc_text">
4854
4855<p>
4856LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4857Collection</a> (GC) requires the implementation and generation of these
4858intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4859These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4860stack</a>, as well as garbage collector implementations that require <a
4861href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4862Front-ends for type-safe garbage collected languages should generate these
4863intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4864href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4865</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4866
4867<p>The garbage collection intrinsics only operate on objects in the generic
4868 address space (address space zero).</p>
4869
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4870</div>
4871
4872<!-- _______________________________________________________________________ -->
4873<div class="doc_subsubsection">
4874 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4875</div>
4876
4877<div class="doc_text">
4878
4879<h5>Syntax:</h5>
4880
4881<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4882 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4883</pre>
4884
4885<h5>Overview:</h5>
4886
4887<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4888the code generator, and allows some metadata to be associated with it.</p>
4889
4890<h5>Arguments:</h5>
4891
4892<p>The first argument specifies the address of a stack object that contains the
4893root pointer. The second pointer (which must be either a constant or a global
4894value address) contains the meta-data to be associated with the root.</p>
4895
4896<h5>Semantics:</h5>
4897
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4898<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]4899location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4900the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4901intrinsic may only be used in a function which <a href="#gc">specifies a GC
4902algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4903
4904</div>
4905
4906
4907<!-- _______________________________________________________________________ -->
4908<div class="doc_subsubsection">
4909 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4910</div>
4911
4912<div class="doc_text">
4913
4914<h5>Syntax:</h5>
4915
4916<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4917 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4918</pre>
4919
4920<h5>Overview:</h5>
4921
4922<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4923locations, allowing garbage collector implementations that require read
4924barriers.</p>
4925
4926<h5>Arguments:</h5>
4927
4928<p>The second argument is the address to read from, which should be an address
4929allocated from the garbage collector. The first object is a pointer to the
4930start of the referenced object, if needed by the language runtime (otherwise
4931null).</p>
4932
4933<h5>Semantics:</h5>
4934
4935<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4936instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4937garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4938may only be used in a function which <a href="#gc">specifies a GC
4939algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4940
4941</div>
4942
4943
4944<!-- _______________________________________________________________________ -->
4945<div class="doc_subsubsection">
4946 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4947</div>
4948
4949<div class="doc_text">
4950
4951<h5>Syntax:</h5>
4952
4953<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4954 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4955</pre>
4956
4957<h5>Overview:</h5>
4958
4959<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4960locations, allowing garbage collector implementations that require write
4961barriers (such as generational or reference counting collectors).</p>
4962
4963<h5>Arguments:</h5>
4964
4965<p>The first argument is the reference to store, the second is the start of the
4966object to store it to, and the third is the address of the field of Obj to
4967store to. If the runtime does not require a pointer to the object, Obj may be
4968null.</p>
4969
4970<h5>Semantics:</h5>
4971
4972<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4973instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4974garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4975may only be used in a function which <a href="#gc">specifies a GC
4976algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4977
4978</div>
4979
4980
4981
4982<!-- ======================================================================= -->
4983<div class="doc_subsection">
4984 <a name="int_codegen">Code Generator Intrinsics</a>
4985</div>
4986
4987<div class="doc_text">
4988<p>
4989These intrinsics are provided by LLVM to expose special features that may only
4990be implemented with code generator support.
4991</p>
4992
4993</div>
4994
4995<!-- _______________________________________________________________________ -->
4996<div class="doc_subsubsection">
4997 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4998</div>
4999
5000<div class="doc_text">
5001
5002<h5>Syntax:</h5>
5003<pre>
5004 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
5005</pre>
5006
5007<h5>Overview:</h5>
5008
5009<p>
5010The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5011target-specific value indicating the return address of the current function
5012or one of its callers.
5013</p>
5014
5015<h5>Arguments:</h5>
5016
5017<p>
5018The argument to this intrinsic indicates which function to return the address
5019for. Zero indicates the calling function, one indicates its caller, etc. The
5020argument is <b>required</b> to be a constant integer value.
5021</p>
5022
5023<h5>Semantics:</h5>
5024
5025<p>
5026The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
5027the return address of the specified call frame, or zero if it cannot be
5028identified. The value returned by this intrinsic is likely to be incorrect or 0
5029for arguments other than zero, so it should only be used for debugging purposes.
5030</p>
5031
5032<p>
5033Note that calling this intrinsic does not prevent function inlining or other
5034aggressive transformations, so the value returned may not be that of the obvious
5035source-language caller.
5036</p>
5037</div>
5038
5039
5040<!-- _______________________________________________________________________ -->
5041<div class="doc_subsubsection">
5042 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5043</div>
5044
5045<div class="doc_text">
5046
5047<h5>Syntax:</h5>
5048<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5049 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5050</pre>
5051
5052<h5>Overview:</h5>
5053
5054<p>
5055The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5056target-specific frame pointer value for the specified stack frame.
5057</p>
5058
5059<h5>Arguments:</h5>
5060
5061<p>
5062The argument to this intrinsic indicates which function to return the frame
5063pointer for. Zero indicates the calling function, one indicates its caller,
5064etc. The argument is <b>required</b> to be a constant integer value.
5065</p>
5066
5067<h5>Semantics:</h5>
5068
5069<p>
5070The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5071the frame address of the specified call frame, or zero if it cannot be
5072identified. The value returned by this intrinsic is likely to be incorrect or 0
5073for arguments other than zero, so it should only be used for debugging purposes.
5074</p>
5075
5076<p>
5077Note that calling this intrinsic does not prevent function inlining or other
5078aggressive transformations, so the value returned may not be that of the obvious
5079source-language caller.
5080</p>
5081</div>
5082
5083<!-- _______________________________________________________________________ -->
5084<div class="doc_subsubsection">
5085 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5086</div>
5087
5088<div class="doc_text">
5089
5090<h5>Syntax:</h5>
5091<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5092 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5093</pre>
5094
5095<h5>Overview:</h5>
5096
5097<p>
5098The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
5099the function stack, for use with <a href="#int_stackrestore">
5100<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5101features like scoped automatic variable sized arrays in C99.
5102</p>
5103
5104<h5>Semantics:</h5>
5105
5106<p>
5107This intrinsic returns a opaque pointer value that can be passed to <a
5108href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
5109<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5110<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5111state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5112practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5113that were allocated after the <tt>llvm.stacksave</tt> was executed.
5114</p>
5115
5116</div>
5117
5118<!-- _______________________________________________________________________ -->
5119<div class="doc_subsubsection">
5120 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5121</div>
5122
5123<div class="doc_text">
5124
5125<h5>Syntax:</h5>
5126<pre>
5127 declare void @llvm.stackrestore(i8 * %ptr)
5128</pre>
5129
5130<h5>Overview:</h5>
5131
5132<p>
5133The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5134the function stack to the state it was in when the corresponding <a
5135href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
5136useful for implementing language features like scoped automatic variable sized
5137arrays in C99.
5138</p>
5139
5140<h5>Semantics:</h5>
5141
5142<p>
5143See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
5144</p>
5145
5146</div>
5147
5148
5149<!-- _______________________________________________________________________ -->
5150<div class="doc_subsubsection">
5151 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5152</div>
5153
5154<div class="doc_text">
5155
5156<h5>Syntax:</h5>
5157<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5158 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]5159</pre>
5160
5161<h5>Overview:</h5>
5162
5163
5164<p>
5165The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5166a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5167no
5168effect on the behavior of the program but can change its performance
5169characteristics.
5170</p>
5171
5172<h5>Arguments:</h5>
5173
5174<p>
5175<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5176determining if the fetch should be for a read (0) or write (1), and
5177<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5178locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5179<tt>locality</tt> arguments must be constant integers.
5180</p>
5181
5182<h5>Semantics:</h5>
5183
5184<p>
5185This intrinsic does not modify the behavior of the program. In particular,
5186prefetches cannot trap and do not produce a value. On targets that support this
5187intrinsic, the prefetch can provide hints to the processor cache for better
5188performance.
5189</p>
5190
5191</div>
5192
5193<!-- _______________________________________________________________________ -->
5194<div class="doc_subsubsection">
5195 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5196</div>
5197
5198<div class="doc_text">
5199
5200<h5>Syntax:</h5>
5201<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5202 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5203</pre>
5204
5205<h5>Overview:</h5>
5206
5207
5208<p>
5209The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5210(PC) in a region of
5211code to simulators and other tools. The method is target specific, but it is
5212expected that the marker will use exported symbols to transmit the PC of the
5213marker.
5214The marker makes no guarantees that it will remain with any specific instruction
5215after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5216optimizations. The intended use is to be inserted after optimizations to allow
5217correlations of simulation runs.
5218</p>
5219
5220<h5>Arguments:</h5>
5221
5222<p>
5223<tt>id</tt> is a numerical id identifying the marker.
5224</p>
5225
5226<h5>Semantics:</h5>
5227
5228<p>
5229This intrinsic does not modify the behavior of the program. Backends that do not
5230support this intrinisic may ignore it.
5231</p>
5232
5233</div>
5234
5235<!-- _______________________________________________________________________ -->
5236<div class="doc_subsubsection">
5237 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5238</div>
5239
5240<div class="doc_text">
5241
5242<h5>Syntax:</h5>
5243<pre>
5244 declare i64 @llvm.readcyclecounter( )
5245</pre>
5246
5247<h5>Overview:</h5>
5248
5249
5250<p>
5251The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5252counter register (or similar low latency, high accuracy clocks) on those targets
5253that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5254As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5255should only be used for small timings.
5256</p>
5257
5258<h5>Semantics:</h5>
5259
5260<p>
5261When directly supported, reading the cycle counter should not modify any memory.
5262Implementations are allowed to either return a application specific value or a
5263system wide value. On backends without support, this is lowered to a constant 0.
5264</p>
5265
5266</div>
5267
5268<!-- ======================================================================= -->
5269<div class="doc_subsection">
5270 <a name="int_libc">Standard C Library Intrinsics</a>
5271</div>
5272
5273<div class="doc_text">
5274<p>
5275LLVM provides intrinsics for a few important standard C library functions.
5276These intrinsics allow source-language front-ends to pass information about the
5277alignment of the pointer arguments to the code generator, providing opportunity
5278for more efficient code generation.
5279</p>
5280
5281</div>
5282
5283<!-- _______________________________________________________________________ -->
5284<div class="doc_subsubsection">
5285 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5286</div>
5287
5288<div class="doc_text">
5289
5290<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5291<p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5292width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5293<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5294 declare void @llvm.memcpy.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5295 i8 &lt;len&gt;, i32 &lt;align&gt;)
5296 declare void @llvm.memcpy.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5297 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5298 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5299 i32 &lt;len&gt;, i32 &lt;align&gt;)
5300 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5301 i64 &lt;len&gt;, i32 &lt;align&gt;)
5302</pre>
5303
5304<h5>Overview:</h5>
5305
5306<p>
5307The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5308location to the destination location.
5309</p>
5310
5311<p>
5312Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5313intrinsics do not return a value, and takes an extra alignment argument.
5314</p>
5315
5316<h5>Arguments:</h5>
5317
5318<p>
5319The first argument is a pointer to the destination, the second is a pointer to
5320the source. The third argument is an integer argument
5321specifying the number of bytes to copy, and the fourth argument is the alignment
5322of the source and destination locations.
5323</p>
5324
5325<p>
5326If the call to this intrinisic has an alignment value that is not 0 or 1, then
5327the caller guarantees that both the source and destination pointers are aligned
5328to that boundary.
5329</p>
5330
5331<h5>Semantics:</h5>
5332
5333<p>
5334The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5335location to the destination location, which are not allowed to overlap. It
5336copies "len" bytes of memory over. If the argument is known to be aligned to
5337some boundary, this can be specified as the fourth argument, otherwise it should
5338be set to 0 or 1.
5339</p>
5340</div>
5341
5342
5343<!-- _______________________________________________________________________ -->
5344<div class="doc_subsubsection">
5345 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5346</div>
5347
5348<div class="doc_text">
5349
5350<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5351<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5352width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5353<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5354 declare void @llvm.memmove.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5355 i8 &lt;len&gt;, i32 &lt;align&gt;)
5356 declare void @llvm.memmove.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5357 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5358 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5359 i32 &lt;len&gt;, i32 &lt;align&gt;)
5360 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5361 i64 &lt;len&gt;, i32 &lt;align&gt;)
5362</pre>
5363
5364<h5>Overview:</h5>
5365
5366<p>
5367The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5368location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5369'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5370</p>
5371
5372<p>
5373Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5374intrinsics do not return a value, and takes an extra alignment argument.
5375</p>
5376
5377<h5>Arguments:</h5>
5378
5379<p>
5380The first argument is a pointer to the destination, the second is a pointer to
5381the source. The third argument is an integer argument
5382specifying the number of bytes to copy, and the fourth argument is the alignment
5383of the source and destination locations.
5384</p>
5385
5386<p>
5387If the call to this intrinisic has an alignment value that is not 0 or 1, then
5388the caller guarantees that the source and destination pointers are aligned to
5389that boundary.
5390</p>
5391
5392<h5>Semantics:</h5>
5393
5394<p>
5395The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5396location to the destination location, which may overlap. It
5397copies "len" bytes of memory over. If the argument is known to be aligned to
5398some boundary, this can be specified as the fourth argument, otherwise it should
5399be set to 0 or 1.
5400</p>
5401</div>
5402
5403
5404<!-- _______________________________________________________________________ -->
5405<div class="doc_subsubsection">
5406 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5407</div>
5408
5409<div class="doc_text">
5410
5411<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5412<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5413width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5414<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5415 declare void @llvm.memset.i8(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5416 i8 &lt;len&gt;, i32 &lt;align&gt;)
5417 declare void @llvm.memset.i16(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5418 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5419 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5420 i32 &lt;len&gt;, i32 &lt;align&gt;)
5421 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5422 i64 &lt;len&gt;, i32 &lt;align&gt;)
5423</pre>
5424
5425<h5>Overview:</h5>
5426
5427<p>
5428The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5429byte value.
5430</p>
5431
5432<p>
5433Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5434does not return a value, and takes an extra alignment argument.
5435</p>
5436
5437<h5>Arguments:</h5>
5438
5439<p>
5440The first argument is a pointer to the destination to fill, the second is the
5441byte value to fill it with, the third argument is an integer
5442argument specifying the number of bytes to fill, and the fourth argument is the
5443known alignment of destination location.
5444</p>
5445
5446<p>
5447If the call to this intrinisic has an alignment value that is not 0 or 1, then
5448the caller guarantees that the destination pointer is aligned to that boundary.
5449</p>
5450
5451<h5>Semantics:</h5>
5452
5453<p>
5454The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5455the
5456destination location. If the argument is known to be aligned to some boundary,
5457this can be specified as the fourth argument, otherwise it should be set to 0 or
54581.
5459</p>
5460</div>
5461
5462
5463<!-- _______________________________________________________________________ -->
5464<div class="doc_subsubsection">
5465 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5466</div>
5467
5468<div class="doc_text">
5469
5470<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5471<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]5472floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5473types however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5474<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5475 declare float @llvm.sqrt.f32(float %Val)
5476 declare double @llvm.sqrt.f64(double %Val)
5477 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5478 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5479 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5480</pre>
5481
5482<h5>Overview:</h5>
5483
5484<p>
5485The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5486returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5487<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5488negative numbers other than -0.0 (which allows for better optimization, because
5489there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5490defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5491</p>
5492
5493<h5>Arguments:</h5>
5494
5495<p>
5496The argument and return value are floating point numbers of the same type.
5497</p>
5498
5499<h5>Semantics:</h5>
5500
5501<p>
5502This function returns the sqrt of the specified operand if it is a nonnegative
5503floating point number.
5504</p>
5505</div>
5506
5507<!-- _______________________________________________________________________ -->
5508<div class="doc_subsubsection">
5509 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5510</div>
5511
5512<div class="doc_text">
5513
5514<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5515<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]5516floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5517types however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5518<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5519 declare float @llvm.powi.f32(float %Val, i32 %power)
5520 declare double @llvm.powi.f64(double %Val, i32 %power)
5521 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5522 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5523 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5524</pre>
5525
5526<h5>Overview:</h5>
5527
5528<p>
5529The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5530specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5531multiplications is not defined. When a vector of floating point type is
5532used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5533</p>
5534
5535<h5>Arguments:</h5>
5536
5537<p>
5538The second argument is an integer power, and the first is a value to raise to
5539that power.
5540</p>
5541
5542<h5>Semantics:</h5>
5543
5544<p>
5545This function returns the first value raised to the second power with an
5546unspecified sequence of rounding operations.</p>
5547</div>
5548
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5549<!-- _______________________________________________________________________ -->
5550<div class="doc_subsubsection">
5551 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5552</div>
5553
5554<div class="doc_text">
5555
5556<h5>Syntax:</h5>
5557<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5558floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5559types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5560<pre>
5561 declare float @llvm.sin.f32(float %Val)
5562 declare double @llvm.sin.f64(double %Val)
5563 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5564 declare fp128 @llvm.sin.f128(fp128 %Val)
5565 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5566</pre>
5567
5568<h5>Overview:</h5>
5569
5570<p>
5571The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5572</p>
5573
5574<h5>Arguments:</h5>
5575
5576<p>
5577The argument and return value are floating point numbers of the same type.
5578</p>
5579
5580<h5>Semantics:</h5>
5581
5582<p>
5583This function returns the sine of the specified operand, returning the
5584same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5585conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5586</div>
5587
5588<!-- _______________________________________________________________________ -->
5589<div class="doc_subsubsection">
5590 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5591</div>
5592
5593<div class="doc_text">
5594
5595<h5>Syntax:</h5>
5596<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5597floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5598types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5599<pre>
5600 declare float @llvm.cos.f32(float %Val)
5601 declare double @llvm.cos.f64(double %Val)
5602 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5603 declare fp128 @llvm.cos.f128(fp128 %Val)
5604 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5605</pre>
5606
5607<h5>Overview:</h5>
5608
5609<p>
5610The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5611</p>
5612
5613<h5>Arguments:</h5>
5614
5615<p>
5616The argument and return value are floating point numbers of the same type.
5617</p>
5618
5619<h5>Semantics:</h5>
5620
5621<p>
5622This function returns the cosine of the specified operand, returning the
5623same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5624conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5625</div>
5626
5627<!-- _______________________________________________________________________ -->
5628<div class="doc_subsubsection">
5629 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5630</div>
5631
5632<div class="doc_text">
5633
5634<h5>Syntax:</h5>
5635<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5636floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5637types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5638<pre>
5639 declare float @llvm.pow.f32(float %Val, float %Power)
5640 declare double @llvm.pow.f64(double %Val, double %Power)
5641 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5642 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5643 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5644</pre>
5645
5646<h5>Overview:</h5>
5647
5648<p>
5649The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5650specified (positive or negative) power.
5651</p>
5652
5653<h5>Arguments:</h5>
5654
5655<p>
5656The second argument is a floating point power, and the first is a value to
5657raise to that power.
5658</p>
5659
5660<h5>Semantics:</h5>
5661
5662<p>
5663This function returns the first value raised to the second power,
5664returning the
5665same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5666conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5667</div>
5668
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5669
5670<!-- ======================================================================= -->
5671<div class="doc_subsection">
5672 <a name="int_manip">Bit Manipulation Intrinsics</a>
5673</div>
5674
5675<div class="doc_text">
5676<p>
5677LLVM provides intrinsics for a few important bit manipulation operations.
5678These allow efficient code generation for some algorithms.
5679</p>
5680
5681</div>
5682
5683<!-- _______________________________________________________________________ -->
5684<div class="doc_subsubsection">
5685 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5686</div>
5687
5688<div class="doc_text">
5689
5690<h5>Syntax:</h5>
5691<p>This is an overloaded intrinsic function. You can use bswap on any integer
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5692type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5693<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5694 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5695 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5696 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5697</pre>
5698
5699<h5>Overview:</h5>
5700
5701<p>
5702The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5703values with an even number of bytes (positive multiple of 16 bits). These are
5704useful for performing operations on data that is not in the target's native
5705byte order.
5706</p>
5707
5708<h5>Semantics:</h5>
5709
5710<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5711The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5712and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5713intrinsic returns an i32 value that has the four bytes of the input i32
5714swapped, 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]5715i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5716<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5717additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5718</p>
5719
5720</div>
5721
5722<!-- _______________________________________________________________________ -->
5723<div class="doc_subsubsection">
5724 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5725</div>
5726
5727<div class="doc_text">
5728
5729<h5>Syntax:</h5>
5730<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]5731width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5732<pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5733 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5734 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5735 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5736 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5737 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5738</pre>
5739
5740<h5>Overview:</h5>
5741
5742<p>
5743The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5744value.
5745</p>
5746
5747<h5>Arguments:</h5>
5748
5749<p>
5750The only argument is the value to be counted. The argument may be of any
5751integer type. The return type must match the argument type.
5752</p>
5753
5754<h5>Semantics:</h5>
5755
5756<p>
5757The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5758</p>
5759</div>
5760
5761<!-- _______________________________________________________________________ -->
5762<div class="doc_subsubsection">
5763 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5764</div>
5765
5766<div class="doc_text">
5767
5768<h5>Syntax:</h5>
5769<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]5770integer bit width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5771<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5772 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5773 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5774 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5775 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5776 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5777</pre>
5778
5779<h5>Overview:</h5>
5780
5781<p>
5782The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5783leading zeros in a variable.
5784</p>
5785
5786<h5>Arguments:</h5>
5787
5788<p>
5789The only argument is the value to be counted. The argument may be of any
5790integer type. The return type must match the argument type.
5791</p>
5792
5793<h5>Semantics:</h5>
5794
5795<p>
5796The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5797in a variable. If the src == 0 then the result is the size in bits of the type
5798of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5799</p>
5800</div>
5801
5802
5803
5804<!-- _______________________________________________________________________ -->
5805<div class="doc_subsubsection">
5806 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5807</div>
5808
5809<div class="doc_text">
5810
5811<h5>Syntax:</h5>
5812<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]5813integer bit width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5814<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5815 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5816 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5817 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5818 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5819 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5820</pre>
5821
5822<h5>Overview:</h5>
5823
5824<p>
5825The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5826trailing zeros.
5827</p>
5828
5829<h5>Arguments:</h5>
5830
5831<p>
5832The only argument is the value to be counted. The argument may be of any
5833integer type. The return type must match the argument type.
5834</p>
5835
5836<h5>Semantics:</h5>
5837
5838<p>
5839The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5840in a variable. If the src == 0 then the result is the size in bits of the type
5841of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5842</p>
5843</div>
5844
5845<!-- _______________________________________________________________________ -->
5846<div class="doc_subsubsection">
5847 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5848</div>
5849
5850<div class="doc_text">
5851
5852<h5>Syntax:</h5>
5853<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5854on any integer bit width.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5855<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5856 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5857 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5858</pre>
5859
5860<h5>Overview:</h5>
5861<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5862range of bits from an integer value and returns them in the same bit width as
5863the original value.</p>
5864
5865<h5>Arguments:</h5>
5866<p>The first argument, <tt>%val</tt> and the result may be integer types of
5867any bit width but they must have the same bit width. The second and third
5868arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5869
5870<h5>Semantics:</h5>
5871<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5872of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5873<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5874operates in forward mode.</p>
5875<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5876right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5877only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5878<ol>
5879 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5880 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5881 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5882 to determine the number of bits to retain.</li>
5883 <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]5884 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5885</ol>
5886<p>In reverse mode, a similar computation is made except that the bits are
5887returned in the reverse order. So, for example, if <tt>X</tt> has the value
5888<tt>i16 0x0ACF (101011001111)</tt> and we apply
5889<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5890<tt>i16 0x0026 (000000100110)</tt>.</p>
5891</div>
5892
5893<div class="doc_subsubsection">
5894 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5895</div>
5896
5897<div class="doc_text">
5898
5899<h5>Syntax:</h5>
5900<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5901on any integer bit width.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5902<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5903 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5904 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]5905</pre>
5906
5907<h5>Overview:</h5>
5908<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5909of bits in an integer value with another integer value. It returns the integer
5910with the replaced bits.</p>
5911
5912<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5913<p>The first argument, <tt>%val</tt>, and the result may be integer types of
5914any 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]5915whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5916integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5917type since they specify only a bit index.</p>
5918
5919<h5>Semantics:</h5>
5920<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5921of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5922<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5923operates in forward mode.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5924
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5925<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5926truncating it down to the size of the replacement area or zero extending it
5927up to that size.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5928
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5929<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5930are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5931in <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]5932to the <tt>%hi</tt>th bit.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5933
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5934<p>In reverse mode, a similar computation is made except that the bits are
5935reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5936<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]5937
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5938<h5>Examples:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5939
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5940<pre>
5941 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5942 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5943 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5944 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5945 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5946</pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5947
5948</div>
5949
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]5950<!-- ======================================================================= -->
5951<div class="doc_subsection">
5952 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
5953</div>
5954
5955<div class="doc_text">
5956<p>
5957LLVM provides intrinsics for some arithmetic with overflow operations.
5958</p>
5959
5960</div>
5961
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5962<!-- _______________________________________________________________________ -->
5963<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]5964 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5965</div>
5966
5967<div class="doc_text">
5968
5969<h5>Syntax:</h5>
5970
5971<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]5972on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5973
5974<pre>
5975 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
5976 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
5977 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
5978</pre>
5979
5980<h5>Overview:</h5>
5981
5982<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
5983a signed addition of the two arguments, and indicate whether an overflow
5984occurred during the signed summation.</p>
5985
5986<h5>Arguments:</h5>
5987
5988<p>The arguments (%a and %b) and the first element of the result structure may
5989be of integer types of any bit width, but they must have the same bit width. The
5990second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
5991and <tt>%b</tt> are the two values that will undergo signed addition.</p>
5992
5993<h5>Semantics:</h5>
5994
5995<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
5996a signed addition of the two variables. They return a structure &mdash; the
5997first element of which is the signed summation, and the second element of which
5998is a bit specifying if the signed summation resulted in an overflow.</p>
5999
6000<h5>Examples:</h5>
6001<pre>
6002 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6003 %sum = extractvalue {i32, i1} %res, 0
6004 %obit = extractvalue {i32, i1} %res, 1
6005 br i1 %obit, label %overflow, label %normal
6006</pre>
6007
6008</div>
6009
6010<!-- _______________________________________________________________________ -->
6011<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6012 <a name="int_uadd_overflow">'<tt>llvm.uadd.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.uadd.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.uadd.with.overflow.i16(i16 %a, i16 %b)
6024 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6025 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6026</pre>
6027
6028<h5>Overview:</h5>
6029
6030<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6031an unsigned addition of the two arguments, and indicate whether a carry occurred
6032during the unsigned 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 unsigned addition.</p>
6040
6041<h5>Semantics:</h5>
6042
6043<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6044an unsigned addition of the two arguments. They return a structure &mdash; the
6045first element of which is the sum, and the second element of which is a bit
6046specifying if the unsigned summation resulted in a carry.</p>
6047
6048<h5>Examples:</h5>
6049<pre>
6050 %res = call {i32, i1} @llvm.uadd.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 %carry, 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_ssub_overflow">'<tt>llvm.ssub.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.ssub.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.ssub.with.overflow.i16(i16 %a, i16 %b)
6072 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6073 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6074</pre>
6075
6076<h5>Overview:</h5>
6077
6078<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6079a signed subtraction of the two arguments, and indicate whether an overflow
6080occurred during the signed subtraction.</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 signed subtraction.</p>
6088
6089<h5>Semantics:</h5>
6090
6091<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6092a signed subtraction of the two arguments. They return a structure &mdash; the
6093first element of which is the subtraction, and the second element of which is a bit
6094specifying if the signed subtraction resulted in an overflow.</p>
6095
6096<h5>Examples:</h5>
6097<pre>
6098 %res = call {i32, i1} @llvm.ssub.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 %overflow, 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_usub_overflow">'<tt>llvm.usub.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.usub.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.usub.with.overflow.i16(i16 %a, i16 %b)
6120 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6121 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6122</pre>
6123
6124<h5>Overview:</h5>
6125
6126<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6127an unsigned subtraction of the two arguments, and indicate whether an overflow
6128occurred during the unsigned 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 unsigned subtraction.</p>
6136
6137<h5>Semantics:</h5>
6138
6139<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6140an unsigned 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 unsigned subtraction resulted in an overflow.</p>
6143
6144<h5>Examples:</h5>
6145<pre>
6146 %res = call {i32, i1} @llvm.usub.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_smul_overflow">'<tt>llvm.smul.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.smul.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.smul.with.overflow.i16(i16 %a, i16 %b)
6168 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6169 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6170</pre>
6171
6172<h5>Overview:</h5>
6173
6174<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6175a signed multiplication of the two arguments, and indicate whether an overflow
6176occurred during the signed multiplication.</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 signed multiplication.</p>
6184
6185<h5>Semantics:</h5>
6186
6187<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6188a signed multiplication of the two arguments. They return a structure &mdash;
6189the first element of which is the multiplication, and the second element of
6190which is a bit specifying if the signed multiplication resulted in an
6191overflow.</p>
6192
6193<h5>Examples:</h5>
6194<pre>
6195 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6196 %sum = extractvalue {i32, i1} %res, 0
6197 %obit = extractvalue {i32, i1} %res, 1
6198 br i1 %obit, label %overflow, label %normal
6199</pre>
6200
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6201</div>
6202
Bill Wendlingbda98b62009-02-08 23:00:09 +0000[diff] [blame]6203<!-- _______________________________________________________________________ -->
6204<div class="doc_subsubsection">
6205 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6206</div>
6207
6208<div class="doc_text">
6209
6210<h5>Syntax:</h5>
6211
6212<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6213on any integer bit width.</p>
6214
6215<pre>
6216 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6217 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6218 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6219</pre>
6220
6221<h5>Overview:</h5>
6222
6223<p><i><b>Warning:</b> '<tt>llvm.umul.with.overflow</tt>' is badly broken. It is
6224actively being fixed, but it should not currently be used!</i></p>
6225
6226<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6227a unsigned multiplication of the two arguments, and indicate whether an overflow
6228occurred during the unsigned multiplication.</p>
6229
6230<h5>Arguments:</h5>
6231
6232<p>The arguments (%a and %b) and the first element of the result structure may
6233be of integer types of any bit width, but they must have the same bit width. The
6234second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6235and <tt>%b</tt> are the two values that will undergo unsigned
6236multiplication.</p>
6237
6238<h5>Semantics:</h5>
6239
6240<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6241an unsigned multiplication of the two arguments. They return a structure &mdash;
6242the first element of which is the multiplication, and the second element of
6243which is a bit specifying if the unsigned multiplication resulted in an
6244overflow.</p>
6245
6246<h5>Examples:</h5>
6247<pre>
6248 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6249 %sum = extractvalue {i32, i1} %res, 0
6250 %obit = extractvalue {i32, i1} %res, 1
6251 br i1 %obit, label %overflow, label %normal
6252</pre>
6253
6254</div>
6255
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6256<!-- ======================================================================= -->
6257<div class="doc_subsection">
6258 <a name="int_debugger">Debugger Intrinsics</a>
6259</div>
6260
6261<div class="doc_text">
6262<p>
6263The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
6264are described in the <a
6265href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
6266Debugging</a> document.
6267</p>
6268</div>
6269
6270
6271<!-- ======================================================================= -->
6272<div class="doc_subsection">
6273 <a name="int_eh">Exception Handling Intrinsics</a>
6274</div>
6275
6276<div class="doc_text">
6277<p> The LLVM exception handling intrinsics (which all start with
6278<tt>llvm.eh.</tt> prefix), are described in the <a
6279href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6280Handling</a> document. </p>
6281</div>
6282
6283<!-- ======================================================================= -->
6284<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6285 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6286</div>
6287
6288<div class="doc_text">
6289<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6290 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6291 the <tt>nest</tt> attribute, from a function. The result is a callable
6292 function pointer lacking the nest parameter - the caller does not need
6293 to provide a value for it. Instead, the value to use is stored in
6294 advance in a "trampoline", a block of memory usually allocated
6295 on the stack, which also contains code to splice the nest value into the
6296 argument list. This is used to implement the GCC nested function address
6297 extension.
6298</p>
6299<p>
6300 For example, if the function is
6301 <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]6302 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]6303<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6304 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6305 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6306 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6307 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6308</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]6309 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6310 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6311</div>
6312
6313<!-- _______________________________________________________________________ -->
6314<div class="doc_subsubsection">
6315 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6316</div>
6317<div class="doc_text">
6318<h5>Syntax:</h5>
6319<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6320declare 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]6321</pre>
6322<h5>Overview:</h5>
6323<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6324 This fills the memory pointed to by <tt>tramp</tt> with code
6325 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6326</p>
6327<h5>Arguments:</h5>
6328<p>
6329 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6330 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
6331 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]6332 intrinsic. Note that the size and the alignment are target-specific - LLVM
6333 currently provides no portable way of determining them, so a front-end that
6334 generates this intrinsic needs to have some target-specific knowledge.
6335 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]6336</p>
6337<h5>Semantics:</h5>
6338<p>
6339 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6340 dependent code, turning it into a function. A pointer to this function is
6341 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6342 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6343 before being called. The new function's signature is the same as that of
6344 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
6345 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
6346 of pointer type. Calling the new function is equivalent to calling
6347 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
6348 missing <tt>nest</tt> argument. If, after calling
6349 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
6350 modified, then the effect of any later call to the returned function pointer is
6351 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6352</p>
6353</div>
6354
6355<!-- ======================================================================= -->
6356<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6357 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6358</div>
6359
6360<div class="doc_text">
6361<p>
6362 These intrinsic functions expand the "universal IR" of LLVM to represent
6363 hardware constructs for atomic operations and memory synchronization. This
6364 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]6365 is aimed at a low enough level to allow any programming models or APIs
6366 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6367 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6368 hardware behavior. Just as hardware provides a "universal IR" for source
6369 languages, it also provides a starting point for developing a "universal"
6370 atomic operation and synchronization IR.
6371</p>
6372<p>
6373 These do <em>not</em> form an API such as high-level threading libraries,
6374 software transaction memory systems, atomic primitives, and intrinsic
6375 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6376 application libraries. The hardware interface provided by LLVM should allow
6377 a clean implementation of all of these APIs and parallel programming models.
6378 No one model or paradigm should be selected above others unless the hardware
6379 itself ubiquitously does so.
6380
6381</p>
6382</div>
6383
6384<!-- _______________________________________________________________________ -->
6385<div class="doc_subsubsection">
6386 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6387</div>
6388<div class="doc_text">
6389<h5>Syntax:</h5>
6390<pre>
6391declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
6392i1 &lt;device&gt; )
6393
6394</pre>
6395<h5>Overview:</h5>
6396<p>
6397 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6398 specific pairs of memory access types.
6399</p>
6400<h5>Arguments:</h5>
6401<p>
6402 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6403 The first four arguments enables a specific barrier as listed below. The fith
6404 argument specifies that the barrier applies to io or device or uncached memory.
6405
6406</p>
6407 <ul>
6408 <li><tt>ll</tt>: load-load barrier</li>
6409 <li><tt>ls</tt>: load-store barrier</li>
6410 <li><tt>sl</tt>: store-load barrier</li>
6411 <li><tt>ss</tt>: store-store barrier</li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6412 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6413 </ul>
6414<h5>Semantics:</h5>
6415<p>
6416 This intrinsic causes the system to enforce some ordering constraints upon
6417 the loads and stores of the program. This barrier does not indicate
6418 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6419 which they occur. For any of the specified pairs of load and store operations
6420 (f.ex. load-load, or store-load), all of the first operations preceding the
6421 barrier will complete before any of the second operations succeeding the
6422 barrier begin. Specifically the semantics for each pairing is as follows:
6423</p>
6424 <ul>
6425 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6426 after the barrier begins.</li>
6427
6428 <li><tt>ls</tt>: All loads before the barrier must complete before any
6429 store after the barrier begins.</li>
6430 <li><tt>ss</tt>: All stores before the barrier must complete before any
6431 store after the barrier begins.</li>
6432 <li><tt>sl</tt>: All stores before the barrier must complete before any
6433 load after the barrier begins.</li>
6434 </ul>
6435<p>
6436 These semantics are applied with a logical "and" behavior when more than one
6437 is enabled in a single memory barrier intrinsic.
6438</p>
6439<p>
6440 Backends may implement stronger barriers than those requested when they do not
6441 support as fine grained a barrier as requested. Some architectures do not
6442 need all types of barriers and on such architectures, these become noops.
6443</p>
6444<h5>Example:</h5>
6445<pre>
6446%ptr = malloc i32
6447 store i32 4, %ptr
6448
6449%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6450 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6451 <i>; guarantee the above finishes</i>
6452 store i32 8, %ptr <i>; before this begins</i>
6453</pre>
6454</div>
6455
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6456<!-- _______________________________________________________________________ -->
6457<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6458 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6459</div>
6460<div class="doc_text">
6461<h5>Syntax:</h5>
6462<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6463 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6464 any integer bit width and for different address spaces. Not all targets
6465 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6466
6467<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6468declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
6469declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
6470declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
6471declare 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]6472
6473</pre>
6474<h5>Overview:</h5>
6475<p>
6476 This loads a value in memory and compares it to a given value. If they are
6477 equal, it stores a new value into the memory.
6478</p>
6479<h5>Arguments:</h5>
6480<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6481 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6482 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6483 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6484 this integer type. While any bit width integer may be used, targets may only
6485 lower representations they support in hardware.
6486
6487</p>
6488<h5>Semantics:</h5>
6489<p>
6490 This entire intrinsic must be executed atomically. It first loads the value
6491 in memory pointed to by <tt>ptr</tt> and compares it with the value
6492 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6493 loaded value is yielded in all cases. This provides the equivalent of an
6494 atomic compare-and-swap operation within the SSA framework.
6495</p>
6496<h5>Examples:</h5>
6497
6498<pre>
6499%ptr = malloc i32
6500 store i32 4, %ptr
6501
6502%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6503%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6504 <i>; yields {i32}:result1 = 4</i>
6505%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6506%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6507
6508%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6509%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6510 <i>; yields {i32}:result2 = 8</i>
6511%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6512
6513%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6514</pre>
6515</div>
6516
6517<!-- _______________________________________________________________________ -->
6518<div class="doc_subsubsection">
6519 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6520</div>
6521<div class="doc_text">
6522<h5>Syntax:</h5>
6523
6524<p>
6525 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6526 integer bit width. Not all targets support all bit widths however.</p>
6527<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6528declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
6529declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
6530declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
6531declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6532
6533</pre>
6534<h5>Overview:</h5>
6535<p>
6536 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6537 the value from memory. It then stores the value in <tt>val</tt> in the memory
6538 at <tt>ptr</tt>.
6539</p>
6540<h5>Arguments:</h5>
6541
6542<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6543 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6544 <tt>val</tt> argument and the result must be integers of the same bit width.
6545 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6546 integer type. The targets may only lower integer representations they
6547 support.
6548</p>
6549<h5>Semantics:</h5>
6550<p>
6551 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6552 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6553 equivalent of an atomic swap operation within the SSA framework.
6554
6555</p>
6556<h5>Examples:</h5>
6557<pre>
6558%ptr = malloc i32
6559 store i32 4, %ptr
6560
6561%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6562%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6563 <i>; yields {i32}:result1 = 4</i>
6564%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6565%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6566
6567%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6568%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6569 <i>; yields {i32}:result2 = 8</i>
6570
6571%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6572%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6573</pre>
6574</div>
6575
6576<!-- _______________________________________________________________________ -->
6577<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6578 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6579
6580</div>
6581<div class="doc_text">
6582<h5>Syntax:</h5>
6583<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6584 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]6585 integer bit width. Not all targets support all bit widths however.</p>
6586<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6587declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6588declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6589declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6590declare 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]6591
6592</pre>
6593<h5>Overview:</h5>
6594<p>
6595 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6596 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6597</p>
6598<h5>Arguments:</h5>
6599<p>
6600
6601 The intrinsic takes two arguments, the first a pointer to an integer value
6602 and the second an integer value. The result is also an integer value. These
6603 integer types can have any bit width, but they must all have the same bit
6604 width. The targets may only lower integer representations they support.
6605</p>
6606<h5>Semantics:</h5>
6607<p>
6608 This intrinsic does a series of operations atomically. It first loads the
6609 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6610 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6611</p>
6612
6613<h5>Examples:</h5>
6614<pre>
6615%ptr = malloc i32
6616 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6617%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6618 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6619%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6620 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6621%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6622 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6623%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6624</pre>
6625</div>
6626
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6627<!-- _______________________________________________________________________ -->
6628<div class="doc_subsubsection">
6629 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6630
6631</div>
6632<div class="doc_text">
6633<h5>Syntax:</h5>
6634<p>
6635 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]6636 any integer bit width and for different address spaces. Not all targets
6637 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6638<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6639declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6640declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6641declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6642declare 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]6643
6644</pre>
6645<h5>Overview:</h5>
6646<p>
6647 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6648 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6649</p>
6650<h5>Arguments:</h5>
6651<p>
6652
6653 The intrinsic takes two arguments, the first a pointer to an integer value
6654 and the second an integer value. The result is also an integer value. These
6655 integer types can have any bit width, but they must all have the same bit
6656 width. The targets may only lower integer representations they support.
6657</p>
6658<h5>Semantics:</h5>
6659<p>
6660 This intrinsic does a series of operations atomically. It first loads the
6661 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6662 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6663</p>
6664
6665<h5>Examples:</h5>
6666<pre>
6667%ptr = malloc i32
6668 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6669%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6670 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6671%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6672 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6673%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6674 <i>; yields {i32}:result3 = 2</i>
6675%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6676</pre>
6677</div>
6678
6679<!-- _______________________________________________________________________ -->
6680<div class="doc_subsubsection">
6681 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6682 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6683 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6684 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6685
6686</div>
6687<div class="doc_text">
6688<h5>Syntax:</h5>
6689<p>
6690 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6691 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6692 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6693 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6694<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6695declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6696declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6697declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6698declare 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]6699
6700</pre>
6701
6702<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6703declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6704declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6705declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6706declare 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]6707
6708</pre>
6709
6710<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6711declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6712declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6713declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6714declare 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]6715
6716</pre>
6717
6718<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6719declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6720declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6721declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6722declare 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]6723
6724</pre>
6725<h5>Overview:</h5>
6726<p>
6727 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6728 the value stored in memory at <tt>ptr</tt>. It yields the original value
6729 at <tt>ptr</tt>.
6730</p>
6731<h5>Arguments:</h5>
6732<p>
6733
6734 These intrinsics take two arguments, the first a pointer to an integer value
6735 and the second an integer value. The result is also an integer value. These
6736 integer types can have any bit width, but they must all have the same bit
6737 width. The targets may only lower integer representations they support.
6738</p>
6739<h5>Semantics:</h5>
6740<p>
6741 These intrinsics does a series of operations atomically. They first load the
6742 value stored at <tt>ptr</tt>. They then do the bitwise operation
6743 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6744 value stored at <tt>ptr</tt>.
6745</p>
6746
6747<h5>Examples:</h5>
6748<pre>
6749%ptr = malloc i32
6750 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6751%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6752 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6753%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6754 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6755%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6756 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6757%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6758 <i>; yields {i32}:result3 = FF</i>
6759%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6760</pre>
6761</div>
6762
6763
6764<!-- _______________________________________________________________________ -->
6765<div class="doc_subsubsection">
6766 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6767 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6768 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6769 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6770
6771</div>
6772<div class="doc_text">
6773<h5>Syntax:</h5>
6774<p>
6775 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6776 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6777 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6778 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6779 support all bit widths however.</p>
6780<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6781declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6782declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6783declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6784declare 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]6785
6786</pre>
6787
6788<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6789declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6790declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6791declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6792declare 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]6793
6794</pre>
6795
6796<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6797declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6798declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6799declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6800declare 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]6801
6802</pre>
6803
6804<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6805declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6806declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6807declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6808declare 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]6809
6810</pre>
6811<h5>Overview:</h5>
6812<p>
6813 These intrinsics takes the signed or unsigned minimum or maximum of
6814 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6815 original value at <tt>ptr</tt>.
6816</p>
6817<h5>Arguments:</h5>
6818<p>
6819
6820 These intrinsics take two arguments, the first a pointer to an integer value
6821 and the second an integer value. The result is also an integer value. These
6822 integer types can have any bit width, but they must all have the same bit
6823 width. The targets may only lower integer representations they support.
6824</p>
6825<h5>Semantics:</h5>
6826<p>
6827 These intrinsics does a series of operations atomically. They first load the
6828 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6829 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6830 the original value stored at <tt>ptr</tt>.
6831</p>
6832
6833<h5>Examples:</h5>
6834<pre>
6835%ptr = malloc i32
6836 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6837%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6838 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6839%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6840 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6841%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6842 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6843%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6844 <i>; yields {i32}:result3 = 8</i>
6845%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6846</pre>
6847</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6848
6849<!-- ======================================================================= -->
6850<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6851 <a name="int_general">General Intrinsics</a>
6852</div>
6853
6854<div class="doc_text">
6855<p> This class of intrinsics is designed to be generic and has
6856no specific purpose. </p>
6857</div>
6858
6859<!-- _______________________________________________________________________ -->
6860<div class="doc_subsubsection">
6861 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6862</div>
6863
6864<div class="doc_text">
6865
6866<h5>Syntax:</h5>
6867<pre>
6868 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6869</pre>
6870
6871<h5>Overview:</h5>
6872
6873<p>
6874The '<tt>llvm.var.annotation</tt>' intrinsic
6875</p>
6876
6877<h5>Arguments:</h5>
6878
6879<p>
6880The first argument is a pointer to a value, the second is a pointer to a
6881global string, the third is a pointer to a global string which is the source
6882file name, and the last argument is the line number.
6883</p>
6884
6885<h5>Semantics:</h5>
6886
6887<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6888This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6889This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6890annotations. These have no other defined use, they are ignored by code
6891generation and optimization.
6892</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6893</div>
6894
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6895<!-- _______________________________________________________________________ -->
6896<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6897 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6898</div>
6899
6900<div class="doc_text">
6901
6902<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6903<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6904any integer bit width.
6905</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6906<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6907 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6908 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6909 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6910 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6911 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]6912</pre>
6913
6914<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6915
6916<p>
6917The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6918</p>
6919
6920<h5>Arguments:</h5>
6921
6922<p>
6923The first argument is an integer value (result of some expression),
6924the second is a pointer to a global string, the third is a pointer to a global
6925string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6926It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6927</p>
6928
6929<h5>Semantics:</h5>
6930
6931<p>
6932This intrinsic allows annotations to be put on arbitrary expressions
6933with arbitrary strings. This can be useful for special purpose optimizations
6934that want to look for these annotations. These have no other defined use, they
6935are ignored by code generation and optimization.
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6936</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6937</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6938
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6939<!-- _______________________________________________________________________ -->
6940<div class="doc_subsubsection">
6941 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6942</div>
6943
6944<div class="doc_text">
6945
6946<h5>Syntax:</h5>
6947<pre>
6948 declare void @llvm.trap()
6949</pre>
6950
6951<h5>Overview:</h5>
6952
6953<p>
6954The '<tt>llvm.trap</tt>' intrinsic
6955</p>
6956
6957<h5>Arguments:</h5>
6958
6959<p>
6960None
6961</p>
6962
6963<h5>Semantics:</h5>
6964
6965<p>
6966This intrinsics is lowered to the target dependent trap instruction. If the
6967target does not have a trap instruction, this intrinsic will be lowered to the
6968call of the abort() function.
6969</p>
6970</div>
6971
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]6972<!-- _______________________________________________________________________ -->
6973<div class="doc_subsubsection">
Misha Brukman5dd7f4d2008-11-22 23:55:29 +0000[diff] [blame]6974 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]6975</div>
6976<div class="doc_text">
6977<h5>Syntax:</h5>
6978<pre>
6979declare void @llvm.stackprotector( i8* &lt;guard&gt;, i8** &lt;slot&gt; )
6980
6981</pre>
6982<h5>Overview:</h5>
6983<p>
6984 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
6985 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
6986 it is placed on the stack before local variables.
6987</p>
6988<h5>Arguments:</h5>
6989<p>
6990 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
6991 first argument is the value loaded from the stack guard
6992 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
6993 has enough space to hold the value of the guard.
6994</p>
6995<h5>Semantics:</h5>
6996<p>
6997 This intrinsic causes the prologue/epilogue inserter to force the position of
6998 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6999 stack. This is to ensure that if a local variable on the stack is overwritten,
7000 it will destroy the value of the guard. When the function exits, the guard on
7001 the stack is checked against the original guard. If they're different, then
7002 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
7003</p>
7004</div>
7005
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7006<!-- *********************************************************************** -->
7007<hr>
7008<address>
7009 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]7010 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7011 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]7012 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]7013
7014 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7015 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7016 Last modified: $Date$
7017</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]7018
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]7019</body>
7020</html>